Global Path Planning for Mobile Robots with Different Risk Metrics

This work proposes a method to include risk awareness in selecting the global path for a mobile robot moving in an environment shared with humans. We proactively consider prior information on human density distribution in the environment and the potential severity of an encounter in a particular location. We compute a graph considering one path candidate for each possible route in the environment. Each edge has a deterministic cost given by its average length and a stochastic cost that depends on the number of people encountered while the robot moves on it and the severity of each encounter. Then, we plan on the obtained stochastic graph using different coherent risk metrics (Expected Cost, Worst Case, and Conditional Value at Risk), so that the robot can choose the best path according to the risk sensitivity of the specific application scenario.