Focusing on the problems such as too many expansion nodes path redundancy points in path planning for the classical A* algorithm

this paper aims to study the improvement of the classical A* algorithm. Firstly

the concept of slant-eight-neighborhood expansion is proposed

and by combining with the four-neighborhood expansion

a novel dual-neighborhood selection expansion strategy is proposed. By this novel strategy

the number of expansion nodes can be greatly reduced in the path search process. In addition

a new heuristic function is established to adapt to a variety of map environments. Hence

the number of expanded nodes can be reduced by above 50% in the same map environment comparing with the classical A* algorithm

and the path search speed can be improved by one order of magnitude

and the efficiency of the algorithm is significantly improved. Then

the initial path is further optimized by establishing a redundant point elimination strategy and the cubic B-spline curve. Thus

an optimal path that meets the robot motion can be obtained by eliminating the redundant nodes and reducing path turns. Finally

the improved A* algorithm is evaluated and analyzed by comparing with the Dijkstra algorithm

four-neighborhood expansion algorithm and eight-neighborhood expansion algorithm by simulation in four maps with different obstacles. And

also the present algorithm is verified by using the field tests based on the intelligent vehicle platform. The results show that by the improved A* algorithm the path search efficiency can improved significantly

the planned path is more suitable for the robot motion. In summary

the present improved A* algorithm is believed to be feasible and effective.