迈向L5级自动驾驶汽车的发展原则
Towards the Unified Principles for Level 5 Autonomous Vehicles
自动驾驶汽车的快速发展给现有交通出行方式带来了全新面貌和潜在挑战。目前,L3 级及以下驾驶辅助系统已经量产,L4 级在特定场景下的一些应用也逐步开发,通过逐渐提高车辆的自动化、智能化程度来不断向完全自动驾驶发展。然而,针对L5 级自动驾驶汽车的发展思路始终未明确,而现有针对L0~L4级自动驾驶发展过程的研发方式主要基于任务驱动来进行特定场景下的功能开发,难以揭示高等级自动驾驶汽车所需解决问题的本质逻辑和物理机制,进而阻碍了迈向L5 级自动驾驶的途径。本文通过探索高等级自动驾驶系统背后的物理机制,并从驾驶的本质出发,采用推理演绎方式,提出'大脑-小脑-器官'协调平衡框架,基于'乌鸦推理+鹦鹉学舌'的混合模式,探索'自主学习+先验知识'的研究范式,实现自动驾驶汽车'自学习、自适应、自超越'特性。从系统、统一、均衡的角度出发,基于最小作用量原理和统一安全场思想,旨在为高等级自动驾驶汽车,尤其是L5 级自动驾驶的研发提供一种全新的研发思路与有效途径。
The rapid advance of autonomous vehicles (AVs) has motivated new perspectives and potential challenges for existing modes of transportation. Currently, driving assistance systems of Level 3 and below have been widely produced, and several applications of Level 4 systems to specific situations have also been gradually developed. By improving the automation level and vehicle intelligence, these systems
can be further advanced towards fully autonomous driving. However, general development concepts for Level 5 AVs remain unclear, and the existing methods employed in the development processes of Levels 0–4 have been mainly based on task-driven function development related to specific scenarios. Therefore, it is difficult to identify the problems encountered by high-level AVs. The essential logical
and physical mechanisms of vehicles have hindered further progression towards Level 5 systems. By exploring the physical mechanisms behind high-level autonomous driving systems and analyzing the essence of driving, we put forward a coordinated and balanced framework based on the brain–cerebellum–organ concept through reasoning and deduction. Based on a mixed mode relying on the crow inference and parrot imitation approach, we explore the research paradigm of autonomous learning and prior knowledge to realize the characteristics of self-learning, self-adaptation, and self-transcendence for AVs. From a systematic, unified, and balanced point of view and based on least action principles and unified safety field concepts, we aim to provide a novel research concept and develop an effective approach for the research and development of high-level AVs, specifically at Level 5.
自动驾驶汽车 / 最小作用量原理 / 行车安全场 / 自主学习 / 基础范式
Autonomous vehicle / Principle of least action / Driving safety field / Autonomous learning / Basic paradigm
| System | Method | Characteristics | Limitations | Level 5 challenges |
|---|---|---|---|---|
| Perception | Single-sensor perception | Consider the physical principle and data type of each sensor | Limited perceptual performance and targeting specific scenarios [16] | Poor perception performance in complex environments No single type of sensor to cover all scenarios |
| Virtual perception | Simultaneous localization and mapping (SLAM) system:build map based on sensor data in virtual environment Vehicle to everything (V2X): collaborative perception | SLAM: dense and complex computation; fail in dramatically changing road conditions; strongly depend on the accuracy of the input perception data V2X: rely on the infrastructure transformation; require high communication performance and high reliability | Dynamic traffic will seriously affect the accuracy and real-time robustness of positioning No uniform standard for infrastructure deployment | |
| Multiple sensor fusion | Realize the complementary advantages of multiple sensors | Fusion perception of severe weather and light changes is unreliable Sensor data is imperfect, inconsistent, and heterogeneous [17] | No fusion algorithm to simultaneously deal with multiple sensor data problems A reliable sensor fault detection and isolation method still needs to be added to deal with sensor failure | |
| Decision | Hierarchical decision-making [18] | Divided into three levels: situation assessment, behavior decision, and motion planning | Consider limited interactivity, uncertainty, or explosion Difficulty in meeting the driver’s expectations Without considering the impact of limited perception and control ability | Dependent on data to adapt to high-dynamic and random real traffic scenarios Limited self-learning ability hinders its decision-making performance in a high-level autopilot Low computational efficiency to apply to high-level AVs |
