车辆碰撞动力学模型和控制策略实验研究

车辆在高速情况下的碰撞易引发车辆失控。如果驾驶员无法及时或者正确地操作，可能会发生二次事故，这通常会导致严重的经济损失和人员伤亡。基于碰撞动力学模型进行碰撞后控制研究是目前解决车辆碰撞失控后恢复控制的主流方案，从理论上能够对碰撞后车辆的状态进行分析且具备良好的可解释性。但是由于车辆碰撞试验中存在高危险性、高成本、碰撞参数难测量、无法进行大规模试验等问题，导致四自由度车辆碰撞动力学模型只进行了仿真验证，缺乏物理的有效性验证，模型参数和现实真值差别较大，研究无法进展到实物部署阶段。车辆碰撞动力学模型和碰撞后控制器的物理验证是制约碰撞后控制研究的最大障碍。

为解决碰撞动力学模型的有效性验证问题，本文的主要研究工作是进行八自由度车辆碰撞动力学模型的有效性验证，并基于验证后的八自由度车辆碰撞动力学模型和深度强化学习框架构建一种新的碰撞后控制策略，提高车辆碰撞后的行驶安全性。首先，根据牛顿第二定律和刚体动力学，搭建八自由度车辆碰撞动力学模型。模型由四自由度车辆碰撞动力学模型与单轮动力学模型结合而成，采用非线性轮胎模型提高模型在极端工况下的准确性。为了解决动力学模型的物理有效性验证的难题，本文设计并搭建了一套四轮独立驱动的缩比模型车。在基于无量纲原则的设计思路下，所设计的缩比模型车与全尺寸车辆之间存在较好的动态相似性。为了能够精确记录车辆的运动状态，采用动作捕捉系统捕捉车身位姿，并设计碰撞力检测与测量系统来获取碰撞过程中车辆的受力情况。通过 车辆驱动、制动、受碰等多个工况验证动力学模型的有效性。最后，本文基于八自由度车辆碰撞动力学模型和TD3深度强化学习算法，开发了一种新的碰撞后控制策略。针对碰撞工况的车辆动力学仿真环境，本文集成八自由度车辆碰撞动力学模型并根据碰撞力特征简化碰撞过程。在基于TD3的控制策略设计方面，采用广义贝尔函数设计回报函数，多段渐进式训练策略，并将训练场景的碰撞力随机化，在保证网络快速收敛的同时最终策略具备良好的泛化性。通过Simulink-Carsim联合仿真在多种碰撞场景下测试控制器的性能，验证该控制策略能有效提高车辆碰撞后的行驶安全性。

Collisions at high speeds can easily lead to loss of vehicle control. If the driver is unable to operate in a timely or correct manner, a secondary accident may occur, often resulting in significant economic losses and casualties. Research on post-impact control based on collision dynamics models is currently the mainstream solution for regaining control of vehicles after collisions. In theory, it allows for the analysis of the vehicle's post-collision state and has good interpretability. However, due to the high risk, cost, difficulty in measuring collision parameters, and inability to conduct large-scale experiments in vehicle collision tests, the four-degree-of-freedom vehicle collision dynamics model has only been simulated and lacks physical validity verification. There is a significant difference between the model parameters and real values, and the research cannot progress to the stage of physical deployment. The physical validation of vehicle collision dynamics models and post-collision controllers is the greatest obstacle to post-impact control research.

