Predicting Future Actions of Reinforcement Learning Agents

TL;DR

This paper compares methods for predicting future actions of reinforcement learning agents, finding that explicit planning information is most effective for accurate and robust predictions, which enhances safety and interaction in real-world applications.

Contribution

It introduces a comparative analysis of inner state and simulation-based approaches for predicting RL agent actions and events, highlighting the superiority of explicit planning signals.

Findings

Explicit planning agents' plans are highly informative for prediction.

Inner state approaches outperform simulation-based methods in robustness.

Event prediction results are mixed across different approaches.

Abstract

As reinforcement learning agents become increasingly deployed in real-world scenarios, predicting future agent actions and events during deployment is important for facilitating better human-agent interaction and preventing catastrophic outcomes. This paper experimentally evaluates and compares the effectiveness of future action and event prediction for three types of RL agents: explicitly planning, implicitly planning, and non-planning. We employ two approaches: the inner state approach, which involves predicting based on the inner computations of the agents (e.g., plans or neuron activations), and a simulation-based approach, which involves unrolling the agent in a learned world model. Our results show that the plans of explicitly planning agents are significantly more informative for prediction than the neuron activations of the other types. Furthermore, using internal plans proves more robust to model quality compared to simulation-based approaches when predicting actions, while the results for event prediction are more mixed. These findings highlight the benefits of leveraging inner states and simulations to predict future agent actions and events, thereby improving interaction and safety in real-world deployments.