Hierarchical task network-based emergency task planning with incomplete information, concurrency and uncertain duration

Abstract

This paper focuses on the emergency task planning problem with the following characteristics: incomplete initial environment information, concurrent execution and uncertain execution durations. To address this problem, the conditional temporal hierarchical task network (HTN) planning with uncertain durations should be investigated. However, the research work on conditional HTN planning mainly concentrates on the planning domain without temporal features and pays no attention to the handling of concurrency and uncertain durations. This motivates us to propose a planning paradigm based on the conditional HTN planning. The planning paradigm incorporates a temporal reasoning technique suitable for conditional HTN planning to achieve concurrency and a temporal management approach to confirm the satisfiability of temporal constraints efficiently. For temporal reasoning, rules for generating temporal constraints are designed to avoid three types of flaws due to interactions between durative actions; a mechanism for detecting invalid temporal constraints is also presented. Regarding temporal management, Conditional Simple Temporal Network with Uncertainty (CSTNU) is utilized to represent the temporal constraints involving uncontrollable intervals and observations, and an approach for eliminating redundant temporal constraints is proposed to reduce the temporal constraint to be represented in a CSTNU to promote efficiency. Finally, an experimental study of an earthquake rescue domain demonstrates the viability of the proposed planning paradigms.