Study On Motion Planning And Control Of Autonomous Vehicles Based On Model Predictive Control

As the two major sub-modules of the automatic driving system,the motion planning module and the motion control module play an extremely important role.It directly shows the intelligence of the autonomous vehicle and affects the safety,the ride comfort,the energy-saving performance and the traffic efficiency.To design a fast,reliable and robust motion planning and control framework and algorithm of autonomous vehicle is of great significance.This paper mainly investigates how to reduce the heavy computation burden of current MPC-based motion planning and control framework and algorithm,and how to deal with the vehicle dynamics constraints during motion planning process,which is lack of consideration in current research.The main contents are as follows:(1)A vehicle bicycle model and a linear tire model are established.To reduce the computation burden of model predictive control as much as possible meanwhile ensure the control performance,a bicycle dynamics model suitable for the motion planning and control problem of autonomous vehicle is established based on reasonable assumptions.The nonlinear tire model is approximately linearized through the linear tire model.(2)The transformation and solving process of linear time-varying model predictive control(LTV-MPC)problem is analyzed and deduced.The mechanism of nonlinear model predictive control(NMPC)and linear time-varying model predictive control is investigated.The linearized transformation equation of nonlinear model predictive control model is deduced in detail.The prediction model of LTV-MPC algorithm is deduced in detail.The steps of transforming multi-constrained LTV-MPC optimization problem into quadratic programming problem and the solving equation are deduced in detail.(3)A new motion planning and control framework and algorithm of autonomous vehicles is proposed.In order to reduce the computation amount,to avoid the excessive computational burden caused by nonlinear model predictive control,and to achieve a safe,stable,comfortable and energy-efficient motion planning and control process,the traditional nonlinear multi-control-variable MPC-based motion planning and control framework is decoupled.A rule of longitudinal safety priority is proposed,based on which a prediction method for dynamic safety area planning is proposed.Finally,a novel longitudinal-horizontal-separation motion planning and control framework of autonomous vehicles is proposed.A longitudinal motion planning algorithm is developed based on the multi-dynamics-constrained LTV-MPC and the vehicle longitudinal motion differential model.A longitudinal motion control algorithm is developed based on the feed-forward control method and the vehicle bicycle model consisting of a linear tire model.An integrated lateral motion planning and control algorithm is developed based on the multi-constrained LTV-MPC and the vehicle bicycle model consisting of a linear tire model.(4)Co-simulations between MATLAB/Simulink and CarSim are conducted in some cases including front vehicle cut-in,front vehicle cut-off,vehicle-following,vehicle-chasing,front vehicle emergency braking and lane-change,etc.The results show that the proposed motion planning and control framework and algorithm of autonomous vehicles has good performance in motion planning and control.Simulation under a typical complex lane-change scenario shows the proposed framework and algorithm computes faster than LTV-MPC-based integrated motion planning and control framework and algorithm,and has the nearly same performance as NMPC-based separated motion planning and control framework and algorithm does.