Planning Under Non-Rational Perception of Uncertain Spatial Costs

This work investigates the design of risk-perception-aware motion-planning strategies that incorporate non-rational risk associated with uncertain spatial costs. Our proposed method employs Cumulative Prospect Theory (CPT) to generate a perceived map over given environment. CPT-like risks and path-length metrics are then combined define cost function is compliant requirements asymptotic optimality sampling-based motion planners (RRT*). The modeling power CPT illustrated in theory simulation, along comparison other perception models like Conditional Value at Risk (CVaR). Theoretically, we notion expressiveness for model show CPT's higher than CVaR expected risk. We this translates our path planning setting, where observe planner equipped together simultaneous perturbation stochastic approximation (SPSA) can better approximate arbitrary paths an Additionally, simulation captures rich set meaningful paths, representative different perceptions custom compare performance T-RRT* (a continuous spaces) Risk-RRT* risk-aware dynamic human obstacles) through simulations cluttered environments respectively, showing advantage planner.