To address the issue of validating the efficacy of collision dynamics models, the principal research conducted in this paper involves the empirical validation of an eight-degree-of-freedom vehicle collision dynamics model. Building upon this validated model and a deep reinforcement learning framework, a new post-impact control strategy is developed to enhance vehicular safety post-impact. Initially, an eight-degree-of-freedom vehicle collision dynamics model is constructed based on Newton's second law and rigid body dynamics. This model combines a four-degree-of-freedom vehicle collision dynamics model with a single wheel dynamics model, employing a nonlinear tire model to enhance accuracy under extreme conditions. To overcome the challenge of physically validating the dynamics model, a scaled model car with four-wheel independent drive is designed and constructed. Designed following the principle of dimensionless scaling, this scaled model car exhibits good dynamic similarity to a full-sized vehicle. To precisely record the vehicle's motion state, a motion capture system is utilized to capture the vehicle's pose, and a collision force detection and measurement system is designed to gauge the forces acting on the vehicle during collisions. The model's validity is verified through multiple operational scenarios including driving, braking, and collision impacts. Lastly, leveraging the eight-degree-of-freedom vehicle collision dynamics model and the Twin Delayed DDPG (TD3) deep reinforcement learning algorithm, a new post-impact control strategy is developed. For the vehicle dynamics simulation environment tailored to collision scenarios, the paper integrates the eight-degree-of-freedom vehicle collision dynamics model and simplifies the collision process based on the characteristics of collision forces. In the design of the TD3-based control strategy, a generalized Bellman equation is used to design the reward function, employing a multi-stage progressive training strategy and randomizing the collision forces in the training scenarios. This approach ensures rapid network convergence and ensures that the final strategy is highly generalizable. Through Simulink-Carsim joint simulation tests in various collision scenarios, the controller's performance is evaluated, confirming that the control strategy effectively enhances vehicular safety post-impact.

[1] ORGANIZATION W H, et al. Global plan for the decade of action for road safety 2021-2030[R]. WHO Regional Office for the Western Pacific, 2022.
[2] ZHOU J. Active safety measures for vehicles involved in light vehicle-to-vehicle impacts[M]. University of Michigan, 2009.
[3] MCGEHEE D V, MAZZAE E N, BALDWIN G S. Driver reaction time in crash avoidance research: Validation of a driving simulator study on a test track[C]//Proceedings of the human factors and ergonomics society annual meeting: volume 44. Sage Publications Sage CA: Los Angeles, CA, 2000: 3-320.
[4] YANG D, GORDON T J, JACOBSON B, et al. Quasi-linear optimal path controller applied to post impact vehicle dynamics[J]. IEEE transactions on intelligent transportation systems, 2012, 13(4): 1586-1598.
[5] NHTSA. Automated driving systems 2.0: A vision for safety[J]. Washington, DC: US Department of Transportation, DOT HS, 2017, 812: 442.
[6] National Automotive Sampling System and Crash worthiness Data System, Case ID: 179012709, Case Number: 2009-09-088[J/OL]. NASS CDS (2004-2015) search - NHTSA Crash Viewer, 2022. https://crashviewer.nhtsa.dot.gov/LegacyCDS/Search.
[7] KIM B. Optimal Vehicle Motion Control to Mitigate Secondary Crashes after an Initial Impact. [D]. University of Michigan, 2015.
[8] 郭景华, 李克强, 罗禹贡. 智能车辆运动控制研究综述[J]. 汽车安全与节能学报, 2016, 7 (2): 151-159.
[9] 胡云峰, 曲婷, 刘俊, 等. 智能汽车人机协同控制的研究现状与展望[J]. 自动化学报, 2019, 45(7): 1261-1280.
[10] AO D, HUA X, YU G, et al. Robust active post-impact motion control for restraining a second crash[C]//2020 IEEE 16th International Conference on Automation Science and Engineering (CASE). IEEE, 2020: 159-164.
[11] CAO M, WANG R, CHEN N. Integrated feedback compensation control and model predictive control with AFS for secondary collisions mitigation after an initial impact[C]//2018 Annual American Control Conference (ACC). IEEE, 2018: 5542-5547.
[12] YANG D, JACOBSON B, JONASSON M, et al. Closed-loop controller for post-impact vehicle dynamics using individual wheel braking and front axle steering[J]. International Journal of Vehicle Autonomous Systems, 2014, 12(2): 158-179.
[13] KIM B, PENG H. Collision strength estimation and preemptive steering control for post-impact vehicle motion control[C]//12th Int. Sympo. Adv. Veh. Control. 2014: 496-503.
[14] MOK Y M, ZHAI L, WANG C, et al. A Post Impact Stability Control for Four Hub-Motor Independent-Drive Electric Vehicles[J/OL]. IEEE Transactions on Vehicular Technology, 2022, 71(2): 1384-1396. DOI: 10.1109/TVT.2021.3136186.
