A Power Assistant Algorithm Based on Human–Robot Interaction Analysis for Improving System Efficiency and Riding Experience Of

sustainability

Article A Power Assistant Algorithm Based on Human–Robot Interaction Analysis for Improving System Efﬁciency and Riding Experience of E-Bikes

Deok Ha Kim , Dongun Lee , Yeongjin Kim, Sungjun Kim and Dongjun Shin *

Human-Centered Robotics Lab, Department of Mechanical Engineering, Chung-Ang University, Seoul 06974, Korea; [email protected] (D.H.K.); [email protected] (D.L.); [email protected] (Y.K.); [email protected] (S.K.) * Correspondence: [email protected]; Tel.: +82-2-820-5072

Abstract: As robots are becoming more accessible in our daily lives, the interest in physical human– robot interaction (HRI) is rapidly increasing. An electric bicycle (E-bike) is one of the best examples of HRI, because a rider simultaneously actuates the rear wheel of the E-bike in close proximity. Most commercially available E-bikes employ a control methodology known as a power assistant system (PAS). However, this type of system cannot offer fully efficient power assistance for E-bikes since it does not account for the biomechanics of riders. In order to address this issue, we propose a control algorithm to increase the efficiency and enhance the riding experience of E-bikes by implementing the control parameters acquired from analyses of human leg kinematics and muscular dynamics. To validate the proposed algorithm, we have evaluated and compared the performance of E-bikes in three different conditions: (1) without power assistance, (2) assistance with a PAS algorithm, and (3) assistance with the proposed algorithm. Our algorithm required 5.09% less human energy consumption than the PAS algorithm and 11.01% less energy consumption than a bicycle operated without power assistance. Our algorithm also increased velocity stability by 11.89% and acceleration Citation: Kim, D.H.; Lee, D.; Kim, Y.; stability by 27.28%, and decreased jerk by 12.36% in comparison to the PAS algorithm. Kim, S.; Shin, D. A Power Assistant Algorithm Based on Human–Robot Keywords: electric bicycles (E-bikes); energy consumption; human analysis; leg kinematics; muscular Interaction Analysis for Improving dynamics; riding experience System Efficiency and Riding Experience of E-Bikes. Sustainability 2021, 13, 768. https://doi.org/ 10.3390/su13020768 1. Introduction Robots are becoming more prevalent in our society, not just in industrial fields but Received: 1 December 2020 Accepted: 6 January 2021 also in our daily lives [1]. Examples of robots in the latter category include robot vacuum Published: 14 January 2021 cleaners, pet robots, service robots, drones, rescue robots, and so on. With easy accessibility to these robots, physical interaction between humans and robots has become a necessity, Publisher’s Note: MDPI stays neu- not an option. tral with regard to jurisdictional clai- Due to increasing demands for physical interaction between human and robots, the ms in published maps and institutio- importance of human–robot interaction (HRI) technology has been gaining attention [2–5]. nal affiliations. In general, HRI can be classified into two categories: remote interaction and proximate interaction [6]. Remote interaction is that in which the human and robot are separated spatially or even temporally, and proximate interaction is that in which the human and robot are co-located. In particular, proximate interaction is closely related to human–robot Copyright: © 2021 by the authors. Li- safety, which is particularly emphasized in transportation on roads, as with electric bicycles censee MDPI, Basel, Switzerland. (E-bikes), autonomous cars, electric kickboards, etc. An E-bike is a system that requires This article is an open access article close interaction between a human and a motor. By utilizing the hybrid actuation of distributed under the terms and con- human muscles and an in-wheel motor, the riding efficiency of cyclists can be enhanced. In ditions of the Creative Commons At- tribution (CC BY) license (https:// addition, an E-bike is an environmentally friendly and healthy alternative to automobiles, creativecommons.org/licenses/by/ making it a promising candidate for HRI transportation technology [7–10]. 4.0/).

Sustainability 2021, 13, 768. https://doi.org/10.3390/su13020768 https://www.mdpi.com/journal/sustainability Sustainability 2021, 13, 768 2 of 18

Various studies have been conducted on how to enhance the HRI of E-bikes in practical applications. Generally, PAS (Power Assistant System) control for E-bikes, as the name suggests, provides motor power assistance in proportion to the pedaling power. Thus, this method cannot assure efficient riding performance. To enhance the riding efficiency of E-bikes, various control methodologies have been introduced. Hatada et al. adopted a PAS to supplement human power during a periodic pedaling motion [11]. This PAS simply provides torque assistance to reduce human energy consumption by reducing the human effort required during cycling. Spagnol et al. developed a strategy for change in energy consumption that was dominated by the exercise protocol, not the crank length [12]. Hull and Gonzalez proposed an optimization method to minimize the required joint moment by optimizing the crank length with respect to the pedal velocity [13]. Danny and Williams also attempted to determine the optimal crank length to investigate the maximum force that the rider can exert and acquire increased crank power [14]. It should be noted that these studies were limited to enhancing the riding efficiency in terms of either human kinematics or motor energy. Greater motor energy is required to compensate for reduced human energy and vice versa. Thus, it is necessary to reduce both human energy consumption and motor energy consumption simultaneously to ensure high riding efficiency. As mentioned above, previous studies have mainly focused on enhancing the riding efficiency of E-bikes with respect to leg kinematics without considering the effect of human muscular dynamics during cycling. Additionally, the human body is a highly complex system consisting of bones as rigid links and muscles as actuators. In order to evaluate the motion of this complex system, inertial measurement units should be used to measure the joint angle [15]. In addition, because muscle force is significantly influenced by variations in muscle length, velocity, and activation, riding efficiency cannot be determined solely from kinematic information. Even if human or motor energy consumption is reduced, it is difficult to ensure that the riding efficiency of E-bikes will be truly enhanced. Thus, it is crucial to consider muscular dynamics to ensure that the riding efficiency of E-bikes (including both human and motor efficiency) has been truly enhanced. Furthermore, it is essential to note that E-bike use should not sacrifice the riding experience for the sake of higher riding efficiency. The higher efficiency of E-bike systems tends to be achieved at the cost of the riding experience, including considerations such as comfortability, safety, and jerk motions, which are typically underrated in the E-bike industry [16]. While previous studies have considered the comfortability (riding velocity stability) and safety (riding acceleration stability) of E-bikes in improving the riding experience [17–19], these studies do not appear to have either taken energy consumption into account or ensured the riding efficiency of E-bikes. To address this trade-off problem, we propose a power assistant algorithm to not only increase the riding efficiency but also improve the riding experiences of E-bikes. The proposed algorithm consists of feedforward and feedback control. The feedback control is a simple PD (Proportional-Derivative) controller based on bike speed. Meanwhile, the feedforward controller is complemented by riding inefficiency, which is obtained from leg kinematics efficiency and muscular dynamics efficiency. Leg kinematics efficiency is calculated by the power transmission ratio from human joint power to crank power. Leg kinematics efficiency is used to determine the energy loss of human joints with respect to the crank angle as a result of the kinematic constraints of human legs. We should also consider muscular dynamics efficiency, which is calculated by the power transmission ratio from joint power with respect to muscle activation. Muscular dynamics efficiency is used to quantify the level of human energy consumption during muscle force generation at a specific crank angle. By considering human muscle parameters from both the leg kinematics and muscular dynamics perspectives, we can achieve enhanced riding efficiency while improving riding experience, by complementing riding inefficiency. To build the proposed power assistant algorithm, we analyzed leg kinematics and muscular dynamics. A simulation environment was developed in MATLAB/Simulink (Mathworks, Inc.) using OpenSim API. To validate the concept, we built an E-bike testbed Sustainability 2021, 13, x FOR PEER REVIEW 3 of 19

Sustainability 2021, 13, 768 (Mathworks, Inc.) using OpenSim API. To validate the concept, we built an E3-bike of 18 testbed using an in-wheel motor and a real-time controller. Finally, we conducted experiments to evaluate the net metabolic cost, riding experience (by considering the riding velocity and accelerationusing an in-wheel stability), motor and and ajerk real-time associated controller. with Finally,the use we of conductedour proposed experiments algorithm to versus theevaluate PAS thealgorithm net metabolic and E cost,-bike riding operation experience without (by consideringa motor. the riding velocity and accelerationSection stability), 2 describes and the jerk analysis associated of withhuman the usedynamics of our proposed performed algorithm with respect versus to mus- cularthe PAS dynamics algorithm and and leg E-bike kinematics. operation Section without a3 motor.explains the novel algorithm based on the analysisSection of 2Section describes 2. E thexperiments analysis of have human been dynamics conducted performed between with the respect proposed to mus- algorithm cular dynamics and leg kinematics. Section3 explains the novel algorithm based on the and the PAS algorithm, and Section 4 summarizes the results, and Section 5 presents the analysis of Section2. Experiments have been conducted between the proposed algorithm discussionand the PAS and algorithm, conclusions. and Section 4 summarizes the results, and Section5 presents the discussion and conclusions. 2. Human Analysis for E-Bikes 2. Human Analysis for E-Bikes As mentioned previously, we define the E-bike as an HRI system consisting of a rider and aAs bike mentioned (Figure previously, 1). A human we deﬁne body theexhibits E-bike highly as an HRI nonlinear system consistingdynamics of due a rider to its com- and a bike (Figure1). A human body exhibits highly nonlinear dynamics due to its complex plex articulations and muscular mechanics. Thus, to achieve mechanical harmony in a articulations and muscular mechanics. Thus, to achieve mechanical harmony in a system systemcombining combining a human a and human a robot, and human a robot, analysis human is essential analysis to is minimizing essential to interference minimizing inter- ferencebetween between a rider and a arider bike. and As a a result, bike. weAs designeda result, awe novel designed algorithm a novel to control algorithm the motor to control theof the motor E-bike of to the ensure E-bike high to efﬁciencyensure high and efficiency provide a good and ridingprovide experience. a good riding experience.

