# An Intrinsically-Motivated Approach for Learning Highly Exploring and Fast Mixing Policies

## TL;DR

This paper introduces a new approach for learning exploration policies that maximize state-action coverage quickly, using a surrogate objective based on entropy lower bounds, and demonstrates its effectiveness in challenging tasks.

## Contribution

It proposes a novel entropy-based surrogate objective for exploration, along with a model-based RL algorithm, IDE$^{3}$AL, to optimize highly exploring, fast mixing policies.

## Key findings

- The proposed method effectively learns highly exploring policies.
- It outperforms baseline methods on hard-exploration tasks.
- The approach balances theoretical guarantees with computational feasibility.

## Abstract

What is a good exploration strategy for an agent that interacts with an environment in the absence of external rewards? Ideally, we would like to get a policy driving towards a uniform state-action visitation (highly exploring) in a minimum number of steps (fast mixing), in order to ease efficient learning of any goal-conditioned policy later on. Unfortunately, it is remarkably arduous to directly learn an optimal policy of this nature. In this paper, we propose a novel surrogate objective for learning highly exploring and fast mixing policies, which focuses on maximizing a lower bound to the entropy of the steady-state distribution induced by the policy. In particular, we introduce three novel lower bounds, that lead to as many optimization problems, that tradeoff the theoretical guarantees with computational complexity. Then, we present a model-based reinforcement learning algorithm, IDE$^{3}$AL, to learn an optimal policy according to the introduced objective. Finally, we provide an empirical evaluation of this algorithm on a set of hard-exploration tasks.