Diverse multi-path planning with a path-set costmap

This paper addresses a planning method to generate well distributed multiple paths in a control space. For this purpose, we employ and combine rapidly-exploring random tree (RRT), evolutionary algorithm (EA) to compose a diverse multi-path planning (DMPP) algorithm. A population is composed of individuals which represent a path-set. Each individual includes a predefined number of feasible path generated by the RRT, one of the sampling-based planners. The proposed method works by building a population with a set of the predefined number of feasible paths by using the RRT, one of the sampling-based planners. As evolving the population with nature selection and genetic operators, more distributed set of the paths can be acquired. The proposed algorithm leads each path element of path-sets to diverge from each other gradually, so that feasible and different paths are well-generated. In order to evaluate the quality and diversity of a path-set, the costmap approach on path elements are also proposed. Experimental results show that the proposed multi-path planning method works well for generating a set of the diverse paths.