DreamSmooth: Improving Model-based Reinforcement Learning via Reward Smoothing
This addresses a key challenge in model-based reinforcement learning for sparse-reward tasks, offering a simple improvement with broad applicability.
The paper tackles the problem of reward prediction as a bottleneck in model-based reinforcement learning, especially for sparse rewards, by proposing DreamSmooth, a reward smoothing approach that predicts temporally-smoothed rewards. It achieves state-of-the-art performance on long-horizon sparse-reward tasks in sample efficiency and final performance without compromising on common benchmarks.
Model-based reinforcement learning (MBRL) has gained much attention for its ability to learn complex behaviors in a sample-efficient way: planning actions by generating imaginary trajectories with predicted rewards. Despite its success, we found that surprisingly, reward prediction is often a bottleneck of MBRL, especially for sparse rewards that are challenging (or even ambiguous) to predict. Motivated by the intuition that humans can learn from rough reward estimates, we propose a simple yet effective reward smoothing approach, DreamSmooth, which learns to predict a temporally-smoothed reward, instead of the exact reward at the given timestep. We empirically show that DreamSmooth achieves state-of-the-art performance on long-horizon sparse-reward tasks both in sample efficiency and final performance without losing performance on common benchmarks, such as Deepmind Control Suite and Atari benchmarks.