Online Sensing of Human Steering Intervention Torque for Autonomous Driving Actuation Systems

Abstract

Identification of human driver intervention is important for the safety and performance of autonomous vehicles. Effective sensing of driver steering torque will enable human drivers to take over controls when needed. Due to the special operation mode of autonomous vehicles, both the human driver and the actuator motor apply active torque to the steering wheel simultaneously during the take-over process. This makes it impossible to detect the human interventions by simply using a torque sensor due to the coupled torques. To tackle this new problem, a model-based strategy to estimate human steering intervention torque is presented for autonomous vehicles in this paper. The model of autonomous steering wheel actuation system is established first. Online human steering torque estimation algorithm is then proposed and the stability condition is given to guarantee both the estimation convergence and accuracy. Numerical simulation is performed to validate the feasibility of the proposed estimator based upon three typical steering intervention cases. The influence factors of the estimation performance are discussed, and a trade-off policy is utilized to determine the optimal estimation gain. Furthermore, experimental studies are carried out on a passenger car equipped with an autonomous driving actuation system. Both the simulation and experimental results demonstrate the effectiveness of the proposed approach.