Identifying Achievable Goals for Adaptive Replanning Against Runtime Environment Change

Abstract

In automated planning, a plan is synthesized to achieve the given goals in the assumed operational environment. However, during the plan’s execution, the operational environment may changes so that replanning a new plan is necessary against the changing environment. In some situations, it is impossible to achieve some goals anyhow; in other situations, a new plan can achieve the main goals if it chooses to give up some subsidiary goals. Therefore, the ability of goal-awareness is essential to determine which goals to achieve in the new plan. This paper proposes an efficient analysis algorithm to identify the achievability and conflicts of reachability goals and safety goals. The algorithm reuses the existing analysis result and efficiently analyzes only the part that changes when unforeseen change happens in the environment. This time-saving feature makes our algorithm ideal for rapid response and replanning during the plan’s execution. We evaluated the absolute computation time and relative reduction rate through two case studies. The result shows that our analysis algorithm averagely reduces 97.8% of the calculation time, compared with analyzing without reusing existing analysis results.

The research was partially supported by National Institute of Information and Communications Technology (NICT) and JSPS KAKENHI.