On Planning while Learning

TL;DR

This paper explores the feasibility of planning in partially known environments, highlighting that plan verification can be efficient, but plan generation is often computationally hard, emphasizing the importance of offline planning processes.

Contribution

It analyzes the computational complexity of planning while learning, demonstrating that plan verification is tractable while plan synthesis remains intractable in many cases.

Findings

Plan verification can be performed efficiently in natural systems.

Generating plans algorithmically is generally intractable.

Offline plan-design processes are crucial for practical planning.

Abstract

This paper introduces a framework for Planning while Learning where an agent is given a goal to achieve in an environment whose behavior is only partially known to the agent. We discuss the tractability of various plan-design processes. We show that for a large natural class of Planning while Learning systems, a plan can be presented and verified in a reasonable time. However, coming up algorithmically with a plan, even for simple classes of systems is apparently intractable. We emphasize the role of off-line plan-design processes, and show that, in most natural cases, the verification (projection) part can be carried out in an efficient algorithmic manner.