Figure 1.1.Illustration Illustration of of human–robot human–robot interaction interaction system system for an electricfor an electric bicycle. Thebicycle. electric The bicycle electric is bicycle isdriven driven by by human human cycling cycling and anand electric an electric motor motor in parallel. in parallel. In this section, we analyze the human harmonic motion required for the highly efficient In this section, we analyze the human harmonic motion required for the highly effi- cycling of an E-bike. The motion of the rider can be analyzed in terms of two biomechanical cientfeatures: cycling (1) joint of an kinematics E-bike. The and motion (2) muscular of the dynamics.rider can be Throughout analyzed thein terms analysis, of two we biome- chanicalemployed features: a “biomechanical (1) joint efficiency”kinematics parameter and (2) muscular to compare dynamics. the kinematic Throughout and dynamic the analysis,characteristics we employed using a the “biomechanical same dimension. efficiency” parameter to compare the kinematic and dynamic characteristics using the same dimension. 2.1. Leg Kinematics Efficiency 2.1. LegIn this Kinematics section, weEfficiency define the leg kinematics efficiency, which represents the transmission efficiency of human legs in terms of crank output at varying angles. The leg kinematics efficiencyIn this isrepresented section, we by define the ratio the of leg the kinematics crank power efficiency, to the hip, knee,which and represents ankle power the trans- missionof a person efficiency riding an of E-bike. human The legs leg kinematicsin terms of efficiency crank output can be at expressed varying in angles. the form Th ofe leg kin- ematicsthe equation efficiency below. is represented by the ratio of the crank power to the hip, knee, and ankle power of a person riding an E-bike. The legPcrank kinematics efficiency can be expressed in the ηLK = (1)

where ηLK is the leg kinematics efﬁciency during the pedaling motion. Pcrank, Phip, Pknee, and Pankle represent the power measured at the crank, hip joint, knee joint, and ankle joint, respectively. Each joint power is calculated by taking the absolute value of the eccentric motion. The eccentric motion is represented as a negative power when generating a muscle force to adjust the joint stiffness. Generally, negative power can be treated as power gain (or energy reservation) during pedaling, but this is not true for humans from a biomechanics perspective. Since a human exerts certain levels of muscle force to maintain the leg stiffness to hold the pedal in the desired position, negative power means that human muscle is using energy. To calculate the joint power in each joint, we used a Jacobian matrix of a human leg to obtain the joint torque and velocity at human joints. Figure2 illustrates the conﬁguration of each joint and the corresponding parameters. The joint torques can be calculated as the sums of the joint motion torques and external load torques, as shown below:

τJoint = τMotion + τext (2)

T τJoint is represented in vector form as [τhip, τknee, τankle] with respect to the joint motion torque, τMotion, and the external load torque, τext. τMotion can be expressed in the form of Equation (3). .. . . τMotion = A(q)q + B q, q q + G(q) (3) A(q) = Σ m JT J + JT I J (i = hip, knee, ankle) (4) i legi legi ωi i ωi

 . T δA .  q δq q . hip . .  . T .  B q, q = Aq −  q δA q  (5)  δqknee  . T . q δA q δqankle G(q) = −ΣJT m g (i = hip, knee, ankle) (6) legi i   ∂xhip ∂xknee ∂xankle  ∂qhip ∂qknee ∂qankle   ∂yhip ∂yknee ∂yankle     ∂qhip ∂qknee ∂qankle   ∂zhip ∂z ∂z  Jleg  knee ankle  J = =  ∂qhip ∂qknee ∂qankle  (7) J   ω  − − − − − − − − −     0 0 0     0 0 0  1 1 1 . In this Equation (3), A(q) represents the inertial matrix, Bq, q represents the Coriolis and centrifugal matrix, and G(q) represents the gravitational vector. These terms are calculated by Equations (4)–(6). In Equations (4)–(6), mi is scalar mass, and J is the Jacobian

matrix containing Jleg and Jω. Jlegi is the Jacobian column vector of Jleg, Ii is the inertia T matrix of legs, and Jωi is the rotational Jacobian of legs, Jωi = [0, 0, 1] . Note that these h iT terms are functions of the joint angle vectors q = qhip qknee qankle and joint velocity vectors . h . . . iT q = qhip qknee qankle . The second term, τext, can be calculated from the pedal reaction force and the Jacobian matrix (Equation (7)) of human legs. τext is expressed in the form of Equation (8).

T τext = JlegFext (8)

where xi, yi, zi, qi (i = hip, knee, ankle) are the position of the hip, knee, ankle, and corresponding joint angles, Fext is the pedal reaction force and Jleg is the Jacobian matrix of the human leg, as shown in Figure2. Sustainability 2021, 13, x FOR PEER REVIEW 5 of 19

where 푥𝑖, 푦𝑖, 푧𝑖, 푞𝑖 (𝑖 = ℎ𝑖푝, 푘푛푒푒, 푎푛푘푙푒) are the position of the hip, knee, ankle, and corre- Sustainability 2021, 13, 768 5 of 18 sponding joint angles, 퐹푒푥푡 is the pedal reaction force and 퐽푙푒𝑔 is the Jacobian matrix of the human leg, as shown in Figure 2.

(a) (b)

Figure 2.Figure Illustration 2. Illustration of joint parameters of joint parameters and its configuration and its conﬁguration with respect with to respect (a) the a tongle (a) theof each angle of each joint, andjoint, (b) the and torque (b) the applied torque to applied each joint to each and joint mass/inertia. and mass/inertia.

By insertingBy inserting the data the acquired data acquired during pedaling during pedaling (i.e., pedaling (i.e., pedaling force, pedaling force, pedaling angle, angle, and pedalingand pedaling velocity) velocity)into Equation into Equation(8), we can (8), calculate we can the calculate joint torques the joint of the torques hip, knee, of the hip, and ankle.knee, The and calculated ankle. Thejoint calculatedtorques are joint then torques used to are derive then the used joint to power, derive the푃퐽표𝑖푛푡 joint= power, [푃 , 푃 , 푃 h ], as shown ini Equation (9). ℎ𝑖푝 푘푛푒푒PJoint =푎푛푘푙푒Phip, Pknee, Pankle , as shown in Equation (9). 푃 = 휏 ∙ 푞̇ 퐽표𝑖푛푡 퐽표𝑖푛푡 . (9) PJoint = τJoint·q (9) To calculate the leg kinematics efficiency during the pedaling motion, we must first analyze the pedalTo calculate force output. the leg Because kinematics the angle efficiency of the during pedal thechanges pedaling continuously motion, we with must first respect analyzeto the crank the pedalangle, force the magnitude output. Because of the the pedal angle force of theoutput pedal in changes the tangential continuously and with radial directionsrespect to varies the crank constantly. angle, The the magnituderesulting tangential of the pedal and forceradial output force exerted in the tangential on the and single sideradial of directionsthe pedal variesduring constantly. pedaling, Theacquired resulting from tangential the sensory and data, radial are force shown exerted in on the Figure 3.single The sensory side of thedata pedal (pedal during force) pedaling,were collected acquired from from a loadcell the sensory placed data, within are the shown in pedal, whileFigure riding3. The the sensory bike at data 15 km/h (pedal and force) with were constant collected load from friction. a loadcell The bike placed saddle within the was placedpedal, at 0. while25 m ridinghorizontally the bike and at 0.70 15 km/h m vertically and with from constant the crank. load The friction. magnitude The bike of saddle the pedalwas force placed depends at 0.25 on m the horizontally 퐹푒푥푡 and the and tilt 0.70 angle m verticallyof the pedal. from It theshould crank. be Thenoted magnitude that of the pedalthe force pedal exerted force dependsin the tangential on the F extdirectionand the is tilt the angle dominant of the factor pedal. that It should determines be noted that the crankthe power. pedal Thus, force exertedone may in assume the tangential that the direction leg kinematics is the dominantefficiency factoris highest that near determines a crank theangle crank of 90 power.°. This Thus, would, one however, may assume not thatbe entirely the leg kinematicstrue without efficiency considering is highest the near a ◦ force inputcrank at anglehuman of joints. 90 . This would, however, not be entirely true without considering the force input at human joints. Therefore, it is necessary to consider both the force output and the force input to derive the leg kinematics efficiency. The force data can be acquired by riding a normal bicycle without assistance, which is then computed into the OpenSim API. Then, the necessary parameters acquired via OpenSim API and the sensory data acquired to depict Figure3 were computed into Equations (1)–(6) to derive the leg kinematics efficiency. The results are shown in Figure4a. In contrast to Figure3, we can observe that high leg kinematics efficiency is observed at crank angles of 164◦ and 326◦. This is because both the hip and the knee are extended at 164◦ and 326◦, resulting in legs with large moment arms. Because the legs function to convert joint power into crank power, the highest leg kinematics efficiency can be acquired with the largest moment arm. On the other hand, a low leg kinematics Sustainability 2021, 13, x FOR PEER REVIEW 6 of 19

Sustainability 2021, 13, 768 6 of 18

efﬁciency is observed at crank angles of 23◦ and 217◦. This is because of the antagonistic characteristics of pedaling. While one leg is exerting high pedaling force, the opposing Sustainability 2021, 13, x FOR PEER REVIEW 6 of 19 leg exerts a minimal pedaling force to rotate the crank position only. Thus, a subsequent decrease in leg kinematic efﬁciency is observed, as shown in Figure4a.

Figure 3. Tangential and radial pedaling force of right foot with respect to the crank angle. Blue points and lines indicate tangential force. Red points and lines indicate radial force.

