Model-Based Offline Planning

TL;DR

This paper introduces Model-Based Offline Planning (MBOP), a method that learns models from offline data to enable planning-based control, achieving near-optimal policies and zero-shot goal conditioning in robotics tasks without system interaction.

Contribution

The paper presents a novel offline learning approach that generates controllable models for planning, improving interpretability and constraint handling over model-free policies.

Findings

Achieves near-optimal policies with minimal real-system interaction.

Demonstrates zero-shot goal-conditioned policy generation.

Effectively leverages planning to respect environmental constraints.

Abstract

Offline learning is a key part of making reinforcement learning (RL) useable in real systems. Offline RL looks at scenarios where there is data from a system's operation, but no direct access to the system when learning a policy. Recent work on training RL policies from offline data has shown results both with model-free policies learned directly from the data, or with planning on top of learnt models of the data. Model-free policies tend to be more performant, but are more opaque, harder to command externally, and less easy to integrate into larger systems. We propose an offline learner that generates a model that can be used to control the system directly through planning. This allows us to have easily controllable policies directly from data, without ever interacting with the system. We show the performance of our algorithm, Model-Based Offline Planning (MBOP) on a series of robotics-inspired tasks, and demonstrate its ability leverage planning to respect environmental constraints. We are able to find near-optimal polices for certain simulated systems from as little as 50 seconds of real-time system interaction, and create zero-shot goal-conditioned policies on a series of environments. An accompanying video can be found here: https://youtu.be/nxGGHdZOFts