Zero-Shot Reinforcement Learning Under Partial Observability
This work addresses a limitation in zero-shot RL for real-world applications where full observability is often unavailable, though it is incremental as it extends existing memory-based approaches to the zero-shot setting.
The paper tackles the problem of zero-shot reinforcement learning under partial observability, showing that standard methods degrade in such settings and that memory-based architectures improve performance over memory-free baselines in domains with partially observed states, rewards, and dynamics.
Recent work has shown that, under certain assumptions, zero-shot reinforcement learning (RL) methods can generalise to any unseen task in an environment after reward-free pre-training. Access to Markov states is one such assumption, yet, in many real-world applications, the Markov state is only partially observable. Here, we explore how the performance of standard zero-shot RL methods degrades when subjected to partially observability, and show that, as in single-task RL, memory-based architectures are an effective remedy. We evaluate our memory-based zero-shot RL methods in domains where the states, rewards and a change in dynamics are partially observed, and show improved performance over memory-free baselines. Our code is open-sourced via: https://enjeeneer.io/projects/bfms-with-memory/.