Therefore, it is necessary to consider both the force output and the force input to derive the leg kinematics efficiency. The force data can be acquired by riding a normal bicycle without assistance, which is then computed into the OpenSim API. Then, the necessary parameters acquired via OpenSim API and the sensory data acquired to depict Figure 3 were computed into Equations (1)–(6) to derive the leg kinematics efficiency. The results are shown in Figure 4a. In contrast to Figure 3, we can observe that high leg kinematics efficiency is observed at crank angles of 164° and 326°. This is because both the hip and the knee are extended at 164° and 326°, resulting in legs with large moment arms. Because the legs function to convert joint power into crank power, the highest leg kinematics efficiency can be acquired with the largest moment arm. On the other hand, a low leg kinematics efficiency is observed at crank angles of 23° and 217°. This is because of the antagonistic characteristics of pedaling. While one leg is exerting high pedaling force, the op- Figure 3. TangentialposingFigure 3.leg andTangential exerts radial a pedaling minimal and radial force pedaling pedaling of right force foot force withto of rotate right respect footthe to crank with the crank respect posit angle.ion to theonly. Blue crank Thus, angle. a sub- Blue points and linessequentpoints indicate and decrease lines tangential indicate in leg force. tangentialkinematic Red points force. efficiency and Red lines points is observed,indicate and lines radial as indicate shown force. radial in Figure force. 4a.

Therefore, it is necessary to consider both the force output and the force input to derive the leg kinematics efficiency. The force data can be acquired by riding a normal bicycle without assistance, which is then computed into the OpenSim API. Then, the necessary parameters acquired via OpenSim API and the sensory data acquired to depict Figure 3 were computed into Equations (1)–(6) to derive the leg kinematics efficiency. The results are shown in Figure 4a. In contrast to Figure 3, we can observe that high leg kinematics efficiency is observed at crank angles of 164° and 326°. This is because both the hip and the knee are extended at 164° and 326°, resulting in legs with large moment arms. Because the legs function to convert joint power into crank power, the highest leg kinematics efficiency can be acquired with the largest moment arm. On the other hand, a low leg kinematics efficiency is observed at crank angles of 23° and 217°. This is because of the antagonistic characteristics of pedaling. While one leg is exerting high pedaling force, the opposing leg exerts a minimal pedaling force to rotate the crank position only. Thus, a sub-

sequent decrease in leg kinematic(a) efficiency is observed, as shown in Figure(b 4a.)

FigureFigure 4. 4. ComparisonComparison of of riding riding efficiency efficiency based based on on human human kinematics: kinematics: (a ()a Leg) Leg kinematics kinematics effi- efficiency ciency with respect to the crank angle; (b) Muscular dynamics efficiency according to the crank with respect to the crank angle; (b) Muscular dynamics efficiency according to the crank angle. angle. 2.2. Muscular Dynamics Efficiency Previously, we analyzed the riding efficiency based on human leg kinematics only. However, to examine human inefficiency during pedaling motion, it is crucial to consider human muscular dynamics. To do so, we must analyze how much energy human muscles consume to produce active muscle force. According to Hill’s model, active muscle force is determined by muscle length, velocity, and activation [20]. First of all, the muscle capability required to produce the maximum force is observed at the no-contraction state and gradually decreases as the muscle is shortened or lengthened. As shown in Figure5, the muscle force tendency exhibits a symmetrical shape as muscles shorten and lengthen, with the highest muscle force occurring at a normalized muscle

(a) (b) Figure 4. Comparison of riding efficiency based on human kinematics: (a) Leg kinematics efficiency with respect to the crank angle; (b) Muscular dynamics efficiency according to the crank angle. Sustainability 2021, 13, x FOR PEER REVIEW 7 of 19

2.2. Muscular Dynamics Efficiency Previously, we analyzed the riding efficiency based on human leg kinematics only. However, to examine human inefficiency during pedaling motion, it is crucial to consider human muscular dynamics. To do so, we must analyze how much energy human muscles consume to produce active muscle force. According to Hill’s model, active muscle force is determined by muscle length, velocity, and activation [20]. First of all, the muscle capability required to produce the maximum force is observed Sustainability 2021, 13, 768 7 of 18 at the no-contraction state and gradually decreases as the muscle is shortened or lengthened. As shown in Figure 5, the muscle force tendency exhibits a symmetrical shape as muscles shorten and lengthen, with the highest muscle force occurring at a normalized lengthmuscle of length 1. For of example, 1. For example, the isometric the isometric length length (for a normalized(for a normalized length length of (1)) of of (1) the) of quadricepsthe quadriceps is observed is observed when when the kneethe knee joint joint angle angle is close is close to −to30 −30◦ [°21 [2].1]. This This is is the the range range withinwithin which which the the crank crank angle angle changes changes from from 60 60◦ °to to 120 120◦,°, and and in in this this range, range, the the quadriceps quadriceps cancan exertexert thethe highest muscle force. force. This This implies implies that that the the capability capability of each of each muscle muscle to pro- to produceduce its itsmaximum maximum muscle muscle force force depends depends on on the the crank crank angle. angle.

(a) (b)

FigureFigure 5. 5.Change Change in in muscle muscle force force while while shortening shortening and and lengthening lengthening right right quadriceps quadriceps during during pedaling ped- motionaling motion (a). Change (a). Change in muscle in muscle force during force during muscle muscle shortening shortening and lengthening and lengthening (b). (b).

Second,Second, wewe mustmust alsoalso consider the muscle muscle velocity, velocity, which which is is another another crucial crucial factor factor in inthe the accurate accurate evaluation evaluation of of the the muscle muscle force. force. Even Even when when the the muscle muscle length length and and muscle muscle ac- activationtivation are are the the same, same, the the muscle muscle force can vary withwith respectrespect toto thethe normalized normalized muscle muscle velocity,velocity, as as shown shown in in Figure Figure5. 5. The The normalized normalized muscle muscle force force exhibits exhibits a sigmoida sigmoid shape shape with with thethe center center positioned positioned at at a a muscle muscle velocity velocity of of 0. 0. During During muscle muscle shortening, shortening, muscles muscles contract contract toto generate generate muscle muscle force. force. Greater Greater muscle muscle force force can can be be generated generated when when the the muscle muscle velocity velocity isis low low (when (when muscles muscles contract contract slowly). slowly). When When the the muscle muscle velocity velocity is is high high (when (when muscles muscles contractcontract quickly), quickly), the the muscle muscle force force decreases. decreases. Muscle Muscle lengthening lengthening occurs occurs when when a a muscle muscle isis elongated elongated by by its its antagonistic antagonistic pair. pair. Unlike Unlike muscle muscle shortening, shortening, greater greater muscle muscle force force can can bebe generated generated whenwhen the muscle velocity velocity is is high high (when (when muscles muscles lengthen lengthen quickly). quickly). This This im- impliesplies that that the the muscle muscle force force can can vary vary with with respect respect to pedaling velocity, eveneven forfor the the same same crankcrank angle. angle. Finally,Finally, it it isis necessary to to consider consider muscle muscle activation activation to toevaluate evaluate the the muscle muscle force force and andenergy energy consumption. consumption. Basically, Basically, human human muscles muscles are activated are activated when when the brain the brainsends sends an ex- an excitation signal to the muscles [21]. In general, both the muscle force and energy consumption increase in proportion to the muscle activation. This tendency can be observed in Figure5, where the normalized muscle force increases with respect to the muscle activation. This implies that a human’s capability to exert muscle force, along with energy consumption, can vary (even for the same crank position and velocity) depending on the level of muscle activation. The muscle activation can be expressed in the form of Equation (10). Fmuscle,current aMuscle = (10) Fmuscle,max

where Fmuscle,current represents the current muscle force being exerted by the human and the Fmuscle,max is the maximum force that the human is capable of exerting. Sustainability 2021, 13, 768 8 of 18

In summary, muscle force can vary with respect to pedaling velocity, even for the same crank angle, because of the muscle velocity. The muscle force may also vary for a given crank angle and pedaling velocity because of muscle activation. Among the three muscle parameters, muscle activation is the key factor that is directly related to both muscle force and energy consumption [19]. Once the muscles are activated, both thermal energy and mechanical energy are required to contract the muscles. When the muscles contract, the hip, knee, and ankle move to produce joint power. As a result, the muscular dynamics efﬁciency can be represented in the form of the following equation:

∑ Phip + |Pknee| + |Pankle| ηMD = (11) ∑ aMuscle

where ∑ aMuscle is the sum of the muscle activation during pedaling. After calculating the ∑ aMuscle using the musculoskeletal dynamics simulator OpenSim (by Simbios, Stanford, CA, USA) and MATLAB (Mathworks Inc., Natick, MA, USA), we simulated the muscular dynamics efficiency, as shown in Figure4b [ 22]. When riding the E-bike, the activations of the hip, knee, and ankle flexor and extensor groups are used to derive the ∑ aMuscle [23,24], which are determined by the static optimization method in OpenSim. In this simulation, the most recent Lai Arnold Wakeling model was adopted for use in performing the musculoskeletal analysis [25]. Figure4b shows ηMD with respect to the corresponding crank angle. The muscular dynamics efficiency exhibits a wave form similar to that of the leg kinematics efficiency, but exhibits a different phase. This is because human muscles are coupled to one another when executing a pedaling motion. For instance, the quadriceps and hamstring are the large muscles that play a key role in executing the pedaling motion. To execute the pedaling motion, these muscles act antagonistically to move not only the hip joint but also the knee joint. Thus, the phase of the muscular dynamics may differ as a result of the activation of coupled muscles, regardless of the leg kinematics.

2.3. Overall Riding Efficiency Analysis In the previous section, the leg kinematics and the muscular dynamics efficiencies involved in the pedaling motion, which are closely related to one another, were described. The activation process enables muscles to contract and generate joint power, which is then converted to crank power. Based on the correlation between leg kinematics and muscular dynamics, a human’s overall riding efficiency ηR,Bike can be expressed as a multiplication function of the leg kinematics efficiency ηHK and the muscular dynamics efficiency ηMD, as expressed in the form of Equation (12).

