Research On Behavior Realization Method Of Unmanned Vehicle In Park

The development and application of unmanned driving is of great importance for safe travel and the reduction of traffic congestion.Planning a trajectory that satisfies safety,comfort and real-time for a specific driving behaviour as well as precise tracking control of the planned trajectory is the key to the realisation of unmanned driving behaviour.Based on the above-mentioned status quo,this thesis focuses on unmanned vehicle behaviour trajectory planning and behaviour tracking control technology in the park environment with lane changing and overtaking behaviour as examples.Firstly,the vehicle kinematics and dynamics models required for behavioural realisation are established,and the rationality of the models built is verified by joint simulation with Matlab/simulink and Carsim,providing the basis for subsequent research into behavioural planning and tracking content.Secondly,in order to meet the driving safety and comfort of lane changing and overtaking behaviours,a rolling planning method for behavioural trajectories based on horizontal and vertical decoupling under the Frenet coordinate system is proposed.Based on the forward prediction time range and the beginning and end state points,the trajectory is fitted and combined into a trajectory cluster,an elliptical envelope model with variable relative travel speed is used for safety detection,and the trajectory is optimised based on the constraints of comfort,safety and efficiency of the trajectory and periodic rolling planning,until the whole lane changing and overtaking behaviour planning process is completed safely.Furthermore,an improved pure tracking algorithm with MPC algorithm based on dynamics model is proposed to address the problems of low tracking accuracy and poor real-time performance due to abrupt changes in road curvature and lack of consideration of vehicle dynamics model in the trajectory tracking process.The algorithm is improved by adding forward pre-sighting heading information according to the tracking curvature mutation,and adaptive adjustment of the pre-sighting distance is achieved to improve the tracking accuracy.The MPC tracking algorithm is designed based on the dynamics model to address the lack of consideration of vehicle dynamics in the geometric model tracking algorithm,and is validated by simulation experiments using the Carsim simulation platform.Finally,the lane change and overtaking behaviour planning and tracking algorithm is validated by a modified unmanned experimental platform under different environmental conditions where dynamic and static vehicles are present in the road,and compared with the implementation of the driver model lane change and overtaking behaviour to verify that the lane change and overtaking behaviour planning and tracking control algorithm studied in this thesis has some practical application value.