Network Randomization: A Simple Technique for Generalization in Deep Reinforcement Learning
This addresses generalization issues in deep RL for high-dimensional state spaces, offering a simple but effective solution for tasks such as robotics and exploration.
The paper tackles the problem of deep reinforcement learning agents failing to generalize to unseen environments by proposing network randomization, a technique that perturbs input observations to learn robust features, and it significantly outperforms other methods on tasks like 2D CoinRun and 3D robotics control.
Deep reinforcement learning (RL) agents often fail to generalize to unseen environments (yet semantically similar to trained agents), particularly when they are trained on high-dimensional state spaces, such as images. In this paper, we propose a simple technique to improve a generalization ability of deep RL agents by introducing a randomized (convolutional) neural network that randomly perturbs input observations. It enables trained agents to adapt to new domains by learning robust features invariant across varied and randomized environments. Furthermore, we consider an inference method based on the Monte Carlo approximation to reduce the variance induced by this randomization. We demonstrate the superiority of our method across 2D CoinRun, 3D DeepMind Lab exploration and 3D robotics control tasks: it significantly outperforms various regularization and data augmentation methods for the same purpose.