Pcrank ηR,Bike = ηMD·ηHK = (12) ∑ aMuscle

The sum of the muscle activation ∑ aMuscle is defined as the input, while the crank power Pcrank is defined as the output. As mentioned in Section 2.2, the leg kinematics efficiency, and the muscular dynamics efficiency, although closely related to one another, exhibit different tendencies. There are two factors that determine the maximum force that the rider can exert at any crank angle: (1) the crank angle and (2) the cycling velocity. First, the length of the muscle is affected by the change in the crank angle. Second, the contraction speed of the muscle is affected by the change in cycling velocity. Thus, the maximum muscle force that the rider can exert at any crank angle changes continuously. The resulting leg kinematics efficiency and muscular dynamics efficiency, acquired previously in Figure4, are illustrated in Figure6 in comparison to the overall riding efficiency. Due to their distinct characteristics, there is a clear difference between the two efficiencies, as shown in Figure6a. The overall riding efficiency (the product of the leg kinematics and muscular dynamics efficiencies) is represented by the black line, and the leg kinematics Sustainability 2021, 13, x FOR PEER REVIEW 9 of 19

are two factors that determine the maximum force that the rider can exert at any crank angle: (1) the crank angle and (2) the cycling velocity. First, the length of the muscle is affected by the change in the crank angle. Second, the contraction speed of the muscle is affected by the change in cycling velocity. Thus, the maximum muscle force that the rider can exert at any crank angle changes continuously. The resulting leg kinematics efficiency and muscular dynamics efficiency, acquired previously in Figure 4, are illustrated in Fig- Sustainability 2021, 13, 768 ure 6 in comparison to the overall riding efficiency. 9 of 18 Due to their distinct characteristics, there is a clear difference between the two efficiencies, as shown in Figure 6a. The overall riding efficiency (the product of the leg kinematics and muscular dynamics efficiencies) is represented by the black line, and the leg kinematicsefficiency iseffi representedciency is represented by the blue by line. the Theblue overallline. The bike overall efficiency bike efficiency exhibits aexhibits tendency a tendencythat is similar that is to similar that of to the that muscular of the muscular dynamics dynamics efficiency efficiency shown in shown Figure in6 b,Figure but very 6b, butdifferent very different from the from kinematic the kinematic efficiency. efficiency. This tendency This tendency shows that shows the muscular that the muscular dynamics dynamicsefficiency efficiency has a more has dominant a more dominant effect on theeffect overall on the riding overall efficiency. riding efficiency. Thus, it is Thus, evident it isthat evident we must that consider we must not consider only the not leg only kinematics, the leg kinematics but also the, but muscular also the dynamics muscular when dy- namicsanalyzing whe cyclingn analyzing activity. cycling activity.

(a) (b)

FigureFigure 6. 6. PhasePhase comparison comparison of two of two E-bike E-bike efficiencies: efficiencies: (a) comparison (a) comparison between between leg kinematic leg kinematic effi- ciencyefficiency and andoverall overall efficiency; efficiency; (b) comparison (b) comparison between between the muscular the muscular dynamics dynamics efficiency efficiency and over- and alloverall efficiency. efficiency.

BecauseBecause the the overall overall riding riding efficiency efficiency considers considers both both the the leg leg kinematics kinematics eff efficiencyiciency and and thethe muscular muscular dynamics dynamics efficiency, efficiency, it it can can be be utilized utilized to to compensate compensate for for the the human human ineffi- ineffi- ciencyciency during the pedalingpedaling motion.motion. ByBy takingtaking the the inverse inverse of of the the overall overall riding riding efficiency, efficiency, we wecan can derive derive the the pedaling pedaling inefficiency, inefficiency,IR,Bike 퐼푅,퐵𝑖푘푒= 1/= η1/R휂,Bike푅,퐵𝑖푘푒, as, as shown shown in in Figure Figure7. In7. In contrast contrast to tothe the leg leg kinematics kinematics efficiency efficiency (high (high efficiency efficiency at 164 at◦ 164and° and 326◦ )326 and°) theand muscular the muscular dynamics dynamicsefficiency efficiency (high efficiency (high efficiency at 125◦ andat 125 300° and◦), a 300 high°), overalla high overall riding efficiencyriding efficiency is observed is ob- at servedcrank anglesat crank of angles 103◦ and of 103 284° ◦and. The 284 lowest°. The lowest overall overall riding riding efficiency efficiency is observed is observed at crank at crankangles angles of 32 ◦ofand 32° 197and◦ 197, while°, while the the leg leg kinematics kinematics efficiency efficiency and and the the muscular muscular dynamics dynamicsefficiency efficiency were were lowest lowest at 23 at◦ 23and° and 217 217◦ and° and 0◦ and0° and 188 188◦, respectively.°, respectively. Thus, Thus, it is it evident is evident that thatthe humanthe human inefficiency inefficiency cannot cannot be solely be solely determined determined from the from leg the kinematics leg kinematics alone, butalone also, butdepends also depends on the muscular on the muscular dynamics. dynamics. Note that Note the thatηR,Bike thecan 휂푅,퐵𝑖푘푒 be calculated can be calculated independently inde- pendentlyfrom the cycling from the velocity, cycling as velocity, long as weas knowlong as the we muscle know activationthe muscle and activation the crank and power. the crankIR,Bike power.represents 퐼푅,퐵𝑖푘푒 the represents magnitude the of musclemagnitude activation of muscle with activation respect to with a unit respect of crank to a power. unit ofBecause crank power. muscle Because activation muscle serves ac astivation an indicator serves when as an evaluatingindicator whe humann evaluating energy consump- human energytion [20 consumption], a greater IR ,[Bike20],indicates a greater that퐼푅,퐵𝑖푘푒 the indicates rider has that consumed the rider more has energyconsumed in exerting more en- an ergyequal in muscle exerting force. an equal Therefore, muscle we force. can visualize Therefore, the we riding can efficiencyvisualize andthe riding inefficiency efficiency from anda human inefficienc metabolicy from perspective a human metabolic rather than perspective a kinematics rather perspective. than a kinematics perspective. Sustainability 2021, 13, x FOR PEER REVIEW 10 of 19 Sustainability 2021, 13, 768 10 of 18

Figure 7. Human energy cost rate for equivalent cycling power, IR,Bike, according to the crank angle. Figure 7. Human energy cost rate for equivalent cycling power, IR,Bike, according to the crank angle. Cycling inefficiencyCycling reflects inefficiency both leg reflects kinematics both leg and kinematics muscular anddynamics. muscular The dynamics. black line Thewas blackfitted line was fitted using a fifth-order Fourier transformation. using a fifth-order Fourier transformation.

3. E-Bike Control3. E-Bike Based Control on Human Based Inefficiency on HumanInefficiency The E-bike isThe a system E-bike that is a requires system the that hybrid requires actuation the hybrid of human actuation muscle of humanand an musclein- and an wheel motorin-wheel to rotate motor a single to rotate-degree a- single-degree-of-freedomof-freedom (1-DoF) wheel. (1-DoF) In order wheel. to do In so, order the to do so, the two actuatorstwo must actuators work alongside must work one alongside another onein parallel another to inminimize parallel tocross minimize interference. cross interference. For instance,For if an instance, in-wheel if anmotor in-wheel exerts motor a large exerts force a in large the forcerange inwithin the range which within human which human muscles can musclesperform canat high perform efficiency, at high the efficiency, human m theuscles human would muscles have wouldto expend have more to expend more effort to exerteffort force to due exert to forcethe interference due to the interferenceof the in-wheel of the motor. in-wheel This motor.is because This human is because human muscles needmuscles to counterbalance need to counterbalance the instantaneous the instantaneous high pedal high velocity pedal output velocity (interfer- output (interference) ence) from thefrom in- thewheel in-wheel motor, motor, which whichrequires requires the muscles the muscles to consume to consume more energy more energy as a as a result result of the ofvarying the varying muscle muscle velocity. velocity. To minimize To minimize the loss the in riding loss in efficiency riding efficiency along with along with the the human musclehuman energy muscle energyconsumption, consumption, the in-wheel the in-wheel motor motormust actuate must actuate in the in range the range within within whichwhich the human the human muscles muscles perform perform at low at efficiency. low efficiency. Based on an Basedinvestigation on an investigation of the human of theinefficiency human inefficiency region during region the during pedaling the pedalingmo- motion, tion, we developedwe developed a new control a new controlalgorithm algorithm to improve to improve the efficiency the efficiency and riding and experi- riding experience ence of the Eof-bike the system E-bike systemthat addresses that addresses the human the part human and part the bike and part. the bike As mentioned part. As mentioned in in Section 1,Section the proposed1, the proposed algorithm algorithm consists consists of a feedback of a feedback PD control PD control based basedon a bike on a bike speed speed and a andprecalculated a precalculated feedforward feedforward control, control, which which is complemented is complemented by the by riding the riding in- inefficiency efficiency basedbased on on leg leg kinematics kinematics and and muscular muscular dynam dynamics.ics. The The proposed proposed control control sche- schematic for the matic for themotor motor inputinput commandcommand is shown in Figure Figure 88.. The red red box box shows shows the the muscular muscular dynamics dynamics andand leg leg kinematics kinematics of ofa human a human,, and and the the blue blue box box shows shows the the motor motor control control and and dynamics. dynamics. TheThe feedforward feedforward and and feedback feedback motor motor control control schemes schemes are are highli highlightedghted within within the blue the blue dotteddotted line. line. First, we implemented a feedforward control associated with leg kinematics and muscular dynamics, based on the human biomechanical model in the red box. When a person drives a bike at the desired speed, the person activates their muscles to track the desired speed. As mentioned in Section2, muscular dynamics efficiency, and then muscle activation, are needed to activate the muscle, and the necessary muscle activation for activating is transmitted from the brain as the excitation. This process was calculated in advance using experimental data using OpenSim API and MATLAB, because mathematical modeling is very complex and difficult. For the feedforward control, we aim to improve the riding efficiency and riding experience by compensating for the inefficient pedaling range. Thus, we set the target range for the motor assistance based on IR,Bike, which is the measure of inefficiency, as mentioned in Section 2.3. Sustainability 2021, 13, 768 11 of 18 Sustainability 2021, 13, x FOR PEER REVIEW 11 of 19

