Safe Reinforcement Learning for Real-World Engine Control Data and Scripts
Fuente:
Zenodo
Enregistré dans:
| Auteurs principaux: | , , , , , |
|---|---|
| Format: | Recurso digital |
| Langue: | anglais |
| Publié: |
Zenodo
2024
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866901910846439424 |
|---|---|
| author | Bedei, Julian Badalian, Kevin Koch, Lucas Winkler, Alexander Schaber, Patrick Andert, Jakob |
| author_facet | Bedei, Julian Badalian, Kevin Koch, Lucas Winkler, Alexander Schaber, Patrick Andert, Jakob |
| contents | <p>This dataset supports our publication, "Safe Reinforcement Learning for Real-World Engine Control". It includes data and scripts for training artificial neural networks used as a reference control strategy, as well as datasets collected during reinforcement learning (RL) policy training on a real-world single-cylinder Homogeneous Charge Compression Ignition (HCCI) engine testbench.</p> <p>The provided resources cover two key RL experiments:</p> <ol> <li>Training the initial control policy for transient load control in direct interaction with the real-world engine.</li> <li>Adapting the policy to increase ethanol energy shares while maintaining safety constraints.</li> </ol> <p>The RL experiments were conducted by applying the Learning and Experiencing Cyclic Interface (<a href="https://github.com/mechatronics-RWTH/lexci-2">LExCI</a>, versions 2.22.0 and before), a free and open-source tool enabling RL with embedded hardware.</p> <p>This dataset enables the reproduction of the artificial neural network-based reference strategy and provides a foundation for analyzing the RL agent’s performance in a safety-critical environment. These resources aim to support further research into applying RL and machine learning to real-world combustion engines.</p> <p>This research was performed as part of the research unit 2401 (FOR2401) “Optimization based Multiscale Control for Low Temperature Combustion Engines” funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) - 277012063. This support is gratefully acknowledged.</p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_14499423 |
| institution | Zenodo |
| language | eng |
| publishDate | 2024 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | Safe Reinforcement Learning for Real-World Engine Control Data and Scripts Bedei, Julian Badalian, Kevin Koch, Lucas Winkler, Alexander Schaber, Patrick Andert, Jakob Reinforcement Learning Deep Deterministic Policy Gradient Safe Learning Transfer Learning Homogeneous Charge Compression Ignition Renewable Fuels <p>This dataset supports our publication, "Safe Reinforcement Learning for Real-World Engine Control". It includes data and scripts for training artificial neural networks used as a reference control strategy, as well as datasets collected during reinforcement learning (RL) policy training on a real-world single-cylinder Homogeneous Charge Compression Ignition (HCCI) engine testbench.</p> <p>The provided resources cover two key RL experiments:</p> <ol> <li>Training the initial control policy for transient load control in direct interaction with the real-world engine.</li> <li>Adapting the policy to increase ethanol energy shares while maintaining safety constraints.</li> </ol> <p>The RL experiments were conducted by applying the Learning and Experiencing Cyclic Interface (<a href="https://github.com/mechatronics-RWTH/lexci-2">LExCI</a>, versions 2.22.0 and before), a free and open-source tool enabling RL with embedded hardware.</p> <p>This dataset enables the reproduction of the artificial neural network-based reference strategy and provides a foundation for analyzing the RL agent’s performance in a safety-critical environment. These resources aim to support further research into applying RL and machine learning to real-world combustion engines.</p> <p>This research was performed as part of the research unit 2401 (FOR2401) “Optimization based Multiscale Control for Low Temperature Combustion Engines” funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) - 277012063. This support is gratefully acknowledged.</p> |
| title | Safe Reinforcement Learning for Real-World Engine Control Data and Scripts |
| topic | Reinforcement Learning Deep Deterministic Policy Gradient Safe Learning Transfer Learning Homogeneous Charge Compression Ignition Renewable Fuels |
| url | https://doi.org/10.5281/zenodo.14499423 |