Adaptive Region Boosting method with biased entropy for path planning in changing environment
Risheng Kang, Tianwei Zhang, Hao Tang, Wenyong Zhao · CAAI Transactions on Intelligence Technology · 2016
Path planning in changing environments with difficult regions, such as narrow passages and obstacle boundaries, creates significant challenges. As the obstacles in W-space move frequently, the crowd degree of C-space changes accordingly. Therefore, in order to dynamically improve the sampling quality, it is appreciated for a planner to rapidly approximate the crowd degree of different parts of the C-space, and boost sample densities with them based on their difficulty levels. In this paper, a novel approach called Adaptive Region Boosting (ARB) is proposed to increase the sampling density for difficult areas with different strategies. What's more, a new criterion, called biased entropy, is proposed to evaluate the difficult degree of a region. The new criterion takes into account both temporal and spatial information of a specific C-space region, in order to make a thorough assessment to a local area. Three groups of experiments are conducted based on a dual-manipulator system with 12 DoFs. Experimental results indicate that ARB effectively improves the success rate and outperforms all the other related methods in various dynamical scenarios.