FigureFigure 8. 8. SchematicSchematic illustration illustration of of control control algorithm showing blockblock diagramdiagram of of E-bike E-bike velocity velocity control control for human–robot interaction (left) and motor control algorithm strategy (right). vdes, vact, and verr for human–robot interaction (left) and motor control algorithm strategy (right). vdes, vact, and verr represent the E-bike’s desired velocity, actual velocity, and velocity error, respectively. Pjoint is the represent the E-bike’s desired velocity, actual velocity, and velocity error, respectively. Pjoint is the leg joint power; τHuman is the human cycling torque; τcmd is the motor command torque; icmd is the leg joint power; τHuman is the human cycling torque; τ is the motor command torque; i is the current input for torque command; τmotor is the actual motorcmd torque; 휃crank and 휃̇ crank. are thecmd crank anglecurrent and input crank for angular torque velocity, command; respectively;τmotor is the Pcrank actual is the motor crank torque;power whichθcrank andis theθ crankproductare theof τ crankHuman anglė and crank angular velocity, respectively; P is the crank power which is the product of and 휃crank; kAl .is the control gain for the proposed algorithm;crank and IR,Bike is the human energy rate for equalτHuman cyclingand θ power.crank; kAl is the control gain for the proposed algorithm; and IR,Bike is the human energy rate for equal cycling power. First, we implemented a feedforward control associated with leg kinematics and muscularIn calculating dynamics, thebasedIR,Bike on forthe thehuman proposed biomechanical feedforward model control, in the∑ redaMuscle box.and WhenPCrank a personare crucial drives parameters a bike at the that desired affect humanspeed, energythe person consumption activates the duringir muscles cycling. to track However, the desiredwith respect speed. to As control, mentioned it is very in Section complicated 2, muscular to employ dynamics∑ aMuscle efficiency,in real-time and then as a muscle control activatiparameter,on, are unlike neededPCrank to activate. This isthe because muscle∑, aandMuscle the necessarymust be calculated muscle activation from the for static ac- tivatingoptimization is transmitted process from in OpenSim. the brain Practically,as the excitation.verr was This not process used inwas static calculated optimization, in ad- vancebecause using it couldexperimental not be data measured. using OpenSim To solve API this and problem, MATLAB we, used because the mathematical pre-measured moddataeling by requesting is very complex the subjects and difficult. to ride the bike at the desired velocity of 15 km/h. Since theFor subjects the feedforward conducted thecontrol, pedaling we aim motion to improve at the required the riding speed, efficiency we can and consider riding ex- the periencepedaling by data compensating as being obtained for the inefficient from conducting pedaling the range. human Thus, parts we of set Figure the target8. The range static optimization requires a substantial amount of time because the skeletal model, the number for the motor assistance based on 퐼푅,퐵𝑖푘푒, which is the measure of inefficiency, as men- of muscles, the pedaling force, the pedaling power and the joint angle of the leg, etc., must tioned in Section 2.3. be considered. In short, a is not suitable for use as a real-time control parameter in In calculating the 퐼 ∑ Muscle for the proposed feedforward control, ∑ 푎 and 푃 the proposed algorithm.푅,퐵𝑖푘푒 푀푢푠푐푙푒 퐶푟푎푛푘 are crucial parameters that affect human energy consumption during cycling. However, To solve this problem, we employed a methodology to estimate the human inefficiency with respect to control, it is very complicated to employ ∑ 푎푀푢푠푐푙푒 in real-time as a control I without having to use ∑ a . We measured and analyzed the pedaling force, parameter,R,Bike unlike 푃 . This is becauseMuscle ∑ 푎 must be calculated from the static op- the crank power, the퐶푟푎푛푘 crank angle, and the pedal푀푢푠푐푙푒 angle for pedaling with and without the timization process in OpenSim. Practically, 푣 was not used in static optimization, be- in-wheel motor for five subjects. The magnitudes푒푟푟 of the overall riding efficiency differed cause it could not be measured. To solve this problem, we used the pre-measured data by for the five subjects because of their different muscle characteristics (length, strength, etc.). requesting the subjects to ride the bike at the desired velocity of 15 km/h. Since the subjects Although the subjects differed in their overall riding efficiency, the five subjects exhibited conducted the pedaling motion at the required speed, we can consider the pedaling data similar tendencies in terms of their overall riding efficiency with equal riding conditions as(i.e., being 15 km/hobtained bike from velocity, conducting constant the load human friction, parts etc.). of WeFigure thus 8. concluded The static that optimization the human requires a substantial amount of time because the skeletal model, the number of muscles, inefficiency can be estimated from the experimental data. The estimated IR,Bike curve was thederived pedaling using force, a fifth-order the pedaling Fourier power transformation and the joint and angle is shownof the leg in Figure, etc., must6 as abe black consid- line ∑ ered.plotted In short, with respect푎푀푢푠푐푙푒 to theis not crank suitable angle. for use as a real-time control parameter in the pro- posedSecond, algorithm. we implemented the feedback control to maintain the desired bike velocity. BecauseTo solve bike this velocity problem, changes we dynamically,employed a methodology we used a proportional–derivative to estimate the human (PD) ineffi- con- ciencytroller 퐼 to푅,퐵𝑖푘푒 prevent without overshoot. having to Feedforward use ∑ 푎푀푢푠푐푙푒 control. We measu alone mayred and not analyzed be able to the maintain pedaling the force,desired the velocity,crank power, because the feedforward crank angle, controland the cannot pedal angle control for the pedaling rider’s velocitywith and accurately. without the in-Thewheel proposed motor for control five subjects. algorithm The not magnitudes only enhances of the efficiency, overall butriding also efficiency improves dif- the feredriding for experience. the five subjectsThis is because because the of proposed their different algorithm muscle prevents characteristics an abrupt change (length, in strength,bike velocity etc.). andAlthough acceleration, the subjects whilst differed reducing in human their overall energy riding consumption, efficiency, when the riding five subjectsan E-bike. exhibited The overall similar control tendencies schematic in terms is depicted of their overall in Figure riding8. Section efficiency3 describes with equal how ridingthe evaluation conditions metrics (i.e., 15 were km/h used bike to evaluatevelocity, theconstant proposed load algorithm. friction, etc.). We thus con- Sustainability 2021, 13, x FOR PEER REVIEW 12 of 19

cluded that the human inefficiency can be estimated from the experimental data. The estimated 퐼푅,퐵𝑖푘푒 curve was derived using a fifth-order Fourier transformation and is shown in Figure 6 as a black line plotted with respect to the crank angle. Second, we implemented the feedback control to maintain the desired bike velocity. Because bike velocity changes dynamically, we used a proportional–derivative (PD) controller to prevent overshoot. Feedforward control alone may not be able to maintain the desired velocity, because feedforward control cannot control the rider’s velocity accurately. The proposed control algorithm not only enhances efficiency, but also improves the riding experience. This is because the proposed algorithm prevents an abrupt change in bike velocity and acceleration, whilst reducing human energy consumption, when riding Sustainability 2021, 13, 768 an E-bike. The overall control schematic is depicted in Figure 8. Section 3 describes12 ofhow 18 the evaluation metrics were used to evaluate the proposed algorithm.

4.4. ExperimentalExperimental ResultsResults 4.1.4.1. ExperimentalExperimental SetupSetup AnAn E-bikeE-bike testbed,testbed, shownshown inin FigureFigure9 ,9, was was developed developed to to evaluate evaluate the the performance performance of of thethe proposedproposed controlcontrol algorithm.algorithm. AA real-timereal-time operationaloperational systemsystem (RTOS)-embedded(RTOS)-embedded LinuxLinux PCPC (STEP PC2, PC2, Neuromeka, Neuromeka, Inc. Inc.,, Seoul, Seoul, Republic Republic of Korea of Korea)) was wasused used with withan EPOS4 an EPOS4 motor motordriver driver (Maxon (Maxon Motor, Motor, Inc., Inc., Sachseln, Sachseln, Switzerland Switzerland)) to to control control the the in in-wheel-wheel motor, thethe AKM-85SXAKM-85SX 250W250W brushlessbrushless DCDC motormotor (Aikema(Aikema Electric,Electric, Inc.,Inc., SuzhouSuzhou City,City, China).China). TheThe in-in- wheelwheel motormotor generatesgenerates torquetorque atat thethe rearrear wheelwheel ofof thethe E-bike.E-bike. AA bikebike powerpower measurementmeasurement devicedevice waswas setset upup toto measuremeasure thethe bikebike powerpower andand applyapply an an externalexternal load load to to the the wheel. wheel. TheThe externalexternal loadload ofof 19.819.8 N*mN*m waswas inducedinduced ontoonto thethe pedalpedal toto matchmatch thethe outdooroutdoor roadroad frictionfriction conditioncondition toto thatthat ofof ourour testbed.testbed. ThisThis valuevalue waswas acquiredacquired byby thethe smartsmart pedalpedal whilewhile ridingriding ourour E-bikeE-bike outdoorsoutdoors onon aa concreteconcrete roadroad withwith minimalminimal slope.slope. TheThe designeddesigned smartsmart pedalpedal waswas composedcomposed ofof aa 6six-axis6six-axis Force/TorqueForce/Torque sensor,sensor, anan inertialinertial measurementmeasurement unitunit (IMU),(IMU), aa BluetoothBluetooth module,module, andand aa TeensyTeensy datadata acquisitionacquisition boardboard (PJRC).(PJRC). TheThe primaryprimary functionfunction ofof thethe smartsmart pedalpedal waswas toto measuremeasure thethe horizontalhorizontal andand verticalvertical componentscomponents of of thethe pedalpedal forceforce andand pedalpedal angle.angle. TheThe measuredmeasured datadata werewere transmittedtransmitted viavia BluetoothBluetooth communicationcommunication andand processedprocessed by by the the Teensy Teensy board. board. We We also also measured measured the the human human metabolic metabolic cost cost by by employing employ- aing K5 a mask K5 mask (COSMED, (COSMED, Inc., Inc. Roma,, Rom Italy)a, Italy to evaluate) to evaluate the human the human energy energy consumption. consumption.

