SafeMIL: Learning Offline Safe Imitation Policy from Non-Preferred Trajectories

TL;DR

SafeMIL is a novel offline imitation learning method that leverages non-preferred trajectories to learn a safety-aware policy, avoiding risky behaviors without requiring online interactions or explicit safety costs.

Contribution

We introduce SafeMIL, a new approach that uses multiple instance learning to predict risky state-action pairs from non-preferred trajectories, enhancing safety in offline imitation learning.

Findings

SafeMIL effectively learns safer policies without sacrificing reward performance.

It outperforms baseline methods in safety and effectiveness.

The approach does not require online interaction or explicit safety cost annotations.

Abstract

In this work, we study the problem of offline safe imitation learning (IL). In many real-world settings, online interactions can be risky, and accurately specifying the reward and the safety cost information at each timestep can be difficult. However, it is often feasible to collect trajectories reflecting undesirable or risky behavior, implicitly conveying the behavior the agent should avoid. We refer to these trajectories as non-preferred trajectories. Unlike standard IL, which aims to mimic demonstrations, our agent must also learn to avoid risky behavior using non-preferred trajectories. In this paper, we propose a novel approach, SafeMIL, to learn a parameterized cost that predicts if the state-action pair is risky via Multiple Instance Learning. The learned cost is then used to avoid non-preferred behaviors, resulting in a policy that prioritizes safety. We empirically demonstrate that our approach can learn a safer policy that satisfies cost constraints without degrading the reward performance, thereby outperforming several baselines.