Research On Planning And Control Of Human-Machine Lateral Shared Driving Based On Game Authority Allocation

Human-machine lateral shared driving is one of the important technical architectures in the transitional stage of automobile intelligence development,with its characteristic binary intelligent body closed-loop control structure.The driver,as a human intelligence,and the automated steering system,as a machine intelligence,jointly operate the vehicle in a decentralised manner to achieve a driving performance that is superior to the independent operation of a single driver and allows the driver to achieve a degree of benefit.Due to the joint participation of human and machine,the problem of the closed-loop control of binary intelligent body for human-machine lateral shared driving has added some new problems to the original research of intelligent vehicle technology,such as the depth of the humanmachine cooperation level,the competition for driving control authority(human-machine conflict)and the driver’s trust crisis for the automated steering system.In this paper,focusing on the joint decision-making-control problem of humanmachine lateral shared driving,the core task is to establish a human-machine driving control authority allocation strategy and optimize the degree of human-machine cooperation,with the objectives of realizing the normal traffic capacity,reducing the driver’s load of humanmachine lateral sharing driving,and enhancing the driver’s acceptance of human-machine lateral shared driving system.This paper draws on the "V" shaped development process and adopts the research sequence of simulating the algorithm first to establish the humanmachine lateral shared driving’s planning and control algorithm based on the bargaining game theory-based driving control authority allocation strategy.Based on the above ideas,the main research contents of this paper are as follows:(1)In view of the demand for the lateral path tracking control algorithms during the pre-simulation development of human-machine lateral shared driving technology,this paper initially develops an automated steering system path-tracking control algorithm and a driver model that simulates the driver,namely: a model predictive control with pre-scanned driving characteristics(PMPC)path tracking control algorithm and a single-view-angle driver model reflecting the real driving process.In order to achieve the basic technical architecture of human-machine lateral shared driving,this paper separately develops a path-tracking control algorithm for lateral control of the automated steering system and a driver model of simulated driver.Firstly,a PMPC path tracking control algorithm is developed,and a joint simulation program is set up using Car Sim vehicle dynamics simulation software and MATLAB/Simulink mathematical software tools to verify its tracking performance by a double-lane-change manuevear.The result shows that the PMPC path tracking control algorithm with the introduction of prescanned driving characteristics delivers better results in terms of tracking accuracy and work speed performance than the traditional model predictive control path tracking control algorithm.Secondly,the theoretical convergence of the single-view-angle driver model is demonstrated in terms of the mechanism of model building,and the assumption of singleview-angle driver basis is proposed.In the joint simulation environment,the operating results of the developed PMPC control algorithm are used as a reference object,and the prescanned distance of the driver model is adjusted by the least squares method to fit ideal operating results.The simulation result shows that the single-view-angle can be directly translated into the target front wheel steering angle when the pre-scanned distance is adjusted to the ideal value.Therefore,the steering translation control process based on proportional control is determined.Based on this driver model,the theoretical derivation method of pre-scanned distance and an empirical formula for the pre-scanned distance is proposed.Using the joint simulation environment of Simulink/Car Sim and a real intelligent vehicle to verify the performance of this single-view-angle driver model,the result indicates that the control process of this driver model is simple and easy,restoring the driving phenomenon that the single-view-angle is consistent with the steering angle trend,with well control performance and matching real driving process.(2)Aiming the driving control authority allocation of human-machine lateral shared driving at the core problem,a strategy of human-machine lateral driving control authority allocation based on bargaining game model is developed.In order to realize the human-machine lateral shared driving and to manage humanmachine driving conflicts properly,the driver with social properties and the automatic steering system with programmatic properties are regarded as a binary intelligent body system.Game theory is one of the best means of solving the problem of distributing the benefit cake for binary intelligent body systems.First,in this paper,a human-machine bargaining game model is proposed based on the technology of corner interaction.The game model considers anti-collision safety benefits,participation benefits and "humancentred" comfort benefits,and dynamically solves the decision-making results for driving control authority through the Nash equilibrium in real time.Secondly,the evaluation index of driver load is proposed.Finally,the proposed human-machine lateral shared driving control authority allocation strategy is verified based on a joint Simulink/Car Sim cosimulation testing platform and the driver-in-the-loop driving simulator.The result shows that the strategy can achieve well shared driving performance,endow human-machine with the applicability of shared driving under multi-working conditions of normal traffic and enhances the driver’s acceptance of the shared driving system.(3)In order to reduce potential target conflicts in human-machine lateral shared driving and to further increase driver’s acceptance of human-machine shared driving systems,a highly human-like lane-change trajectory planning algorithm is proposed in the planning layer of the automated steering system.From the planning layer of the automated steering system,multi-level involvement of human-machine lateral shared driving is achieved to minimise potential conflicts between human and machine targets as much as possible.Firstly,the driver’s lane-change trajectory data at different speeds is collected using a driving simulator,and the lane-change characteristics of multiple drivers are analyzed: the average lane-change duration and the maximum average slope of the lane-change trajectory.Secondly,a basic lane-change trajectory algorithm based on a quintuple polynomial with lane-change duration as the core lane-change trajectory parameter is developed in the Frenet coordinate system.Thirdly,the trajectory clusters are generated according to the plus or minus five times the standard deviation time intervals of the driver’s mean lane-change duration.Then,trajectories with a maximum theoretical lateral acceleration greater than 0.4 g are removed to ensure the dynamic stability of the trajectory clusters.The optimal planning trajectory is solved with the human-like index,lane-change smoothness index and lane-change efficiency index as the core benefits,and an anti-collision detection algorithm for convex polygons based on the vertex embedding method is proposed to ensure the anti-collision safety of the optimal planning trajectory.Finally,the generated human-like lane-change trajectories are compared with the collected trajectory clusters for human-like verification,and a single-view-angle driver model is used to verify the feasibility of tracking control.The result shows that the planned human-like lane-change trajectory is consistent with the driver’s historical trajectories,and the planned human-like lane-change trajectory with the driver’s personality characteristics meets the requirements of vehicle tracking control execution while ensuring stability and anti-collision safety.(4)In order to verify the algorithm for human-machine lateral shared driving planning and control algorithm strategy based on game authority allocation and human-like lanechange trajectory planning proposed in the previous article,this paper uses the joint Simulink/Car Sim simulation platform and driving simulator to implement the feasibility and effectiveness validation of the algorithm strategy.According to the requirements for the verification of the human-machine lateral shared driving planning and control algorithm strategy based on game authority allocation and human-like lane-change trajectory planning proposed in this paper,the joint Simulink/Car Sim simulation platform is set up to implement the overall simulation verification of the algorithm strategy.In addition,driver-in-the-loop test validation is carried out by a driving simulator.The result shows that the proposed human-machine lateral shared driving planning and control algorithm strategy based on bargaining game authority allocation and human-like lane-change trajectory planning delivers good human-machine shared driving operation results and improves drivers’ acceptance of the human-machine shared driving system.