Design And Implementation Of Trajectory Planning For Auto-parking System

With the development of society and the progress of human civilization,human beings have more and more requirements for vehicle's level of intelligence,convenience,and safety.In recent decades,intellectualized modern technology is constantly being applied to vehicle,which improved vehicle's performance of all aspects.As a typical representative of the intelligent driving technology,Auto-parking System can effectively improve the efficiency of parking,and reduce the difficulty of parking to a certain extent.It controls the vehicle automatically in the parking process instead of driver's manipulation.Motion control strategy is the heart of decision making for the Auto-parking System,while the motion control strategy is limited to the whole system architecture.The specific techniques required by different architectures are different,and the final results also vary widely.Through the analysis of the Auto-parking System based on reactive architecture and layered architecture,a hybrid architecture is designed and applied to Auto-parking System,which comprehensively improves the whole system.It avoids bad flexibility caused by the requirement of trajectory mathematical models,and overcomes poor dynamical adjustment to the path due to the open-loop and off-line planning.Under this hybrid architecture,a new hierarchical trajectory planning method based on dynamic window approach and wall following strategy is presented.Trajectory planning is primarily decomposed into time-free path planning problem and time-related velocity planning problem.Through analyzing vehicle's Akermann steering characteristics,a wall following strategy is introduced to fulfill path planning without prior knowledge of trajectory models.And a dynamic window approach based on model predictive control is applied to accomplish vehicle's speed planning.The introduced method is capable of making local adjustments dynamically as the dynamic window approach has the characteristic of feedback optimization.The method is implemented in ROS by using C++ language,and the ROS node form programming are tested by Gazebo.The simulation results show that without the design of trajectory mathematical model,the introduced method can also park cars safely and effectively under the vertical parking scene and parallel parking scene.The method also has a certain guiding for vehicle's autonomous navigation under the high constraint conditions in small scale of area.