Training Verifiably Robust Agents Using Set-Based Reinforcement Learning
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866929463139237888 |
|---|---|
| author | Wendl, Manuel Koller, Lukas Ladner, Tobias Althoff, Matthias |
| author_facet | Wendl, Manuel Koller, Lukas Ladner, Tobias Althoff, Matthias |
| contents | Reinforcement learning often uses neural networks to solve complex control tasks. However, neural networks are sensitive to input perturbations, which makes their deployment in safety-critical environments challenging. This work lifts recent results from formally verifying neural networks against such disturbances to reinforcement learning in continuous state and action spaces using reachability analysis. While previous work mainly focuses on adversarial attacks for robust reinforcement learning, we train neural networks utilizing entire sets of perturbed inputs and maximize the worst-case reward. The obtained agents are verifiably more robust than agents obtained by related work, making them more applicable in safety-critical environments. This is demonstrated with an extensive empirical evaluation of four different benchmarks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2408_09112 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Training Verifiably Robust Agents Using Set-Based Reinforcement Learning Wendl, Manuel Koller, Lukas Ladner, Tobias Althoff, Matthias Machine Learning Robotics Systems and Control Reinforcement learning often uses neural networks to solve complex control tasks. However, neural networks are sensitive to input perturbations, which makes their deployment in safety-critical environments challenging. This work lifts recent results from formally verifying neural networks against such disturbances to reinforcement learning in continuous state and action spaces using reachability analysis. While previous work mainly focuses on adversarial attacks for robust reinforcement learning, we train neural networks utilizing entire sets of perturbed inputs and maximize the worst-case reward. The obtained agents are verifiably more robust than agents obtained by related work, making them more applicable in safety-critical environments. This is demonstrated with an extensive empirical evaluation of four different benchmarks. |
| title | Training Verifiably Robust Agents Using Set-Based Reinforcement Learning |
| topic | Machine Learning Robotics Systems and Control |
| url | https://arxiv.org/abs/2408.09112 |