A Harmonized Approach: Beyond-the-Limit Control for Autonomous Vehicles Balancing Performance and Safety in Unpredictable Environments
- 1 July 2024
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 25 (11) , 15827-15840
Abstract
This paper introduces an adaptive beyond-the-limit controller, aimed at striking a balance between high-performance maneuvers, such as transient drift, and ensuring safety in unpredictable environments. Our work is motivated by the necessity for autonomous beyond-the-limit control adaptable to real-world uncertainties, where reinforcement learning (RL) faces simulation-to-reality gap challenges in safety and performance. Our approach introduces a hybrid control mechanism that integrates data-driven performance optimization with a robust safety-centric control policy. By leveraging expert demonstrations and employing Jump-Start RL framework in Frenet coordinates, we greatly improve the learning efficiency of performance optimization. Further, an integrated safety control policy is designed to mitigate hazards through predictive trajectory planning, thus significantly reducing the risk of accidents in unforeseen situations. Meanwhile, the hybrid control mechanism employs adaptive weighting between performance and safety considerations, allowing for fusion control based on real-time environmental assessments. Through simulation experiments and initial real-vehicle testing, we validate the effectiveness of our adaptive hybrid controller. The findings confirm that our controller consistently ensures integrated safety in unpredictable environments, with an acceptable impact on performance.Keywords
Funding Information
- National Key Research and Development Program of China (Number, 2022YFB2503103)
This publication has 33 references indexed in Scilit:
- Radar sensor based machine learning approach for precise vehicle position estimationScientific Reports, 2023
- An efficient planning method based on deep reinforcement learning with hybrid actions for autonomous driving on highwayInternational Journal of Machine Learning and Cybernetics, 2023
- Autonomous driving at the handling limit using residual reinforcement learningAdvanced Engineering Informatics, 2022
- Multidirectional motion coupling based extreme motion control of distributed drive autonomous vehicleScientific Reports, 2022
- Dynamic Vehicle Drifting With Nonlinear MPC and a Fused Kinematic-Dynamic Bicycle ModelIEEE Control Systems Letters, 2021
- Deep reinforcement learning-based safe interaction for industrial human-robot collaboration using intrinsic reward functionAdvanced Engineering Informatics, 2021
- Minimum curvature trajectory planning and control for an autonomous race carVehicle System Dynamics, 2019
- Drift control for cornering maneuver of autonomous vehiclesMechatronics, 2018
- Survey of Model-Based Reinforcement Learning: Applications on RoboticsJournal of Intelligent & Robotic Systems, 2017
- Staying within the nullcline boundary for vehicle envelope control using a sliding surfaceVehicle System Dynamics, 2012