Study On Shared Control Method For Driver-Automation Cooperative Driving In Highly Automated Vehicles

Cooperative driving is defined as the scheme wherein the driver and automation share the control authority and navigate the vehicle simultaneously through multi-level collaboration.The cooperative driving scheme is believed to be a solution to the safety,ethics and cost issues faced by conventional automated vehicles,and represents the new direction of automated driving technologies.Cooperative driving relies on the shared control method that enables the driver and automation to control the vehicle together.This paper focuses on the shared steering control method in lane-keeping scenarios.There are drawbacks in previous works including unreasonable driver steering model,limited lane-keeping performance and unresolved driver-automation conflicts.This paper is aimed to address these issues in shared control,with the research content divided into three parts:a shared controller designing method,an appropriate driver steering model for shared control with identified parameters,and an approach to optimize the driver comfort in shared control.First,an indirect shared control scheme is presented which can be implemented in active front steering systems or steer-by-wire systems.A weighted-summation method is proposed to blend the desired steering inputs of the driver and automation.A best-response driver steering model for shared control is derived from the conventional predictive driver steering model,along with a generic analytic expression under different shared controllers.Considering the best-response driver steering model,a shared control strategy based on the Nash equilibrium concept and model predictive control(Nash-MPC)is proposed.In the meantime,a fast computation method for the vehicle steady trajectory under Nash strategies is provided.The Nash-MPC controller can achieve better lane-keeping performance in contrast to the Decentralized-MPC controller that does not consider driver-automation interaction.Second,the parameters of the best-response driver steering model are identified.The best-response model under the Decentralized-MPC controller is derived.To identify the model parameters,the knowledge of the driver's desired trajectory is required.Therefore,this paper proposes a linear model to parametrize the driver's desired trajectory in different road configurations,as well as a shared control method based on trajectory-correction to estimate the driver's desired trajectory from his steering input.Experiments were conducted in the driving simulator.The subjects' desired trajectory parameters are first identified using linear regression.Based on the results,the subjects' best-response steering model parameters are identified via the nonlinear least-squares method.The parameter identification result can be used to simulate the best-response driver steering model in shared controller development process.And also,it reveals the steering characteristics of drivers in different levels of cooperative driving.Third,the driver-automation conflict in shared control is analyzed,which is ascribed to the desired trajectory difference between the driver and automation.A shared control strategy based on the Pareto optimality concept and model predictive control(Pareto-MPC)is proposed,which can theoretically optimizes the driver steering effort to minimum.It is proved that the Pareto-MPC controller can be approximated by a Decentralized-MPC controller which tracks the driver's desired trajectory.A desired trajectory adaptation approach is derived using the gradient descent method,which leads to an integral update law of the desired trajectory parameters.Several pragmatic approaches to apply the trajectory adaptation method are discussed.The Pareto-MPC controller can improve driver's comfort in cooperative driving significantly.Lastly,driving simulator experiments were conducted to validate the shared control methods proposed in this paper.Objective performance metrics and subjective questionnaires are designed to evaluate the shared control strategies.The Nash-MPC controller and the Pareto-MPC controller are evaluated and compared in the aspects of lane-keeping performance,driver steering effort,and subjective assessment.