A Policy Synthesis-Based Framework for Robot Rescue Decision-Making in Multi-Robot Exploration of Disaster Sites

摘要

In this work we investigate the problem of intelligent decision-making of robot rescue in the context of disaster-site exploration. We consider the performance of a policy synthesis-based framework with a focus on the robustness of the generated policies with respect to uncertainties in the state estimates. Online state estimation is required to execute a policy; however, this must be done with a limited amount of data and a considerable amount of uncertainty. As a result, there is inherently some amount of error when making a decision. By systematically introducing varying amounts of error into the estimates of state parameters, we provide an empirical sensitivity analysis as it relates to the availability of information and impact on mission performance. We show that our policy synthesis-based framework can achieve superior performance, compared to two feasible baseline approaches with respect to the probability of mission failure, when there is up to 15 % error in the estimation of critical state components.