Approximate Policy Iteration with a Policy Language Bias: Solving Relational Markov Decision Processes

TL;DR

This paper introduces a novel approximate policy iteration method using a relational policy language and a new bootstrapping routine, effectively solving large relational MDPs with sparse rewards.

Contribution

It presents a new API variant that learns in policy space with a relational language and a bootstrapping method for goal-based domains, addressing large relational MDP challenges.

Findings

Successfully solves classical planning domains and stochastic variants

Demonstrates effectiveness in large relational MDPs with sparse rewards

Identifies limitations and future directions for the approach

Abstract

We study an approach to policy selection for large relational Markov Decision Processes (MDPs). We consider a variant of approximate policy iteration (API) that replaces the usual value-function learning step with a learning step in policy space. This is advantageous in domains where good policies are easier to represent and learn than the corresponding value functions, which is often the case for the relational MDPs we are interested in. In order to apply API to such problems, we introduce a relational policy language and corresponding learner. In addition, we introduce a new bootstrapping routine for goal-based planning domains, based on random walks. Such bootstrapping is necessary for many large relational MDPs, where reward is extremely sparse, as API is ineffective in such domains when initialized with an uninformed policy. Our experiments show that the resulting system is able to find good policies for a number of classical planning domains and their stochastic variants by solving them as extremely large relational MDPs. The experiments also point to some limitations of our approach, suggesting future work.