FigureFigure 9.9. ExperimentalExperimental setupsetup forfor evaluatingevaluating thethe proposedproposed algorithm.algorithm. AA testbedtestbed withwith aa metabolicmetabolic cost cost measurementmeasurement device,device, COSMEDCOSMED K5K5 (left(left).). TestbedTestbed conﬁguration configuration ( right(right).).

Five healthy male adults (height, 178.6 ± 5.3 cm; weight, 80.8 ± 7.0 kg; age, 27.8 ± 1.8 years; mean ± SD) were recruited for this experiment. The experimental protocol was approved and regulated by the Chung-Ang University Institutional Review Board (IRB 1041078-201812-HR-229-01). None of the subjects had any history of musculoskeletal- related impairments or surgical operations that might have affected cycling activity. Details of the subjects are shown in Table1.

Table 1. Experimental Subjects.

Subjects Heights (cm) Weight (kg) Age (Year) Subject 1 177.4 85.0 29 Subject 2 182.6 81.1 25 Subject 3 183.0 74.4 27 Subject 4 170.1 73.4 29 Subject 5 180.2 90.0 29 Sustainability 2021, 13, 768 13 of 18

For the 5-min experiments conducted, the riding velocity was set to 15 km/h. This speed was based on the average bike speed in urban areas. The selected bike velocity of 15 km/h is an important criterion in evaluating the riding experience of each algorithm during the 5-min experiments. Due to the applied external load, the riders were unable to ride further after the 5-min experiment, thus the riding time was regulated to 5 min to prevent fatigue/injury of the riders. After the riders performed the 5-min experiment, they required at least a 15 min break until their metabolic signs were back to the stable sign again (within the COSMED software). Motor power was restricted to 40 W so that the motor could exert a crank torque up to 6.5 Nm at a bike velocity of 15 km/h, which is equivalent to 25% of the required crank torque. Restriction of the motor power was necessary because using an excessive motor power would result in E-bike operation with motor power rather than manpower. Table2 summarizes the algorithms we compared and their experimental conditions. There was no change in the slope and external load conditions based on the assumption that the majority of the bike riders prefer to ride their E-bike within the urban or rural area, where slope and road conditions are relatively uniform, and not within the off-road areas with extreme road conditions (e.g., mountain bike).

Table 2. Algorithm control variables.

Algorithms Velocity (km/h) Motor Power (W) Without Motor 15.96 ± 0.60 - Power Assistance System (PAS) 15.99 ± 0.89 41.17 ± 4.40 Proposed Algorithm 15.63 ± 0.67 40.61 ± 0.40

We evaluated the performance of the E-bike for three cases: (1) without a motor (human powered), (2) with the PAS algorithm, and (3) with the proposed algorithm. The PAS algorithm was chosen for comparison because it is the algorithm most commonly employed with commercial E-bikes [11]. The PAS algorithm commands motor torque in proportion to the instantaneous human torque. The P gains for the PAS and the proposed algorithm were set to 4 and 5, respectively. The D gain was set to 0.01 for the proposed algorithm. Since participants were instructed to maintain a constant bike velocity of 15 km/h based on the sensory data, and similar pedaling torque was observed during the experiment for all ﬁve subjects. Thus, the same control gains can be applied to each of the subjects. To evaluate performance for the three cases, we evaluated the net metabolic cost, E-bike system energy, riding experience, and riding jerk.

4.2. Net Metabolic Cost and Energy Consumption We compared the net metabolic cost for each case. Figure 10a shows the net metabolic cost data collected during the 5-min experiments. The case without a motor had the highest net metabolic cost of the three cases because the bike was driven by 100% manpower. The net metabolic cost in the PAS case was 6.24% (p = 0.0409) lower than in the case without a motor. The proposed algorithm demonstrated improved performance, as the net metabolic cost was reduced by 11.01% (p = 0.0007) in comparison to that of the case without a motor, and by 5.09% (p = 0.0407) in comparison to the PAS algorithm case. From these results, we conclude that the proposed algorithm successfully reduces human energy consumption by compensating for the human inefficiency. Figure 10b shows the total energy consumption of the human and the motor while cycling. The PAS algorithm aids the rider during pedaling as the motor provides additional rear wheel torque by feedback control. The total energy consumption in the PAS case (518.8 W) was observed to be higher than in the case without a motor (509.1 W). This is because the PAS algorithm does not account for human inefficiency and requires the rider to respond to abrupt changes in pedaling velocity due to the assisted torque. Consequently, there is a sudden increase in muscle velocity, and the rider needs to use more energy. The proposed algorithm, however, consumes less energy (493.9 W), as the assisted torque is provided in the inefficient pedaling range of the human. Therefore, it is necessary to Sustainability 2021, 13, x FOR PEER REVIEW 14 of 19

(a) (b)

FigureFigure 10. 10. ExperimentalExperimental results results of ofthree three different different algorithms algorithms showing showing change change in the in thefollowing following:: (a) Human(a) Human energy energy consumption consumption (net (net metabolic metabolic cost cost with with N N= =5, 5,ANOVA, ANOVA, pp == 0.0012). 0.0012). The The metabolic metabolic costcost is is successfully successfully reduced reduced by by the the proposed proposed algorithm algorithm under under the the same same driving driving condition; condition; ( (bb)) Total Total net cycling power. Although the change in motor power is marginal (approximately 0.5 W), the net cycling power. Although the change in motor power is marginal (approximately 0.5 W), the proposed algorithm reduces net cycling power (both motor and human power) under the same proposed algorithm reduces net cycling power (both motor and human power) under the same driving condition. driving condition. Figure 10b shows the total energy consumption of the human and the motor while 4.3. Riding Experience cycling. The PAS algorithm aids the rider during pedaling as the motor provides addi- tionalWe rear compared wheel torque both theby feedback riding velocity control. and The the total acceleration energy consumption stability achieved in the in PAS the casedifferent (518.8 cases W) was to evaluate observed the to be riding higher experience. than in the Velocity case without stability a motor can be (509.1 a measure W). This of ishow because comfortable the PAS the algorithm ridingexperience does not account is, based for on human the instantaneous inefficiency and velocity requires during the ridercycling to respond [16–18]. Whento abrupt the velocitychanges reducesin pedaling during velocity cycling, due the to riderthe assi muststed exert torque. an effortConse- to quently,speed up there the E-bike, is a sudden thus causing increase discomfort in muscle during velocity, cycling. and the Velocity rider stabilityneeds to is use assessed more energy.by determining The proposed the ratio algorithm, between however, the number consumes of velocity less energy measurements (493.9 W), that as the fall assisted within torquethe boundary is provided and thein the total inefficient number ofpedaling velocity range measurements. of the human. The Therefore, boundary isit basedis neces- on saryan average to consider velocity both of the 15 leg km/h kinematics (the average and the bike muscular velocity dynamics within the efficiencies urban area), to witheffec- a ± tively2 km/h reduce deviation the energy band consumption. [26–28]. Acceleration stability is another crucial factor that reflects how safe the rider feels 4.3.during Riding cycling Experience [17–19 ]. It is a measure of the riding instability, based on instantaneous acceleration and deceleration, which may cause riding imbalance or road accidents. Acceleration We compared both the riding velocity and the acceleration stability achieved in the stability is determined based on the bike acceleration bounds of −0.4 to 0.4 rad/s3 [16]. different cases to evaluate the riding experience. Velocity stability can be a measure of In order to evaluate the riding experience, we conducted experiments to measure the howbike comfortable velocity and the acceleration riding experience in each case is, based with respecton the instantaneous to the riding distance velocity and during subject, cy- clingas shown [16–18 in]. Figure When 11 the. If velocity the bike reduces velocity during and acceleration cycling, the fall rider within must the ex rangeert an bounded effort to speedby the up dashed the E- blackbike, lines,thus causing high velocity discomfort and acceleration during cycling. stability Velocity can bestability achieved. is assessed by determiningUsing these the data, ratio we between calculated the velocity number and of velocity acceleration measurements stability as thethat proportions fall within theof data boundary that fell and within the total the defined number boundaries of velocity in measurements. Table3. The higher The theboundary ratio is, is the based higher on anthe average E-bike isvelocity within of the 15 velocitykm/h (the and average acceleration bike velocity threshold. within The the results urban show area), that with the a ±2proposed km/h deviation algorithm band achieved [26–28 a]. greater riding experience (in both velocity and acceleration stability perspective) than the PAS algorithm. The PAS algorithm achieves relatively low velocity stability because the velocity of the E-bike suddenly increases as the motor provides excessive assistant torque when in the high efficiency range of human pedaling, resulting in inevitable interference between the human and the motor. This effect is closely related to riding efficiency because a sudden increase in velocity results in a sudden increase in muscle length. When this happens, the muscle does not have enough time to provide sufficient muscle force and the overall riding efficiency dramatically decreases. As a result, the rider feels great discomfort during the pedaling motion. It is noted that Sustainability 2021, 13, x FOR PEER REVIEW 15 of 19

Sustainability 2021, 13, 768 Acceleration stability is another crucial factor that reflects how safe the rider feels during15 of 18 cycling [17–19]. It is a measure of the riding instability, based on instantaneous acceleration and deceleration, which may cause riding imbalance or road accidents. Acceleration stability is determined based on the bike acceleration bounds of −0.4 to 0.4 rad/s3 [16]. accelerationIn order stabilityto evaluate in the this riding paper experience, has relatively we conducted low values experiments when compared to measure to the the results bikein Refs. velocity[17– 19and]. Itacceleration is observed in thateach ourcase bike’s with respect power to measurement the riding distance device and could subject, induce a ashigh shown rolling in Figure friction 11. on If thethe wheels,bike velocity and might and acceleration cause sudden fall accelerationwithin the range spikes, bounded which can byincrease the dashed deviation. black lines, high velocity and acceleration stability can be achieved.

