MGB-HA *: Multi-Goal Bi-directional Hybrid A*

The problem of multi-goal path planning is to quickly find a smooth and safe path that satisfies kinematic constraints over multiple target locations, and to avoid getting trapped in repeated searches for previously arrived goals during the search process. We propose a Multi-Goal Bi-directional Hybrid A*(MGB-HA*) method to solve this problem. It incorporates the distance information between the nodes and the latest arrived goal by designing the actual and heuristic function, which greatly reduces the repeated search for the goals that have been arrived before. We also employ a bi-directional search strategy and use dubins curve to connect the bi-directional search encounter to reduce the search time. Finally, the path optimization is solved using L-BFGS, then the optimal path can be generated which effectively reduces unnecessary turns and is more suitable for robot path tracking. The simulation experiment results show that compared with the basic Extend HA*, the number of nodes is reduced by 50.5% and the time is reduced by 72.9%; compared with HA*, the time is reduced by 36.1% in small map and 39.6% in big map.