| Driven by the development of information technology and intelligence in the automotive industry,research into autonomous driving technology is becoming increasingly prevalent.With the development of big data,cloud computing,the Internet of Things and other information technology,artificial intelligence technology represented by deep neural networks(DNN)has been developed rapidly and has significantly crossed the "technology gap" between science and application.Technologies such as autonomous driving have achieved a technological breakthrough from "unusable and not working well" to "usable".Among the many application scenarios for autonomous driving technology,the nearperfect handling of obstacles in the road by autonomous vehicles is a necessary capability for the metamorphosis of the vehicle automation level from L4 to L5.This research is based on the subjective dynamics(obstacle behaviour decision,obstacle avoidance trajectory planning and tracking control)of dealing with obstacles in the breakthrough phase of autonomous driving technology L4 to L5.In order to simplify the complexity of trajectory planning and vehicle control,the Frenet framework is first introduced and the coordinate transformation between the state quantities is reasoned for the problem of obstacle avoidance behaviour decision and obstacle avoidance trajectory planning.To derive an optimal solution for the obstacle avoidance trajectory,the decision problem of obstacle avoidance behaviour is transformed into a convex optimization problem,and a convex space is opened up with the help of dynamic programming to derive a coarse solution for the obstacle avoidance planning trajectory.The coarse solution of an optimised obstacle avoidance trajectory in convex space using quadratic programming yields an obstacle avoidance trajectory that satisfies the expectations of safety,comfort and efficiency.In order to solve the problem of tracking the trajectory of obstacle avoidance,the control strategy of decoupling the lateral and longitudinal motions is used in order to solve the complexity of the vehicle’s lateral and longitudinal motions,the difficulty of guaranteeing the real-time control,and the mutual interference of the lateral and longitudinal motions.After the vehicle control layer receives the obstacle avoidance trajectory information,in order to control the expected obstacle avoidance trajectory of the vehicle tracking,a lateral and longitudinal control strategy based on the optimized PID algorithm and LQR algorithm is constructed.In order to verify the effectiveness of the obstacle avoidance decision and trajectory planning algorithms and control strategies,this study combines the features of each of the three software packages Pre Scan,Car Sim and Matlab/Simulink to carry out a highly realistic joint simulation test and to visualise the real-time process of the simulation test.The test results show that the final solution of the obstacle avoidance planning trajectory is derived based on dynamic and quadratic planning,and that obstacle avoidance can be accomplished safely and efficiently through the action of the lateral and longitudinal control strategy.The joint simulation test results thus verify the realism and reliability of the algorithms and control strategies in this paper.Finally,a game theory-based optimisation strategy is proposed to address the "change-over" behaviour of the dynamic planning algorithm in the joint simulation test.The basic idea of optimisation is to break the "upward and downward" relationship between the decision planning and control levels and to give the control level the right to make intelligent choices.In the construction of an optimisation model based on game strategies,a dynamic game type with incomplete information is first adapted to the vehicle after a rational analysis,and the game strategy of the insider is developed with the help of a signal game model and a game decision tree.Considering the complexity and feasibility of the game optimisation strategy,the whole game process is divided into P1,P2 and P3,three stages.In the P3 game stage,considering the limitations of the insider’s access to information and the irrational behaviour in the actual game,an evolutionary game is proposed to optimise the insider’s game strategy and finally achieve the optimisation of the vehicle obstacle avoidance and trajectory tracking control strategy. |