FigureFigure 11. 11. BikeBike velocity velocity and and bike bike acceleration acceleration for forthe the five ﬁve subjects. subjects. Based Based on the on data the datashown shown in this in this figure,ﬁgure, we we calculated calculated velocity velocity and and acceleration acceleration stability stability.. The The dashed dashed black black lines lines and andgray grayshading shading indicateindicate thresholds thresholds for for velocity velocity and and acceleration acceleration stability. stability. Velocity Velocity stability stability is a factor is a related factor to related de- to termindetermininging the therate rate of instantaneous of instantaneous velocity. velocity. Acceleration Acceleration stability stability is a factor is a factor related related to determin to determininging the rate of instantaneous acceleration. The ratios of velocity and acceleration stability were evalu- the rate of instantaneous acceleration. The ratios of velocity and acceleration stability were evaluated ated based on the amount of acquired data within the threshold limits relative to the total data acquired.basedon the amount of acquired data within the threshold limits relative to the total data acquired.

TableUsing 3. Riding these experience data, we calculated of subjects velocity based on and algorithms accelerat used.ion stability as the proportions Ratio ofof Bike data Velocity that fell within within the Thresholdthe defined (%) boundaries Ratio in Table of Bike 3. The Acceleration higher the within ratio the is, Threshold the higher (%) Subjects Without Motorthe E-bikePAS is within theProposed velocity Algorithm and acceleration Without threshold. Motor The PASresults showProposed that the Algorithm pro- Subject 1 88.03posed algorithm 72.91 achieved a 94.83 greater riding experience 61.49 (in both 28.86velocity and acceleration 56.21 Subject 2 97.71stability perspective) 98.46 than the 98.08 PAS algorithm. The 67.55 PAS algorithm 50.88 achieves relatively 66.23 low Subject 3 92.05 90.09 93.42 77.11 33.65 76.67 Subject 4 94.76velocity stability 87.02 because the 98.44 velocity of the E-bike 63.70 suddenly increases 35.71 as the motor 52.94 pro- Subject 5 98.92vides excessive 76.12 assistant torque 99.28 when in the high 74.34 efficiency range 41.61 of human pedaling, 75.06 resulting in inevitable interference between the human and the motor. This effect is closely Average Standard 94.30 ± 4.40 84.92 ± 10.44 96.81 ± 2.54 68.84 ± 6.72 38.14 ± 8.47 65.42 ± 10.73 Deviation related to riding efficiency because a sudden increase in velocity results in a sudden increase in muscle length. When this happens, the muscle does not have enough time to provideOn sufficient the other muscle hand, force riding and without the overall a motor riding results efficiency in velocity dramatically and acceleration decreases. sta- Asbility a result, levels the similar rider tofeels those great achieved discomfort with during the proposed the pedaling algorithm. motion. However, It is noted as that shown accelin Figureeration 10 stability, the total in energythis paper associated has relatively with low the values proposed when algorithm compared is to lower the results than that inassociated Refs. [17– with19]. It the is observed case without that aour motor. bike’s Therefore, power measurement we conclude device that could the proposed induce a algo- highrithm rolling can provide friction a on rider the with wheels velocity, and might and acceleration cause sudden stability acceleration similar spikes, to that achievablewhich canwithout increase a motor, deviation. while consuming less energy. Overall, energy consumption can be mini- mized by compensating for the human inefficient range in pedaling. As a result, minimal interference between the two actuators (human muscles and the motor) can be achieved by maximizing the muscles’ efficiency, while utilizing the motor’s torque assistance in the inefficient range only.

Table 3. Riding experience of subjects based on algorithms used.

Ratio of Bike Acceleration within the Ratio of Bike Velocity within the Threshold (%) Subjects Threshold (%) Without Motor PAS Proposed Algorithm Without Motor PAS Proposed Algorithm Subject 1 88.03 72.91 94.83 61.49 28.86 56.21 Subject 2 97.71 98.46 98.08 67.55 50.88 66.23 Subject 3 92.05 90.09 93.42 77.11 33.65 76.67 Subject 4 94.76 87.02 98.44 63.70 35.71 52.94 Subject 5 98.92 76.12 99.28 74.34 41.61 75.06 Average Standard 84.92 ± 38.14 94.30 ± 4.40 96.81 ± 2.54 68.84 ± 6.72 65.42 ± 10.73 Deviation 10.44 ± 8.47

On the other hand, riding without a motor results in velocity and acceleration stability levels similar to those achieved with the proposed algorithm. However, as shown in Figure 10, the total energy associated with the proposed algorithm is lower than that associated with the case without a motor. Therefore, we conclude that the proposed algorithm can provide a rider with velocity and acceleration stability similar to that achievable without a motor, while consuming less energy. Overall, energy consumption can be min- imized by compensating for the human inefficient range in pedaling. As a result, minimal interference between the two actuators (human muscles and the motor) can be achieved by maximizing the muscles’ efficiency, while utilizing the motor’s torque assistance in the inefficient range only.

4.4. Crank Jerk The last criterion to be evaluated is the crank jerk, which is a measure of crank motion smoothness during pedaling. High crank jerk may cause discomfort to the rider as a result of a loss in kinetic energy transmission from the human to the bike, and may require more Sustainability 2021, 13, 768 effort for the rider to transmit crank power [29]. Second, a high jerky motion causes dis-16 of 18 turbance during pedaling, and results in faster muscle fatigue for the rider [30]. Based on the experimental conditions described in Section 4.1, we analyzed the crank jerk during pedaling. Figure 12 shows the crank jerk for the five subjects for each case duringduring 5-min5-min experiments. experiments. As As Figure Figure 12a 12 shows,a shows, by by compensating compensating for forhuman human inefficiency inefﬁciency duringduring pedaling, pedaling, the the proposed proposed algorithm algorithm results results in smaller in smaller crank jerks crank than jerks the thanPAS algo- the PAS algorithmrithm does. does. ThisThis is consistent is consistent with with the results the results shown shown in Fi ingure Figure 9. As9. Asmentioned mentioned before, before, jerkyjerky motionsmotions cause cause kinetic kinetic energy energy loss loss and and muscle muscle fatigue. fatigue. As Assuch, such, the theamount amount of energy of energy consumedconsumed with the the PAS PAS algorithm algorithm tends tends to to be be higher higher because because of the of the larger larger jerky jerky motions motions itit produces.produces.

(a) (b)

FigureFigure 12. CrankCrank jerk jerk comparison comparison for forthe thesmoothness smoothness of cycling of cycling.. (a) Average (a) Average crank jerk crank for each jerk case for. each case.(b) Average (b) Average and standard and standard deviation deviation of crank ofjerk crank error jerk for the error five for subjects. the ﬁve Crank subjects. jerk was Crank reduced jerk was by the proposed algorithm by 12.36% and 6.86% with respect to the PAS algorithm and the case reduced by the proposed algorithm by 12.36% and 6.86% with respect to the PAS algorithm and the without a motor, respectively. case without a motor, respectively.

However, the proposed algorithm resulted in a relatively higher crank jerk for Subject 2 than was observed without a motor. This is because Subject 2 was not satisﬁed with the saddle shape and felt discomfort throughout the experiment, which inﬂuenced the crank jerk. The other subjects experienced relatively low crank jerk when riding the E-bike operated with the proposed algorithm. As a result, an average crank jerk reduction of 12.36% was observed relative to operation with the PAS algorithm. The results are shown in Figure 12b.

5. Discussion In this paper, in order to address this problem associated with the PAS algorithm, we propose a new control algorithm intended to enable highly efficient pedaling and provide an enhanced riding experience for the rider. We conducted analyses of human leg kinematics and muscular dynamics during bike pedaling using MATLAB and OpenSim. In this analysis process, we defined a term referred to as the overall riding efficiency, ηR,Bike, based on the muscle activation concept. For the proposed motor control algorithm, both feedforward and feedback control are used to enhance the system efficiency. While feedforward control compensates for the IR,Bike, feedback control maintains a consistent bike velocity. For the purpose of real-time control, pedaling inefficiency based on muscular dynamics and leg kinematics were predicted using a fifth-order Fourier transformation. In order to evaluate the proposed control algorithm, we compared its performance with that of the PAS algorithm. The human energy consumption associated with our algorithm was 5.09% lower than that with the PAS algorithm and 11.04% lower than that without a motor. The total average energy consumption associated with the proposed algorithm (493.9 W) was lower than that with PAS (518.8 W) and without a motor. In addition, the proposed algorithm achieved an 11.9% higher velocity stability and a 27.28% higher acceleration stability than PAS. These results demonstrate that the proposed algorithm makes an E-bike system more efficient, while providing enhanced velocity and acceleration stability. The proposed algorithm also produces 12.36% lower jerk than PAS, which suggests that the proposed algorithm reduces muscle fatigue and energy transmission loss by improving the smoothness of pedaling. These results were possible because our algorithm successfully aids pedaling within the inefficient range and performs velocity feedback effectively. Sustainability 2021, 13, 768 17 of 18

The proposed approach can enhance the riding efﬁciency of E-bikes while consuming less energy by minimizing the interference in a system in which a human and robot work in close proximity. The current work with our bike measurement device can generate limited friction conditions, and thus various road conditions, such as gravel, dirt, steep roads, etc., are not fully reﬂected in the results reported here. In future work, we plan to evaluate the performance of the proposed algorithm in various environments and adjust the control parameters accordingly.

