Hao Wang and Simon J. Julier
Motion Planning, Hazardous environments
Many path planning algorithms assume that the environment is either perfectly known or perfectly unknown (in which case the environment is assumed to be empty). We consider the intermediate case in which partial prior information, in the form of a probabilistic occupancy cell map, is available. Using heuristics for clustering the spatial distribution of paths based on most common routes, we derive an algorithm, called the PD planner, which exploits this probabilistic information. Unlike local entropy-based planners, it can account for global effects such as the need for a robot to “Back Up” if it becomes stuck in a blind alley. The performance of the algorithm is assessed in a simulated highly damaged indoor scenario, where we show that exploiting the global impacts of uncertainty has the potential to significantly reduce both travel time and travel distance.
Important Links:
Go Back