[15] WANG C, WANG Z, ZHANG Z, et al. Integrated post-impact planning and active safety control for autonomous vehicles[J/OL]. IEEE Transactions on Intelligent Vehicles, 2023: 1-13. DOI: 10.1109/TIV.2023.3236150.
[16] YIN Y, LI S E, LI K, et al. Self-learning drift control of automated vehicles beyond handling limit after rear-end collision[J]. Transportation Safety and Environment, 2020, 2(2): 97-105.
[17] ZHAO T, YURTSEVER E, CHLADNY R, et al. Collision avoidance with transitional drift control[C]//2021 IEEE International Intelligent Transportation Systems Conference (ITSC). IEEE, 2021: 907-914.
[18] 王星宇. 汽车碰后稳定性控制方法研究[D]. 吉林大学, 2020.
[19] ZHOU J, PENG H, LU J. Collision model for vehicle motion prediction after light impacts[J]. Vehicle System Dynamics, 2008, 46(S1): 3-15.
[20] KIM B, PENG H. Optimal vehicle motion control to mitigate secondary crashes after an initial impact[C]//Dynamic Systems and Control Conference: volume 46186. American Society of Mechanical Engineers, 2014: V001T10A002.
[21] BABU V, THOMSON K R, SAKATIS C. LS-DYNA 3D interface component analysis to predict FMVSS 208 occupant responses[R]. SAE Technical Paper, 2003.
[22] SOLANKI K, OGLESBY D, BURTON C, et al. Crashworthiness Simulations Comparing PAM-CRASH and LS-DYNA[R]. SAE Technical Paper, 2004.
[23] PARSEH M, NYBACKA M, ASPLUND F. Motion planning for autonomous vehicles with the inclusion of post-impact motions for minimising collision risk[J]. Vehicle System Dynamics, 2022: 1-27.
[24] 郑何妍, 卢耀辉, 张德文, 等. 汽车正面耐碰撞性有限元仿真分析[J]. 重庆理工大学学报（自然科学）, 2018.
[25] 邹铁方, 张勇刚, 陈元新. 基于 Pc-Crash 的车辆侧滑事故再现方法[J]. 中国安全科学学报, 2013(1): 77-82.
[26] KIM B J, PENG H. Vehicle stability control of heading angle and lateral deviation to mitigate secondary collisions[C]//11th International Symposium on Advanced Vehicle Control. 2012: 1-6.
[27] CHEN G, ZHAO X, GAO Z, et al. Dynamic drifting control for general path tracking of autonomous vehicles[J]. IEEE Transactions on Intelligent Vehicles, 2023.
[28] JIA F, JING H, LIU Z. A novel nonlinear drift control for sharp turn of autonomous vehicles [J]. Vehicle System Dynamics, 2023: 1-21.
[29] WANG C, WANG Z, ZHANG L, et al. Post-impact motion planning and tracking control for autonomous vehicles[J]. Chinese Journal of Mechanical Engineering, 2022, 35(1): 54.
[30] BERGMANN D P, DENZEL J, BADEN A, et al. Innovative scaled test platform e-genius-mod-scaling methods and systems design[J]. Aerospace, 2019, 6(2): 20.
[31] PERETZ A, EINAV O, HASHMONAY B A, et al. Development of a laboratory-scale system for hybrid rocket motor testing[J]. Journal of Propulsion and Power, 2011, 27(1): 190-196.
[32] WENJIN W, GUANXUE W, BEN L, et al. Control system design and experiment for largescale high-speed unmanned underwater vehicle[J]. Chinese Journal of Ship Research, 2020, 15 (2): 95-103.
[33] PFITSCH D, GORDON B, RICE J, et al. Development and deployment of autonomous scale submarine models for hydrodynamic testing of US Navy submarine maneuvering characteristics [C]//OCEANS 2016 MTS/IEEE Monterey. IEEE, 2016: 1-6.
[34] GOLDFAIN B, DREWS P, YOU C, et al. Autorally: An open platform for aggressive autonomous driving[J]. IEEE Control Systems Magazine, 2019, 39(1): 26-55.
