On-board Electrical, Electronics and Pose Estimation System for Hyperloop Pod Design Nihal Singh*, Jay Karhade*, Ishika Bhattacharya*, Prathamesh Saraf*, Plava Kattamuri*, Alivelu Manga Parimi Birla Institute of Technology and Science, Pilani Email: [email protected], [email protected], [email protected], [email protected], [email protected], [email protected] *All authors have equal contribution Abstract—Hyperloop is a high-speed ground-based transporta- Master Computer serves as the connection between the remote tion system utilizing sealed tubes, with the aim of ultimately laptop and embedded micro-controllers. A dedicated Flight transporting passengers between metropolitan cities in efficiently Controller whose sole responsibility performs state estimation, designed autonomous capsules. In recent years, the design and development of sub-scale prototypes for these Hyperloop pods has and Front and Rear Analog modules read non-critical sensor set the foundation for realizing more practical and scalable pod data for monitoring. In addition, there is a fiducial detector architectures. This paper proposes a practical, power and space module and a module for braking. The rLoop design [6], too, optimized on-board electronics architecture, coupled with an uses a combination of retro-reflective photo-detection sensors end-to-end computationally efficient pose estimation algorithm. and inertial measurement units for navigation. Considering the high energy density and discharge rate of on- board batteries, this work additionally presents a robust system The heart of the system, encapsulated by the power node, for fault detection, protection and management of batteries, along focuses on power conditioning, battery monitoring, and overall with the design of the surrounding electrical system. Performance system safety. Prolonged operation of the pod at maximum evaluation and verification of proposed algorithms and circuits capacity, can, however, result in increased temperatures in the has been carried out by software simulations using both Python cells, which eventually may lead to a phenomenon known as and Simulink. Index Terms—Hyperloop, distributed architecture, battery electrical arcing. management system, pose estimation, fault detection To prevent arcing, the rLoop team [6] designed a vessel of carbon dioxide to store the battery pack, whereas the Georgia Tech pod [7] built a water cooling based system for batteries I. INTRODUCTION and other electronics. Other teams like the one from UC Santa The Hyperloop is a novel mode of high speed passenger Barbara [8] programmed their system for partial or complete and freight transportation, based on an open-source vactrain shut down and execution of emergency protocols in the case design released by a joint team from Tesla and SpaceX. It of detected malfunctions. is a form of low friction ground transport aiming for speeds Another important part of the hyperloop pod is the pose over 700mph with the help of partially evacuated low-pressure estimation in real-time. Generally, the parameters that are tubes. Work on air-pressure driven transportation dates back measured during pose estimation are acceleration, velocity to the early 1800s with the introducton of the concept of an and position and attitude. Pose estimation tracking is done for “atmospheric railway” [1] and coinage of the term Vactrains a variety of reasons ranging from safety to motion-planning [2]. Interest in the concept was renewed with Elon Musk’s and even vehicle control. A range of sensors are used for original white paper on Hyperloop Alpha [3]. The emphasis obtaining the pod’s pose, however, each of them suffer from on the need for more efficient transport systems is increasing their own drawbacks which has been discussed in the pose- in recent times, and there has been considerable evolution in estimation section further. This is facilitated by a robust set arXiv:2012.09412v1 [eess.SY] 17 Dec 2020 the various technical aspects of the pod. of position estimation algorithms. Effective pose estimation is The nervous system of any general pod design is comprised constituted by a low latency system with high resolution and of its software and embedded systems. The expectations and accuracy of the estimates. Nikolaev [4], in his work on the challenges of this combined subsystem have been well-detailed Waterloo pod’s software system, proposes a minimal Kalman in the paper on the Goose-3 pod [4]. Using a divide-and- filter that simply fuses the data from 3 Inertial Measurement conquer approach, the tasks were divided between (1) the Units (IMUs). Drift and possible existing correlation between embedded system, responsible for collecting data from on- the sensors due to sensing the same physical quantity of a board sensors and executing control panel commands, and moving body are some of the limitations of this method. (2) the communication system, responsible for transmission The MIT Hyperloop Report [5], has proposed an Extended of data between the remote control panel and the embedded Kalman Filter(EKF) that takes into account ground truths by system. Key features include the presence of a CAN-BUS and reading the fiducial markers throughout the tube length which a Master-Slave Design with master and hub units. In the MIT not only serves as a comparison stage to the IMUs, but can design [5], the system is split into seven important blocks. The act as a recalibration step at constant intervals thus reducing Fig. 1. On-board System Architecture Model the bias and drift from true estimates even further. While hardware. Often, in the interest of better performance, due this serves as a basic Kalman filtering update, it is required consideration is not given to monitoring the health of the to analyze the velocity, acceleration and position models on-board power systems. 2) In the Electrical System Design, and introduce an element of independent sensing for each safety has been strongly emphasized on, with focus on arcing parameter. Alternate systems taken by other teams like UCI [9] mitigation and fault-detection. This falls under the purview include the ellipse-N commercial model which is an out-of- of the software subsystem, as does pose estimation. 3) For the-box model for pose estimation. The ellipse-N series uses pose estimation, the proposed design focuses on developing Global Positioning System(GPS)/Global Navigation Satellite a robust positioning technique that can utilize on-board and System(GNSS) coupled with IMU sensing which while offers internal tube structure resources with minimal assumptions extremely accurate measurements and integrate some of the and external installations to provide accurate estimates. It most accurate positioning algorithms may suffer inside the makes use of a simple, but diverse, range of sensors namely: tube due to satellite position attenuation and failure [10]. A IMUs, Tachometers, Absolute Optical Encoders and Fiducial recent patent [11] assumes passive elements throughout the markers and simulated acceleration data following a pre- tube length which would yield accurate and precise estimates, planned trajectory. It has been assumed for simplification but it is not immediately clear as to what the economic costs of the algorithm, that the motion is along a linear axis. A would entail. The focus hence should be to develop a robust Kalman filter is used for fusing estimates and the updation positioning technique that can utilize onboard and internal tube follows a hierarchical fashion updating the estimates in the structure resources with minimal assumptions and external order: Acceleration, Velocity followed by position. Attitude installations to provide accurate estimates. estimation is done in an independent loop and involves only IMUs. Each of these subsystems has been further elaborated This paper introduces a custom on-board hardware ar- on in the following sections. chitecture for end-to-end interfacing and internal control of electrical and electronics components. The work presented II. ELECTRONICS SYSTEM DESIGN in this paper further proposes 1) detailed position, velocity and acceleration estimation algorithms specific to Hyperloop Sensing and monitoring of the pod’s environment and pod systems, along with and an efficient electrical system. internal functionalities is critical for a hyperloop pod. The Additionally on Electronics System Design, the board-level data gathered by various sensors is processed to determine the architecture proposed leverages salient features of the afore- core parameters for all on-board control and actuation. The mentioned embedded system designs while prioritizing dis- presented system architecture model builds on the work of tributed sensing and control along with the efficient use of other cited literature to propose a distributed noise-immune design with minimal hardware overhead. Each subsequent section details the various components used in engineering this system as shown in Fig 1, along with descriptions of their hardware layout and functionalities. A. CAN Bus The electronics system design is based on a Controller Area Network (CAN)-BUS shown in Fig 1, popularly used in automobiles to simplify connections between different elec- tronic control units. This reduces the chances of single point failures by making the system more noise-robust. Real-time, reliability and flexibility make it an indispensable network communication technology applied in the automobile network communication field [12]. It makes communication