A quasi Hybrid A* method is introduced for motion planning of autonomous driving car, based on MPC and Bezier spline
In this algorithm, the following factors are considered:
- Non-Holonomic feature of the vehicle
- Speed of the vehicle
- Start heading and steering angle
- Goal heading
- Grid size is 50mx50m
Non-Holonomic feature of the vehicle is taken into consideraion, so the path is continuous and drivable instead of discrete in the conventinoal A* algorithm.
Due to the usage of Bezier spline, no expand grid or heuristic layer is pre-computed as in Hybrid A*, making it very efficient.
The speed of vehicle is considered, thus it can minimize the error using model predictive control.
The pictures below show that the algorithm can plan a path according the starting/goal heading and steering. Also it can avoid obstacles and manage to return to Bezier spline as the reference path. The search tree (green dotted lines) is pruned due to the advantage of th Bezier spline.
Obstacle avoidance
Search tree
