Adaptive Clustering Quasi-Line Search Path Planning Algorithm Based On Sampling

In this paper, the adaptive clustering quasi-line search path planning algorithm (ACPP) is proposed from the viewpoint of non-convex optimization in an artificial potential field (APF). In ACPP, the initial path is the trajectory of the optimization process from the initial point to the target one. In a complex environment, the drivable area is divided by the edges into many relatively isolated regions, which leads to a non-convex path planning problem. Therefore, the clustering based on path nodes is set adaptively to make the potential function of each isolated region convex in the potential field. In each isolated region, the environment can be perceived by updating the parameters with sampling based on probability distribution functions, and these functions have similar influence to the cost function in optimization. Inspired by convex optimization, the quasi-line search algorithm is proposed for sampling points. Therefore, ACPP reduces the number of samples, dramatically, and has advantages of both sampling-based and APF-based algorithms in path planning. Based on the line segment based map which can be obtained from sensors readily, a gridding strategy is used to further reduce the time complexity. A series of simulation and experiential results validate the effectiveness of ACPP in virtual and real-world environments, respectively.