PuRR-RRT*: Behavioral Path Planning in Uncertain, Risky and Rewarding Environments
Aamodh Suresh, Carlos Nieto, Sonia Martı́nez · 2023
Effective robotic deployment in uncertain risky and rewarding environments demand diverse reasoning and planning capabilities from robots. In this work, we propose a novel behavioral planning algorithm to navigate in such environments with continuous and uncertain sources of risks and rewards. Agents can express a variety of different behaviors, leading to various environment assessments and correspondingly different planned paths. We take inspiration from behavioral decision making models from Cumulative Prospect Theory (CPT), to construct a class of novel perceived loss functions to capture these different behaviors. We then incorporate these perceived losses into path costs and leverage sampling based planning techniques from RRT*. Our planner Perceived uncertain Risk and Reward RRT* (PuRR-RRT*) plans asymptotically optimal paths, consistent with any given behavioral profile, resulting in a diverse AI for path planning. We then illustrate the proposed algorithm in virtual experiments conducted in a ROS-Unity environment embedded with risk and reward sources. We show that our proposed planner is capable of producing a larger range of diverse paths, aligned with the required behavior.