| End-to-end decision making [19] | Based on machine learning, the vehicle control information is directly output from the input of environment perception information | Simple application scenarios Unclear decision mechanism Limited generalization ability of multiple scenes Lack of consideration of the interaction between traffic participants | Massive and comprehensive data need to be collected in advance for training The uninterpretability makes it a great uncertainty in the application of advanced automatic driving | |
| Control | Lateral control [20] | The vehicle lateral control is realized by means of adaptive control, fuzzy logic control, sliding mode control, etc. | Traditional control methods fail to solve the problem of multiobjective control of AVs | High-level AVs need to improve the safety and efficiency of overall traffic rather than control in single direction |
| Longitudinal control [21] | The vehicle longitudinal control is realized by means of incremental proportion-integration-differentiation (PID) control, adaptive cruise control, etc. | The control effect is sensitive to the modeling accuracy Challenges in solving high-dimensional nonlinear analysis | Complex scenes require many computations in control process | |
| Multiobjective cooperative control [22] | Combined with learning methods (i.e., deep reinforcement learning), multivehicle cooperative control can be realized for complex scenarios considering multiobjects | Realize intelligent adjustment of personalized parameters and achieve multiobjective collaborative control while poor real-time performance occurs in complex scenes The stability and convergence of the model cannot be guaranteed | How to obtain the optimal solution of dynamic multiobjective control for high-level AVs; optimize the operation performance of complex conditions; and obtain the real-time, robust, and optimal control results |
| [1] |
Antsaklis PJ, Rahnama A. Control and machine intelligence for system autonomy. J Intell Robot Syst 2018;91:23–34. |
| [2] |
Fridman L. Human-centered autonomous vehicle systems: principles of effective shared autonomy. 2018. arXiv:1810.01835. |
| [3] |
SAE On-Road Automated Vehicles Standards Committee. J3016. Taxonomy and definitions for terms related to on-road motor vehicle automated driving systems. Washington, DC: SAE International; 2014. |
| [4] |
Kamil F, Tang SH, Khaksar W, Zulkifli N, Ahmad SA. A review on motion planning and obstacle avoidance approaches in dynamic environments. Adv Robot Autom 2015;4(2):1000134. |
| [5] |
Ulbrich S, Reschka A, Rieken J, Ernst S, Bagschik G, Dierkes F, et al. Towards a functional system architecture for automated vehicles. 2017. arXiv:1703.08557. |
| [6] |
Badue C, Guidolini R, Carneiro RV, Azevedo P, Cardoso VB, Forechi A, et al. Selfdriving cars: a survey. 2019. arXiv:1901.04407. |
| [7] |
Dingus TA, Guo F, Lee S, Antin JF, Perez M, Buchanan-King M, et al. Driver crash risk factors and prevalence evaluation using naturalistic driving data. Proc Natl Acad Sci 2016;113(10):2636–41. |
| [8] |
Muhrer E, Vollrath M. The effect of visual and cognitive distraction on driver’s anticipation in a simulated car following scenario. Transp Res Part F 2011;14 (6):555–66. |
| [9] |
Ou YK, Liu YC, Shih FY. Risk prediction model for drivers’ in-vehicle activities— application of task analysis and back-propagation neural network. Transp Res Part F 2013;18:83–93. |
| [10] |
Tas ÖS, Kuhnt F, Zöllner JM, Stiller C. Functional system architectures towards fully automated driving. In: Proceedings of 2016 IEEE Intelligent Vehicles Symposium (IV); 2016 Jun 19–22; Gotenburg, Sweden; 2016. p. 304–9. |
| [11] |
Burns LD. A vision of our transport future. Nature 2013;497(7448):181–2. |
| [12] |
Khastgir S, Dhadyalla G, Birrell S, Redmond S, Addinall R, Jennings P. Test scenario generation for driving simulators using constrained randomization technique. SAE technical paper. Washington, DC: SAE International; 2017. No.:2017-01-1672. |
| [13] |
Hubmann C, Becker M, Althoff D, Lenz D, Stiller C. Decision making for autonomous driving considering interaction and uncertain prediction of surrounding vehicles. In: Proceedings of 2017 IEEE Intelligent Vehicles Symposium (IV); 2017 Jun 11–14; Los Angeles, CA, USA; 2017. |
| [14] |
Urmson C, Anhalt J, Bagnell D, Baker C, Bittner R, Clark MN, et al. Autonomous driving in urban environments: boss and the urban challenge. J Field Robot 2008;25(8):425–66. |
| [15] |
