Autonomous parking trajectory planning method suitable for multiple scenarios

Abstract

The invention discloses an autonomous parking trajectory planning method suitable for multiple scenarios. The autonomous parking trajectory planning method comprises the steps of (1) building a parking map; (2) determining a current vehicle position and a target parking location; (3) determining physical parameters of a vehicle and initializing a virtual protection box expansion coefficient of thevehicle; (4) expressing a kinematic differential equation of the vehicle as a vehicle motion state constraint; (5) establishing a vehicle and obstacle collision avoidance constraint by using a virtual protection frame and the virtual protection box expansion coefficient; (6) establishing state variable and input variable constraints; (7) determining an optimization objective and building an optimization problem; (8) solving the optimization problem by a solver and obtaining an alternative autonomous parking trajectory; and (9) recalculating the virtual protection box expansion coefficient, obtaining the final autonomous parking trajectory if the parking condition is met, or else, updating the virtual protection box expansion coefficient and repeating the steps (4) to (8). By using the autonomous parking trajectory planning method, a collision-free parking trajectory can be planned in multiple scenarios.

Description

technical field [0001] The invention belongs to the field of automatic driving trajectory planning, and in particular relates to an autonomous parking trajectory planning method applicable to multiple scenarios. Background technique [0002] Autonomous parking is an important function in autonomous driving. At present, some commercial vehicles with autonomous parking have appeared on the market. The traditional autonomous parking trajectory planning method mainly includes two links. First, the parking map is divided into network Grid, using search algorithms such as A*, state lattice, etc. to calculate the parking path in the grid, and then use the optimization method to calculate the path speed according to the parking path, and finally form the parking trajectory. This solution is complex and cannot Guaranteed optimality, deviations are prone to occur during the controller implementation phase. [0003] Another method to calculate the parking trajectory is the optimizatio...

AI Technical Summary

Problems solved by technology

[0002] Autonomous parking is an important function in autonomous driving. At present, some commercial vehicles with autonomous parking have appeared on the market. The traditional autonomous parking trajectory planning method mainly includes two links. First, the parking map is divided into network Grid, using search algorithms such as A*, state lattice, etc. to calculate the parking path in the grid, and then use the optimization method to calculate the path speed according to the parking path, and finally form the parking trajectory. This solution is complex and cannot Guaranteed optimality, prone to deviations during the controller implementation phase

[0003] Another method to calculate the parking trajectory is the optimization method based on the vehicle kinematics model. This method uses optimal control to combine the two steps of path planning and pace, and directly calculates the parking trajectory. However, this method generally requires Manually select the number of optimization points, and after discretizing the continuous optimization problem, there is often a constraint violation problem, which eventually leads to collisions between the path and obstacles, causing parking safety hazards

[0004] Currently, there are some methods that use the dual method to calculate the shortest distance between obstacles and vehicles, and some trajectory planning methods treat vehicles and obstacles as polygons, and use the area method to represent vehicles and obstacles to avoid collision constraints. These methods are implemented in In the process, there are still constraints violations in the optimization problem. Although some studies have proposed the method of re-refining optimization points to increase the number of optimization points in the area where constraint violations occur, this method often leads to more complex problems. The time becomes longer, so the formulation of collision avoidance constraints and optimization problems needs to be improved in principle

Images