Research on Path Planning Problem of Smart Car Considering Kinematic Constraints

Abstract

With the development of artificial intelligence, intelligent cars have gradually appeared in people's lives and brought many conveniences to people's lives and work. On the street, we can see many busy figures of unmanned delivery cars; In some factories, smart cars shuttle back and forth between product production processes for material distribution; In some hospitals, we can also find that some smart cars replace medical staff to undertake the task of sending medicines and tests. For intelligent cars, autonomous collision-free path planning and trajectory tracking of the planned path are the most basic and core tasks of intelligent vehicles. The existing path planning algorithms are mainly divided into search-based algorithms (Dijkstra, A*, etc.) and sample-based algorithms (PRM, RRT, etc.), and different types of algorithms have good performance in their respective applicable scenarios. However, these path planning algorithms all use the same simplification assumption, that is, the intelligent car is regarded as a freely moving point or sphere in the path planning task, and the driving experience in life tells us that the movement of the wheel-driven intelligent car is constrained by the turning radius of the vehicle, so the path planned by the simplified intelligent car for a point is often not suitable for the driving of the intelligent car in reality. Considering this constraint, this project models the kinematics of the intelligent trolley with four-wheel steering, and adds kinematic constraints to the path planning, so as to plan the path that conforms to the kinematics model of the intelligent trolley. The path planning algorithm proposed in this topic will be tested in ROS and Gazebo simulation environments, and compared with the traditional search-based A* algorithm, the results of multiple scenarios verify the effectiveness of the algorithm. Finally, this project summarizes and looks forward to this research, which has completed the construction of low-cost physical intelligent vehicles, and plans to complete the verification of the real vehicle function by combining the designed path planning algorithm with the real-time mapping and positioning system of the car in the future.