Research On Model Predictive Control-Based Local Path Planning Algorithm For Unmanned Vehicles

In order to cope with a series of traffic that environmental and other problems brought about by the increasing number of motor vehicles every year,unmanned vehicles have become Research hotspots in recent years in the context of the increasingly mature active safety technology of vehicles and the rapid development of Internet technology.Local path planning that is one of the key technologies of driverless vehicles directly determines the success of driverless vehicles.This thesis has conducted a series of studies on the local path planning of driverless vehicles in consideration of Vehicle dynamics and kinematics constraints,structured road constraints,traffic rules constraints,and computational efficiency,and designed Local path planning algorithm for driverless vehicles based on model prediction principles.Relying on experiment platform designed by the research group,the relevant content of this thesis will be tested in experiment platform.The research content of this thesis is as follows:(1)The current status of domestic and foreign driverless vehicle research and typical path planning methods in recent years are investigateed,summarizing path planning methods are summarizied into five categories,and analyzed these characteristics and advantages and disadvantages.(2)Based on the 2-degree-of-freedom vehicle dynamics model,a non-linear model of the vehicle with the front wheel angle as the input and the vehicle heading angle as the output is established to achieve discretization and serve as the prediction model for the model predictive control algorithm.At the same time,on the basis of the two-wheeled vehicle model,the trajectory tracking method of Pure Pursuit is analyzed,and the vehicle lateral control algorithm based on Pure Pursuit and the longitudinal speed tracking algorithm based on PID are designed.(3)The grid method is used to model the driving environment of the vehicle,and the feasible domain solution method of the principle of convex approximate obstacle avoidance is introduced to pre-process the grid environment model to obtain the feasible domain constraints of the model predictive control to improve the efficiency of the algorithm.(4)the objective function is formulated,Combining the principles of model predictive control,considering the shortest path,road alignment,successive lane changes,lateral acceleration,front wheel rotation angle change rate and other factors,and the quadratic planning problem of path optimization based on MPC is derived,and Solve using sequence-based quadratic programming.(5)Using the convenience of Simulink algorithm development and the diversity of LabVIEW hardware acquisition interfaces,a driverless vehicle experiment platform based on Simulink + LabVIEW hybrid architecture was built,and the path planning algorithm was experimentally verified to verify the path following / The effectiveness of the algorithm under the three road traffic conditions of straight road with obstacles / curve with obstacles.