Q-Value Weighted Regression: Reinforcement Learning with Limited Data
About
Sample efficiency and performance in the offline setting have emerged as significant challenges of deep reinforcement learning. We introduce Q-Value Weighted Regression (QWR), a simple RL algorithm that excels in these aspects. QWR is an extension of Advantage Weighted Regression (AWR), an off-policy actor-critic algorithm that performs very well on continuous control tasks, also in the offline setting, but has low sample efficiency and struggles with high-dimensional observation spaces. We perform an analysis of AWR that explains its shortcomings and use these insights to motivate QWR. We show experimentally that QWR matches the state-of-the-art algorithms both on tasks with continuous and discrete actions. In particular, QWR yields results on par with SAC on the MuJoCo suite and - with the same set of hyperparameters - yields results on par with a highly tuned Rainbow implementation on a set of Atari games. We also verify that QWR performs well in the offline RL setting.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Social Navigation | CrowdNav Circle Crossing 5 ORCA Pedestrians, Visible | Success Rate62.48 | 10 | |
| Social Navigation | CrowdNav Circle Crossing 5 ORCA Pedestrians Invisible | Success Rate6.68 | 10 | |
| Social Navigation | CrowdNav Circle Crossing Scenario 5 Social Force Pedestrians (Invisible) | Success Rate3.6 | 10 | |
| Social Navigation | CrowdNav Circle Crossing Scenario 5 Social Force Pedestrians (Visible) | Success Rate58.04 | 10 |