6. Conclusions With the growing interest in HRI, the use of robots for both industrial and household applications is increasing rapidly. Among the many applications of HRI, those involv- ing proximate interaction are tasks that involve physical interaction between humans and robots. Safety is the utmost concern for this type of application. Some examples of proximate interaction are E-bikes, autonomous vehicles, and electric scooters. E-bikes have been gaining attention as eco-friendly vehicles for urban transportation. For this reason, E-bikes are widely used in densely populated areas in Europe, China, and other places around the world. However, the PAS algorithm used for E-bikes does not account for the characteristics of the muscles during the phases of pedaling. Instead, it provides assistance to the rider in the high-efﬁciency pedaling range. As a result, interference between the human operator and the motor is inevitable. As a result of this inefﬁcient actuation, excessive energy consumption is required for human muscles to compensate for the interference. Because our approach was developed on the basis of human analysis, the application is not limited to E-bike systems but rather can also be extended to numerous types of systems that involve close physical human–robot interaction, such as smart factory robots, kitchen robots, social robots, and others.

Author Contributions: Developed the algorithm concept and drafted the paper and reviewed and revised the paper, D.H.K.; reviewed and revised the paper for submission, and investigated the experimental data, D.L.; setup hardware for the testbed and conducted experiments, Y.K. and S.K.; supervised overall work and revised the paper, D.S. All authors have read and agreed to the published version of the manuscript. Funding: This study was supported in part by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT & Future Planning (No.NRF-2018R1C1B6008549). Also, this study was supported by the Industrial Technology Inno- vation Program (No. 20007058, Development of safe and comfortable human augmentation hybrid robot suit) funded by the Ministry of Trade, Industry & Energy (MOTIE, Korea). And, this study was supported by the Chung-Ang University Graduate Research Scholarship in 2017. Institutional Review Board Statement: This study was approved and regulated by the Chung-Ang University Institutional Review Board (IRB 1041078-201812-HR-229-01). Informed Consent Statement: Informed consent was obtained from all subjects involved in the study. Data Availability Statement: No new data were created or analyzed in this study. Data sharing is not applicable to this article. Acknowledgments: The authors would like to thank people in the Human-Centered Robotics Laboratory at Chung-Ang University for their valuable comments and feedback. Conﬂicts of Interest: The authors declare no conﬂict of interest.

References 1. Shin, D.; Sardellitti, I.; Khatib, O.; Cutkosky, M. Design and control of a bio-inspired human-friendly robot. Int. J. Robot. Res. 2010, 29, 571–584. [CrossRef] 2. Wang, P. Dynamics and Control of Rider-Bicycle Systems. Ph.D. Thesis, Rutgers University, Piscataway, NJ, USA, 2018. 3. Wang, P.; Gong, Y.; Yi, J.; Liu, T. An integrated stationary/moving balance control of an autonomous bikebot. In Proceedings of the IEEE American Control Conference, Philadelphia, PA, USA, 10–12 July 2019; pp. 3273–3278. Sustainability 2021, 13, 768 18 of 18

4. Kim, N.; Yun, S.; Shin, D. A bio-inspired lightweight wrist for high-dof robotic prosthetic arms. IEEE/ASME Trans. Mechatron. 2019, 24, 2674–2683. [CrossRef] 5. Lee, D.; Kim, D.H.; Che, C.H.; In, J.B.; Shin, D. Highly durable bidirectional joint with twisted string actuators and variable radius pulley. IEEE/ASME Trans. Mechatron. 2020, 25, 2674–2683. [CrossRef] 6. Goodrich, M.A.; Schultz, A.C. Human–robot interaction: A survey. Found. Trends® Hum. Comput. Interact. 2007, 1, 203–275. [CrossRef] 7. Kim, N.H.; Kim, J.M.; Khatib, O.; Shin, D. Design optimization of hybrid actuation combining macro-mini actuators. Int. J. Precis. Eng. Manuf. 2017, 18, 519–527. [CrossRef] 8. Hansen, K.B.; Nielsen, T.A.S. Exploring characteristics and motives of long distance commuter cyclists. Transp. Policy. 2014, 35, 57–63. [CrossRef] 9. Ton, D.; Duives, D.C.; Cats, O.; Hoogendoorn-Lanser, S.; Hoogendoorn, S.P. Cycling or walking? Determinants of mode choice in the Netherlands. Transp. Res. Part A Policy Pract. 2019, 123, 7–23. [CrossRef] 10. Koszowski, C.; Gerike, R.; Hubrich, S.; Götschi, T.; Pohle, M.; Wittwer, R. Active mobility: Bringing together transport planning, urban planning, and public health. In Towards User-Centric Transport in Europe; Springer: Cham, Switzerland, 2019; pp. 149–171. 11. Hatada, K.; Hirata, K. Energy-efficient power assisting methods for periodic motions and its experimental verification. In Proceedings of the 2012 IEEE International Conference on Industrial Technology, Athens, Greece, 19–21 March 2012; pp. 858–863. 12. Spagnol, P.; Alli, G.; Spelta, C.; Lisanti, P.; Todeschini, F.; Savaresi, S.M.; Morelli, A. A full hybrid electric bike: How to increase human efficiency. In Proceedings of the 2012 American Control Conference (ACC), Montreal, QC, Canada, 27–29 June 2012; pp. 2761–2766. 13. Hull, M.L.; Gonzalez, H. Bivariate optimization of pedaling rate and crank arm length in cycling. J. Biomech. 1988, 21, 839–849. [CrossRef] 14. Too, D.; Williams, C. Determination of the optimal crank arm length to maximize peak power production in an upright cycling position. Hum. Power 2018, 25, 1–26. 15. Cafolla, D.; Chen, I.-M.; Ceccarelli, M. An experimental characterization of human torso motion. Front. Mech. Eng. 2015, 10, 311–325. [CrossRef] 16. Gorenflo, C.; Rios, I.; Golab, L.; Keshav, S. Usage patterns of electric bicycles: An analysis of the WeBike project. J. Adv. Transp. 2017, 2017, 1–14. [CrossRef] 17. Hsu, R.C.; Liu, C.T.; Chan, D.Y. A reinforcement-learning-based assisted power management with QoR provisioning for human–electric hybrid bicycle. Trans. Ind. Electron. 2012, 59, 3350–3359. [CrossRef] 18. Lee, J.S.; Jiang, J.W. Enhanced fuzzy-logic-based power-assisted control with user-adaptive systems for human-electric bikes. IET Intell. Transp. Syst. 2019, 13, 1492–1498. [CrossRef] 19. Liu, C.T.; Hsu, R.C. A fuzzy Q-learning based assisted power management method for comfortable riding of pedelec. In Proceed- ings of the 2015 6th International Conference on Automation, Robotics and Applications (ICARA), Queenstown, New Zealand, 17–19 February 2015; pp. 580–585. 20. Bhargava, L.J.; Pandy, M.G.; Anderson, F.C. A phenomenological model for estimating metabolic energy consumption in muscle contraction. J. Biomech. 2004, 37, 81–88. [CrossRef] 21. Marks, R. The effect of isometric quadriceps strength training in mid-range for osteoarthritis of the knee. Arthritis Rheum. Off. J. Am. Coll. Rheumatol. 1993, 6, 52–56. [CrossRef] 22. Thelen, D.G. Adjustment of muscle mechanics model parameters to simulate dynamic contractions in older adults. J. Biomech. Eng. 2003, 125, 70–77. [CrossRef] 23. Ting, L.H.; Kautz, S.A.; Brown, D.A.; Zajac, F.E. Phase reversal of biomechanical functions and muscle activity in backward cycling. J. Neurophysiol. 1999, 81, 544–551. [CrossRef] 24. Childers, W.L.; Kistenberg, R.S.; Gregor, R.J. The biomechanics of cycling with a transtibial amputation: Recommendations for prosthetic design and direction for future research. Prosthet. Orthot. Intl. 2009, 33, 256–271. [CrossRef] 25. Lai, A.K.; Arnold, A.S.; Wakeling, J.M. Why are antagonist muscles co-activated in my simulation? A musculoskeletal model for analysing human locomotor tasks. Ann. Biomed. Eng. 2017, 45, 2762–2774. [CrossRef] 26. Shi, F. Improving urban non-motorized mobility for public affairs trips: A survey and analysis of innovative official bicycles in Nanjing city, China. J. Urban Manag. 2019, 8, 396–407. [CrossRef] 27. Jensen, P.; Rouquier, J.B.; Ovtracht, N.; Robardet, C. Characterizing the speed and paths of shared bicycle use in Lyon. Transp. Res. Part D Transp. Environ. 2010, 15, 522–524. [CrossRef] 28. Rossi, R.; Mantuano, A.; Pascucci, F.; Rupi, F. Fitting time headway and speed distributions for bicycles on separate bicycle lanes. Transp. Res. Procedia 2017, 27, 19–26. [CrossRef] 29. Szecsi, J.; Krause, P.; Krafczyk, S.; Brandt, T.; Straube, A. Functional output improvement in FES cycling by means of forced smooth cycling. Med. Sci. Sports Exerc. 2007, 39, 764–780. [CrossRef][PubMed] 30. Zhang, L.; Diraneyya, M.M.; Ryu, J.; Haas, C.T.; Abdel-Rahman, E.M. Jerk as an indicator of physical exertion and fatigue. Autom. Constr. 2019, 104, 120–128. [CrossRef]