Saute RL: Almost Surely Safe Reinforcement Learning Using State Augmentation
About
Satisfying safety constraints almost surely (or with probability one) can be critical for the deployment of Reinforcement Learning (RL) in real-life applications. For example, plane landing and take-off should ideally occur with probability one. We address the problem by introducing Safety Augmented (Saute) Markov Decision Processes (MDPs), where the safety constraints are eliminated by augmenting them into the state-space and reshaping the objective. We show that Saute MDP satisfies the Bellman equation and moves us closer to solving Safe RL with constraints satisfied almost surely. We argue that Saute MDP allows viewing the Safe RL problem from a different perspective enabling new features. For instance, our approach has a plug-and-play nature, i.e., any RL algorithm can be "Sauteed". Additionally, state augmentation allows for policy generalization across safety constraints. We finally show that Saute RL algorithms can outperform their state-of-the-art counterparts when constraint satisfaction is of high importance.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Safety-constrained Reinforcement Learning | Safety-Gym SafetyPointCircle1 (evaluation) | Average Reward39.6 | 15 | |
| Velocity Control | Safety Gymnasium Hopper | Return1.66e+3 | 8 | |
| Velocity Control | Safety Gymnasium Swimmer | Return111.6 | 8 | |
| Navigation | Safety Gymnasium PointCircle1 | Return42.4 | 8 | |
| Navigation | Safety Gymnasium PointPush1 | Return0.3 | 8 | |
| Velocity Control | Safety Gymnasium Ant | Return2.71e+3 | 8 | |
| Velocity Control | Safety Gymnasium HalfCheetah | Return2.79e+3 | 8 | |
| Safe Reinforcement Learning | Bullet Safety Gym AntRun (eval environment) | Reward652 | 8 | |
| Navigation | Safety Gymnasium PointPush2 | Return0.1 | 8 | |
| Navigation | Safety Gymnasium CarPush2 | Return0.3 | 8 |