[35] WANG R, CHEN Y, FENG D, et al. Development and performance characterization of an electric ground vehicle with independently actuated in-wheel motors[J]. Journal of Power Sources, 2011, 196(8): 3962-3971.
[36] KATZOURAKIS D I, PAPAEFSTATHIOU I, LAGOUDAKIS M G. An open-source scaled automobile platform for fault-tolerant electronic stability control[J]. IEEE Transactions on Instrumentation and Measurement, 2010, 59(9): 2303-2314.
[37] LAPAPONG S, GUPTA V, CALLEJAS E, et al. Fidelity of using scaled vehicles for chassis dynamic studies[J]. Vehicle System Dynamics, 2009, 47(11): 1401-1437.
[38] DOMBERG F, WEMBERS C C, PATEL H, et al. Deep drifting: Autonomous drifting of arbitrary trajectories using deep reinforcement learning[C]//2022 International Conference on Robotics and Automation (ICRA). IEEE, 2022: 7753-7759.
[39] PARSEH M, NYBACKA M, ASPLUND F. Motion planning for autonomous vehicles with the inclusion of post-impact motions for minimising collision risk[J]. Vehicle system dynamics, 2023, 61(6): 1707-1733.
[40] GOH J Y, GOEL T, CHRISTIAN GERDES J. Toward automated vehicle control beyond the stability limits: drifting along a general path[J]. Journal of Dynamic Systems, Measurement, and Control, 2020, 142(2): 021004.
[41] CUTLER M, HOW J P. Autonomous drifting using simulation-aided reinforcement learning [C]//2016 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2016: 5442-5448.
[42] CAO M, HU C, WANG J, et al. Adaptive complementary filter-based post-impact control for independently-actuated and differentially-steered autonomous vehicles[J]. Mechanical Systems and Signal Processing, 2020, 144: 106852.
[43] LIEBEMANN E, MEDER K, SCHUH J, et al. Safety and performance enhancement: The Bosch electronic stability control (ESP)[J]. SAE Paper, 2004, 20004(2004): 21-0060.
[44] SCHOENEBURG R, BREITLING T. Enhancement of active and passive safety by future PRESAFE systems[C]//Proceedings of the 19th ESV Conference, Washington, DC, USA. 2005.
[45] YANG D, GORDON T J, JACOBSON B, et al. A nonlinear post impact path controller based on optimised brake sequences[J]. Vehicle System Dynamics, 2012, 50(sup1): 131-149.
[46] LI Z, GAO C, ZHU Z, et al. Post-Impact Control to Mitigate the Secondary Collision by Combining LQR with Feed-Forward Control[C]//2022 International Conference on Advanced Robotics and Mechatronics (ICARM). IEEE, 2022: 937-942.
[47] 申棋仁, 代凯, 蒲永锋, 等. 四轮驱动及其融合技术发展综述[J]. 汽车文摘, 2020, 7.
[48] GANGOPADHYAY B, DASGUPTA P, DEY S. Safe and Stable RL (S 2 RL) Driving Policies Using Control Barrier and Control Lyapunov Functions[J]. IEEE Transactions on Intelligent Vehicles, 2022, 8(2): 1889-1899.
[49] LEE D, KWON M. Stability Analysis in Mixed-Autonomous Traffic With Deep Reinforcement Learning[J]. IEEE Transactions on Vehicular Technology, 2022, 72(3): 2848-2862.
[50] CUTLER M, HOW J P. Efficient reinforcement learning for robots using informative simulated priors[C]//2015 IEEE international conference on robotics and automation (ICRA). IEEE, 2015: 2605-2612.
[51] AHMIC K, ULTSCH J, BREMBECK J, et al. Reinforcement Learning-Based Path following Control with Dynamics Randomization for Parametric Uncertainties in Autonomous Driving [J]. Applied Sciences, 2023, 13(6): 3456.
[52] PACEJKA H. Tire and vehicle dynamics[M]. Elsevier, 2005.
[53] DUGOFF H, FANCHER P S, SEGEL L. An analysis of tire traction properties and their influence on vehicle dynamic performance[J]. SAE transactions, 1970: 1219-1243.
