Oscillation annealing and driver/tire load torque estimation in Electric Power Steering Systems Javier Tordesillas, Valentina Ciarla, Carlos Canudas de Wit
To cite this version:
Javier Tordesillas, Valentina Ciarla, Carlos Canudas de Wit. Oscillation annealing and driver/tire load torque estimation in Electric Power Steering Systems. MSC 2011 - IEEE Multi-Conference on Systems and Control, Sep 2011, Denver, United States. pp.s/n. hal-00642035
HAL Id: hal-00642035 https://hal.archives-ouvertes.fr/hal-00642035 Submitted on 17 Nov 2011
HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Oscillation annealing and driver/tire load torque estimation in Electric Power Steering Systems
J. Tordesillas Ill´an, V. Ciarla, and C. Canudas de Wit
Abstract— The paper presents several aspects of modeling, wheel. In that way the low frequency feelings of the driver observation and control towards a new generation of Electrical will not be affected. The observer in this first version was Power Steering(EPS) systems. In particular we design an derived assuming that the exogenous inputs (driver and tire- optimal control to reject oscillations of the steering column, then we device a new observer to estimate the internal state friction torques) were known. variables of the steering column, the driver applied torque b) Improved LuGre Tire-friction model: The dynamic (steering wheel torque), and the load torque (tire/ground contact LuGre Friction model proposed in [5], [6] is here modified friction). Finally, we also revisited the LuGre tire dynamic friction model by improving the transient behavior between to trade the transition between the stick friction component the sticking phases and the dynamic ones. Simulation of the (that dominated when vehicle is stand still), and the self- proposed control and observer are shown at the end of the alignment torque (produced when vehicle in motion). To paper using the improved LuGre-tire friction model. have a complete tire-friction model is mandatory to test EPS Index Terms— Electric Power Steering systems (EPSs), LQ systems. control, LuGre friction model, observer. c) Driver and load torque estimation: The observer I.INTRODUCTION previously designed was extended to the case where the Nowadays, the existing Electric Power Steering systems exogenous inputs are unknown. This was done assuming the (EPSs) are developed basing on a general driver profile [1], standard hypothesis for observation under unknown input, [2]. Future generation of EPSs should be able to propose a namely a slow rate of variations of these inputs. In spite torque assistance which may be adapted for different kinds of this hypothesis, our results show that these exogenous of driver population: young drivers, aged people, disabled torques where correctly estimated with a very good precision people, etc. and very little phase lag. A first important step towards the above goals consists in setting a control framework that includes a realistic model 2 φ˙ of a steering column accounting for all other torque loads Tyre/Road involved in a real driving situation (torsion torque due to Friction Model β force sensors flexibility, applied driver forces, and tire-road V contact friction forces). The control framework should also include observers that allow to recover signals which are not τa τv 3 y sensed and a control law that compensates for the column Mechanical flexibilities. This is completed with a booster-torque-law Reference Model u u (power steering torque) specific to the driver population Model ass in question. The general architecture that we propose is shown in Fig. 1. This paper concerns with the block 1 xˆ (observer/control), and the block 2 (tire-friction model) of −K Observer 1 Fig. 1, with the additional contribution that the observer also estimate the driver and load torque. The design of the specific Fig. 1. General architecture of the EPSs: block 1 concerns the observer and reference model (block 3) for non-standard drivers is under the control; block 2 includes the tyre-friction model and block 3 includes current investigation. the reference model. The contributions of this paper are: a) Optimal output control feedback: Based on the steer- The paper is organized following the 3 main items as ing column model proposed by [3], [4], we redesign a linear indicated above. optimal control that seeks to cancel oscillation due to column stiffness. This results in an output optimal feedback with an observer included. In addition to cancel oscillation the control design seeks to preserve the open-loop gain between II. OSCILLATIONS ANNEALING the applied driver torque and the motion of the steering To begin with, it is essential to define the mechanical C. Canudas-de-Wit and V. Ciarla are with the Laboratoire d’Automatique de Grenoble, UMR CNRS, Grenoble 5528, France (e-mails: carlos.canudas- system that we will take into account. Fig.2 shows an [email protected], [email protected]) explanatory schema of the system. TABLE I where T = − Bv 0 − k . Fig. 3 shows the frequency CONSTANT PARAMETERS OF THE EPSS Jv Jv response and puts in evidence that the open loop system has Symbol Value Description a significant peak for a frequency of about 12 Hz that cause 2 Jv steering wheel inertia 0.025 [kg · m ] 2 Jm motor inertia 0.00033 [kg · m ] substantial oscillations on the steering wheel and that should Jc steering motor inertia neglected Jw aggregated wheel and rack inertia neglected be avoided so as to improve the driving comfort. k steering column stiffness 100 [N · m] N1 vehicle steering angle to steering colon ratio 13.67 N2 motor-steering column gear ratio 17 B. Full-state optimal feedback control Bv damping coeff. associated to the steering wheel 0.01 [N · s/m] Bm damping coeff. associated to the motor axis 0.003 [N · s/m] To compensate the oscillations an optimal LQR controller is designed. This controller is computed so that the state- τa feedback law u = −Kx (with K the state-feedback gain) = τ θs θm/N2 minimizes following cost function: v θv ∞ T 2 Jv 1 J = (x Qx + ru ) dt (5) k N 0 N2 Z Jm u where the constant matrices Q > 0 and r > 0 are θm the weighting matrices. For the purpose of annealing the Bv Bm oscillation without impact the low frequency gains of the resulting closed loop, we can select Q such that only the Fig. 2. Mechanical model of the EPSs torsion angle and its time derivative is penalized, i.e. ∞ 2 2 2 J = (q1(x1 − x2) + q2x3 + ru ) dt 0 A. Column model Z The mechanical equations governing the system explained Normalizing this cost with r =1, the problem is simplified above are (see [3]) by selecting Q as q1 −q1 0 Jvθ¨v = τv − k(θv − θs) − Bvθ˙v (1) Q = −q1 q1 0 (6) ¨ 2 ˙ τa JT θs = −k(θs − θv) − N2 Bmθs + + N2u (2) 0 0 q2 N1 This results in a cost function with only two parameter to 2 Jw with JT = Jc + N2 Jm + 2 . The constants of the N1 be tuned. The closed-loop transfer function is given by the model are defined in Table I, while θv, θs and θm are, following equation: respectively, the steering wheel, the motor-shaft and the ¨ motor angles. Both Jc and Jw are neglected for simplicity, θv −1 Gcl = = T (sI − (A − BK)) G1 (7) but to obtain more precise results it is advisable to introduce τv these parameters in the calculations. Let with A − BK the closed-loop state space matrix. T T T x = x1, x2, x3 = θ˙v, θ˙s,θv − θs (3) Bode Diagram The model can be formulated into the state-space form 80
60 x˙ = Ax + Bu + Gw 40 Open Loop Peak T with w = τv, τa and 20 at 12 Hz