Research On Trajectory Planning And Motion Control Algorithms For Autonomous Vehicle In Unstructured Scenes

The application scenarios of autonomous vehicles can be divided into structured scenarios and unstructured scenarios.Compared with structured scenes,unstructured scenes lack the constraints of lane lines and traffic rules.In these scenes,traffic participants have a higher degree of driving freedom,and their motion modes are difficult to be predicted accurately.Therefore,the problem of trajectory planning is more complicated.Meanwhile,the narrow and disordered characteristics of unstructured scenes also put forward higher requirements on the robustness and control accuracy of motion control algorithms.In response to the above problems,this thesis focuses on the trajectory planning and motion control of autonomous vehicles in unstructured scenarios.The main research content includes the following aspects:(1)The trajectory has a spatio-temporal dimension and consists of two parts: path and speed.It is more complicated to solve the trajectory planning problem directly.Therefore,this article decomposes it into two parts: path planning and speed planning.The path planning procedure is responsible for static obstacles.A planning strategy based on graph search algorithm and numerical optimization is adopted.The initial path is generated according to the hybrid A* algorithm,which aims to satisfy the vehicle kinematics and workspace constraints.However,the path is usually not smooth enough.Therefore,an optimal control proposition for path planning tasks is constructed,in which safety and smoothness are considered in the objective function.And precise constraints that meet obstacle avoidance and vehicle kinematics are formulated in the analytical form.The local optimization of the initial path is accomplished by solving this problem.The speed planning procedure deals with dynamic obstacles.A speed planning strategy of decision-making and optimization is designed.The S-T diagram is introduced to describe the relationship between the ego vehicle and dynamic obstacles.Therefore,the speed planning problem can be transformed into a path search problem in the S-T diagram.In the decisionmaking part,the A* algorithm is applied to search for an S-T trajectory in the grid graph,which contains the speed decision information.In the speed optimization part,a segmented sampling algorithm is designed in the S-T diagram.The sampling points at each moment maintain the same driving intention as the result of decision,and the sampling points at different moments are connected by fifth-degree polynomials.The numerical optimization algorithm is adopted to find the polynomial coefficients.And then,a set of feasible speed profiles is obtained.The speed profile with the lowest cost is searched in the collection based on the dynamic programming algorithm.(2)In the motion control procedure,to reduce the dimensionality of the system model,the control of autonomous vehicle is decoupled into the vertical control in the - direction and the horizontal control in the -7)direction based on the Frenet coordinate system.In the longitudinal control module,to follow the desired speed,a hierarchical control structure is designed.The upper controller calculates the expected acceleration based on the model predictive control algorithm.In the lower controller part,the desired throttle opening and braking pressure are calculated based on the principles of vehicle dynamics.In the lateral control module,to realize the tracking of the desired path,the MPC algorithm is adopted.The lateral error model in the Frenet coordinate system is derived based on the twodegree-of-freedom dynamic model.And it can be used as the predictive model of MPC.The lateral deviation can be reasonably approximated by analyzing the positional relationship between the vehicle and the reference trajectory point.The coupling effect of the tire force in the vertical and horizontal directions is considered,and it is equivalent to the boundary constraint of the control quantity.The cost function is designed to minimize lateral deviation.The problem is derived as a constrained quadratic programming problem so that the front wheel rotation angle is calculated in real-time.(3)Taking the local roads in the residential area of the South Campus of Chang’an University as the simulation scene,combining the planning layer and the control layer,the proposed algorithm is comprehensively simulated and verified in terms of safety,smoothness,control accuracy and robustness.