[54] HANG P, CHEN X. Towards autonomous driving: Review and perspectives on configuration and control of four-wheel independent drive/steering electric vehicles[C]//Actuators: volume 10. MDPI, 2021: 184.
[55] MORERA TORRES E. Real-Data-Based Modelling and Torque Vectoring Algorithm for a 4-Wheel-Drive Formula Student Vehicle[D]. Universitat Politècnica de Catalunya, 2020.
[56] KARAMAN S, ANDERS A, BOULET M, et al. Project-based, collaborative, algorithmic robotics for high school students: Programming self-driving race cars at MIT[C/OL]//2017IEEE Integrated STEM Education Conference (ISEC). 2017: 195-203. DOI: 10.1109/ISEC on.2017.7910242.
[57] 李玲. 车辆稳定性五自由度模型的有效性验证及车队稳定时距预测[D]. 吉林大学, 2017.
[58] WENSING P M, WANG A, SEOK S, et al. Proprioceptive actuator design in the mit cheetah: Impact mitigation and high-bandwidth physical interaction for dynamic legged robots[J]. IEEE transactions on robotics, 2017, 33(3): 509-522.
[59] LI Z, CHEN S, GAO C, et al. Lenny-Lee-ustb’s 4WD-CAR-ROMA-ros2 Repository[EB/OL].2024. https://github.com/Lenny-Lee-ustb/4WD-CAR-ROMA-ros2.
[60] SIERRA C, TSENG E, JAIN A, et al. Cornering stiffness estimation based on vehicle lateral dynamics[J]. Vehicle System Dynamics, 2006, 44(sup1): 24-38.
[61] VOSER C, HINDIYEH R Y, GERDES J C. Analysis and control of high sideslip manoeuvres [J]. Vehicle System Dynamics, 2010, 48(S1): 317-336.
[62] WONG J Y. Theory of ground vehicles[M]. John Wiley & Sons, 2022.
[63] BRENNAN S. Similarity conditions for comparing closed-loop vehicle roll and pitch dynamics [C]//Proceedings of the 2004 American Control Conference: volume 4. IEEE, 2004: 3393-3398.
[64] 李玲. 车辆稳定性五自由度模型的有效性验证及车队稳定时距预测[D]. 吉林大学, 2017.
[65] WANG Y, TAO W, NAN Z, et al. A Passive Optical Motion Capture Method towards Occlusion Conditions Based on Multi-vision System[C]//Proceedings of the 2022 10th International Conference on Information Technology: IoT and Smart City. 2022: 67-73.
[66] NHTSA. National Highway Traffic Safety Administration: Vehicle Crash Test Database [EB/OL]. 2024. https://www.nhtsa.gov/research-data/research-testing-databases#/vehicle.
[67] ZOIS H, APEKIS L, OMASTOVÁ M. Electrical properties of carbon black-filled polymer composites[C]//Macromolecular Symposia: volume 170. Wiley Online Library, 2001: 249-256.
[68] JAKOBI N, HUSBANDS P, HARVEY I. Noise and the reality gap: The use of simulation in evolutionary robotics[C]//Advances in Artificial Life: Third European Conference on Artificial Life Granada, Spain, June 4–6, 1995 Proceedings 3. Springer, 1995: 704-720.
[69] SILVER D, LEVER G, HEESS N, et al. Deterministic policy gradient algorithms[C]// International conference on machine learning. Pmlr, 2014: 387-395.
[70] LILLICRAP T P, HUNT J J, PRITZEL A, et al. Continuous control with deep reinforcement learning[A]. 2015.
[71] THRUN S, SCHWARTZ A. Issues in using function approximation for reinforcement learning [C]//Proceedings of the 1993 connectionist models summer school. Psychology Press, 2014: 255-263.
[72] FUJIMOTO S, HOOF H, MEGER D. Addressing function approximation error in actor-critic methods[C]//International conference on machine learning. PMLR, 2018: 1587-1596.
[73] 尹冲. 基于强化学习的智能车辆横向控制研究[D]. 湖南大学, 2022.