Zhou J, Li P, Zhou Y, Wang B, Zang J, Meng L. Toward new-generation intelligent manufacturing. Engineering 2018;4(1):11–20. |
| [16] |
Rasouli A, Tsotsos JK. Autonomous vehicles that interact with pedestrians: a survey of theory and practice. IEEE Trans Intell Transp Syst 2020;21 (3):900–18. |
| [17] |
Jo K, Kim J, Kim D, Jang C, Sunwoo M. Development of autonomous car—part Ⅰ: distributed system architecture and development process. IEEE Trans Ind Electron 2014;61(12):7131–40. |
| [18] |
Noh S, An K. Decision-making framework for automated driving in highway environments. IEEE Trans Intell Transp Syst 2018;19(1):58–71. |
| [19] |
Shalev-Shwartz S, Shammah S, Shashua A. Safe, multi-agent, reinforcement learning for autonomous driving. 2016. arXiv:1610.03295. |
| [20] |
Sun Z, Huang Z, Zhu Q, Li X, Liu D. High-precision motion control method and practice for autonomous driving in complex off-road environments. In: Proceedings of 2016 IEEE Intelligent Vehicles Symposium (IV); 2016 Jun 19– 22; Gothenburg, Sweden; 2016. p. 767–73. |
| [21] |
Ruan Y, Chen H, Li J. Longitudinal planning and control method for autonomous vehicles based on a new potential field model. SAE technical paper. Washington, DC: SAE International; 2017. No: 2017-01-1955. |
| [22] |
González D, Pérez J, Milanés V, Nashashibi F. A review of motion planning techniques for automated vehicles. IEEE Trans Intell Transp Syst 2016;17 (4):1135–45. |
| [23] |
Spielberg NA, Brown M, Kapania NR, Kegelman JC, Gerdes JC. Neural network vehicle models for high-performance automated driving. Sci Robot 2019;4 (28):eaaw1975. |
| [24] |
Mohanan MG, Salgoankar A. A survey of robotic motion planning in dynamic environments. Robot Auton Syst 2018;100:171–85. |
| [25] |
Heaven D. Why deep-learning AIs are so easy to fool. Nature 2019;574 (7777):163–6. |
| [26] |
Chiang HTL, HomChaudhuri B, Smith L, Tapia L. Safety, challenges, and performance of motion planners in dynamic environments. In: Amato NM, Hager G, Thomas S, Torres-Torriti M, editors. Robotics research. Cham: Springer; 2020. p. 793–808. |
| [27] |
Chen G, Cao H, Conradt J, Tang H, Rohrbein F, Knoll A. Event-based neuromorphic vision for autonomous driving: a paradigm shift for bioinspired visual sensing and perception. IEEE Signal Process Mag 2020;37 (4):34–49. |
| [28] |
Behere S, Torngren M. A functional architecture for autonomous driving. In: Proceedings of the First International Workshop on Automotive Software Architecture (WASA); 2015 May 4–8; Montreal, QC, Canada; 2015. |
| [29] |
Prakash R, Prakash O, Prakash S, Abhishek P, Gandotra S. Global workspace model of consciousness and its electromagnetic correlates. Ann Indian Acad Neurol 2008;11(3):146–53. |
| [30] |
Oizumi M, Albantakis L, Tononi G. From the phenomenology to the mechanisms of consciousness: integrated information theory 3.0. PLoS Comput Biol 2014;10(5):e1003588. |
| [31] |
Tsien H. [On systems engineering]. Changsha: Hunan Science and Technology Press; 1988. Chinese. |
| [32] |
Zhu SC. Towards general artificial intelligence: from big data to big task. In: Proceedings of 2019 BAAI; 2019 Oct 31–Nov 1; Beijing: China National Convention Center; 2019. |
| [33] |
Zhu M, Wang X, Wang Y. Human-like autonomous car-following model with deep reinforcement learning. Transp Res Part C 2018;97:348–68. |
| [34] |
Wang J, Wu J, Zheng X, Ni D, Li K. Driving safety field theory modeling and its application in pre-collision warning system. Transp Res Part C 2016;72: 306–24. |
| [35] |
Aven T. A risk concept applicable for both probabilistic and non-probabilistic perspectives. Saf Sci 2011;49(8–9):1080–6. |
| [36] |
Wang J, Huang H, Li Y, Zhou H, Liu J, Xu Q. Driving risk assessment based on naturalistic driving study and driver attitude questionnaire analysis. Accid Anal Prev 2020;145:105680. |
| [37] |
Siburg KF. The principle of least action in geometry and dynamics. Berlin: Springer-Verlag; 2004. |
| [38] |
Gururajan MP. The lazy universe: an introduction to the principle of least action, by J. Coopersmith. Contemp Phys 2018;59(1):95–6. |
| [39] |
Zheng X, Huang H, Wang J, Xu Q. Behavioral decision-making model based on driving risk assessment for intelligent vehicle development. Comput Aided Civ Infrastruct Eng 2019;36(7):820–37. |
| [40] |
Huang H, Zheng X, Yang Y, Liu J, Liu W, Wang J. An integrated architecture for intelligence evaluation of automated vehicles. Accid Anal Prev 2020;145:105681. |
()
/
| 〈 |
|
〉 |