Scaling Pareto-Efficient Decision Making Via Offline Multi-Objective RL
It addresses the challenge of scaling Pareto-efficient decision-making for agents with unknown preferences in offline MORL, which is incremental as it builds on Decision Transformers and dataset creation.
The paper tackles the problem of learning preference-agnostic policies in offline multi-objective reinforcement learning (MORL) using only finite datasets of demonstrations, by introducing D4MORL datasets with 1.8 million annotated demonstrations and proposing PEDA algorithms that approximate the Pareto-front with high hypervolume and low sparsity metrics.
The goal of multi-objective reinforcement learning (MORL) is to learn policies that simultaneously optimize multiple competing objectives. In practice, an agent's preferences over the objectives may not be known apriori, and hence, we require policies that can generalize to arbitrary preferences at test time. In this work, we propose a new data-driven setup for offline MORL, where we wish to learn a preference-agnostic policy agent using only a finite dataset of offline demonstrations of other agents and their preferences. The key contributions of this work are two-fold. First, we introduce D4MORL, (D)atasets for MORL that are specifically designed for offline settings. It contains 1.8 million annotated demonstrations obtained by rolling out reference policies that optimize for randomly sampled preferences on 6 MuJoCo environments with 2-3 objectives each. Second, we propose Pareto-Efficient Decision Agents (PEDA), a family of offline MORL algorithms that builds and extends Decision Transformers via a novel preference-and-return-conditioned policy. Empirically, we show that PEDA closely approximates the behavioral policy on the D4MORL benchmark and provides an excellent approximation of the Pareto-front with appropriate conditioning, as measured by the hypervolume and sparsity metrics.