Path Planning under Risk and Uncertainty of the Environment
Luis Cuevas, Miquel Ramírez, Iman Shames, Chris Manzic · 2021
This manuscript addresses the path planning problem under risk and uncertainty to reduce the chances of undesirable damage or even loss of the autonomous vehicle. Here, it is assumed that the configuration space is partially known and that a set of points is provided to pursue the exploration of the uncertain region. Furthermore, this work considers a set of alternative risk scenarios are available for the uncertain regions of the vehicle workspace. A novel solution framework incorporating stochastic and deterministic optimisation techniques is proposed as a method of handling the tradeoff between exploration and exploitation in the chosen path. The decisions on whether to explore the uncertain region or exploit the known information are made based on a risk/reward function, that is iteratively re-evaluated when new information becomes available under the proposed algorithm. The feasibility of the algorithm is proven and results characterising the cost of exploring the unknown environment are presented here.