| In recent years,as people pay more attention to the active safety of automobiles,intelligent driving assistance systems for automobiles have been extensively developed.Traditional driving assistance systems provide safety guarantees for cars under normal working conditions.Considering the complexity of the real scene,in order to further improve the driving safety of the vehicle,the research on the motion control method of the unmanned vehicle under extreme conditions is particularly important.In the track scene,in order to achieve certain goals,professional drivers will take the initiative to control the vehicle on the edge of extreme sports,and rely on professional driving skills to control the vehicle to drive stably.In the high-speed emergency obstacle avoidance scene,the vehicle will passively enter the extreme sports state.If the vehicle cannot be stabilized in time,it is easy to cause secondary injuries.Regardless of whether the vehicle enters the extreme state actively or passively,if the intelligent driving assistance system can operate the vehicle flexibly under extreme conditions like a professional driver,it can not only improve the safety performance of the vehicle,but also broaden the vehicle control boundary.Drift motion is a typical situation in extreme vehicle motion.Research on vehicle drift motion control is helpful to further understand the motion control mechanism of vehicle extreme conditions.Especially at the moment when unmanned vehicles are ready to go,the research on the drifting motion control methods of unmanned vehicles is of great significance.On the basis of previous research,this paper studies the drift motion control method of unmanned vehicles from the perspective of vehicle dynamics model and artificial intelligence（AI）algorithm,and builds a joint simulation platform to carry out the proposed drift control algorithm.Simulation.The main research contents of the thesis are as follows:（1）First,this article analyzes the drift motion control of unmanned vehicles from the perspective of vehicle dynamics,and designs a drift motion control algorithm for unmanned vehicles based on drift equilibrium.By analyzing the drift balance state,the drift balance value corresponding to different front wheel angles under the expected longitudinal vehicle speed condition is calculated,and the steady state drift process of the vehicle has typical drifts such as front wheel reversal and rear wheel saturation.Movement characteristics.The yaw rate-centroid side slip angle phase plane diagram under different front wheel angles is drawn,and the vehicle state value at the saddle point in the phase diagram is compared with the vehicle state value of the corresponding drift equilibrium state,and the vehicle at the saddle point in the phase diagram is obtained.The state is the conclusion of the drift equilibrium state.On this basis,the drift balance value under the expected path and the expected side slip angle of the center of mass is calculated as the feedforward control quantity,and the linear quadratic regulator（LQR）drift control algorithm is designed to calculate the feedback control quantity.The way of feedback+feedback controls the vehicle to reach the drift equilibrium state,and realizes the steady-state circular drift motion of the unmanned vehicle.The feasibility of the drift control algorithm was verified by establishing a Carsim/Matlab/Simulink co-simulation platform.（2）In order to meet the needs of unmanned vehicle drifting motion control under more complex working conditions,this paper designs an unmanned vehicle drifting motion control algorithm based on drift state estimation.First,the drift motion control algorithm framework based on drift state estimation is introduced in detail,and then the expected vehicle state value（expected heading angular velocity and expected heading angular velocity and expected vehicle state value）are calculated based on the reference path information（path curvature）and the expected vehicle drift state information（expected side slip angle of the center of mass）.Yaw angular acceleration),combined with the vehicle dynamics model and the Magic Formula tire model to derive the expected vehicle drift state value.On this basis,a linear quadratic regulator（LQR）drift control algorithm is designed to realize the drift motion control of unmanned vehicles on more complex roads.The algorithm has both path tracking and drift motion control functions.By establishing a Carsim/Matlab/Simulink co-simulation platform,the feasibility of the drift control algorithm designed in this paper is verified,and compare the robustness of the drift control algorithm in this paper with the drift control algorithm in Reference[81].（3）Considering that it is difficult to establish a simple and accurate vehicle model under extreme conditions,this paper attempts to design an unmanned vehicle drift control strategy based on deep reinforcement learning algorithms from the perspective of artificial intelligence（AI）.First,the theoretical basis of deep reinforcement learning is introduced,and the feasibility of using deep reinforcement learning to realize drifting motion control of unmanned vehicles is analyzed.Then designed a deep reinforcement learning algorithm based on the actor-critic architecture and the Deep Deterministic Strategy Gradient（DDPG）algorithm,selected appropriate environmental variables,and designed it to realize the circle of unmanned vehicles The reward function for drifting motion.In Py Charm,a deep reinforcement learning algorithm file and a file for data transfer and conversion were written,and a Py Charm/Matlab/Simulink/Carsim joint simulation platform was established to train and test the deep reinforcement learning algorithm network.The joint simulation platforms communicate wirelessly through the User Datagram Protocol（UDP）.The automatic repeat training function is realized by writing M scripts in Matlab. |
