# Continual Reinforcement Learning in 3D Non-stationary Environments

## TL;DR

This paper introduces CRLMaze, a new benchmark for continual reinforcement learning in complex 3D non-stationary environments, and proposes a model-free strategy that adapts without extra supervision.

## Contribution

The paper presents CRLMaze as a novel benchmark and a new model-free continual RL method that handles environment changes without additional signals.

## Key findings

- The proposed method performs competitively against four baselines.
- CRLMaze effectively simulates real-world non-stationary environments.
- The strategy requires no access to previous environmental conditions or supervised signals.

## Abstract

High-dimensional always-changing environments constitute a hard challenge for current reinforcement learning techniques. Artificial agents, nowadays, are often trained off-line in very static and controlled conditions in simulation such that training observations can be thought as sampled i.i.d. from the entire observations space. However, in real world settings, the environment is often non-stationary and subject to unpredictable, frequent changes. In this paper we propose and openly release CRLMaze, a new benchmark for learning continually through reinforcement in a complex 3D non-stationary task based on ViZDoom and subject to several environmental changes. Then, we introduce an end-to-end model-free continual reinforcement learning strategy showing competitive results with respect to four different baselines and not requiring any access to additional supervised signals, previously encountered environmental conditions or observations.