Wang, Zhaoqing (Tsinghua University), Huang, Jin (Tsinghua University), Xu, Yuhang (Tsinghua University), Wu, Xiaozhou (Tsinghua University), Li, Xueyun (Wuhan University of Technology), Yang, Mengmeng (Tsinghua University), Zhong, Zhihua (Tsinghua University)

A Human–Machine Shared Control Strategy for Automation Disengagement Cases with Intention-Based Driving Authority Allocation

Scheduled for presentation during the Regular Session "Driver Assistance Systems II" (FrBT1), Friday, September 27, 2024, 13:30−13:50, Salon 1

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on December 26, 2024

Abstract

Automation disengagement cases are the situations where an intelligent vehicle encounters unresolvable cases. Current autonomous vehicles usually directly switch the vehicle driving authority to human driver in these cases. This study develops a human–machine shared control strategy to address automation disengagement cases on highways. First, the driver’s driving intention in disengagement cases is recognized through a double-layer long short-term memory deep neural network. Both the driver action and surrounding vehicle data are considered. The a novel driving authority allocation method is proposed, considering the takeover emergency level and the driver’s adaptation level to the shared control system when allocating authority in the automation disengagement takeover procedure. Subsequently, an Udwadia–Kalaba approach-based vehicle nonlinear controller is designed to assist the driver with the driving intention. The nonlinearity of the vehicle lateral dynamics is also thought about. Finally, the proposed shared control strategy is validated by driver-in-the-loop experiments with 12 participants. The results show that the proposed strategy can help drivers better handle automation disengagement cases. Additionally, driver intention can be accurately recognized and driving authority is smoothly transferred to the human driver, which weakens the situational awareness loss problem in the traditional 0–1 takeover pattern.