Safe Reinforcement Learning for Autonomous Lane Changing Using Set-Based Prediction
Open accessAbstract
Machine learning approaches often lack safety guarantees, which are often a key requirement in real-world tasks. This paper addresses the lack of safety guarantees by extending reinforcement learning with a safety layer that restricts the action space to the subspace of safe actions. We demonstrate the proposed approach using lane changing in autonomous driving. To distinguish safe actions from unsafe ones, we compare planned motions with the set of possible occupancies of traffic participants generated by set-based predictions. In situations where no safe action exists, a verified failsafe controller is executed. We used real-world highway traffic data to train and test the proposed approach. The evaluation result shows that the proposed approach trains agents that do not cause collisions during training and deployment.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Hofmarcher et al.'s semantic segmentation methodology surpasses several well-known segmentation models including ENet and SegNet, with more than half a percent per-class average IoU and more than 80 per-category average IoU.
“With almost more than half a percent per-class average intersection over union (IoU) and more than 80 per-category average IoU, the methodology surpasses several well-known segmentation models including ENet and SegNet [71], [101].”
From A Review of Trustworthy and Explainable Artificial Intelligence (XAI) (2023) · cites this paper as [101]
Cites
LeCun, Y., Bengio, Y., & Hinton, G. (2015) · Nature · doi:10.1038/nature14539 · Cited by 84,126 · see citing works
Proximal Policy Optimization Algorithms
Schulman, J., Wolski, F., Dhariwal, P., Radford, A., & Klimov, O. (2017) · arXiv · doi:10.48550/arxiv.1707.06347 · Cited by 10,625 · see citing works
Playing Atari with Deep Reinforcement Learning
Mnih, V., Kavukcuoglu, K., Silver, D., Graves, A., Antonoglou, I., Wierstra, D., & Riedmiller, M. (2013) · arXiv · doi:10.48550/arxiv.1312.5602 · Cited by 9,471 · see citing works
Pattern Recognition and Machine Learning
Neal, R. M. (2007) · Technometrics · doi:10.1198/tech.2007.s518 · Cited by 4,629 · see citing works
Categorical Reparameterization with Gumbel-Softmax
Jang, E., Gu, S., & Poole, B. (2016) · arXiv · doi:10.48550/arxiv.1611.01144 · Cited by 3,616 · see citing works
High-Dimensional Continuous Control Using Generalized Advantage Estimation
Schulman, J., Moritz, P., Levine, S., Jordan, M., & Abbeel, P. (2015) · arXiv · doi:10.48550/arxiv.1506.02438 · Cited by 2,103 · see citing works
Krajewski, R. W., Bock, J., Kloeker, L., & Eckstein, L. (2018) · IEEE International Conference on Intelligent Transportation Systems (ITSC) · doi:10.1109/itsc.2018.8569552 · Cited by 1,519 · see citing works
(2015) · Journal of Machine Learning Research · Cited by 1,119 · see citing works
Optimal trajectory generation for dynamic street scenarios in a Frenét Frame
Werling, M., Ziegler, J., Kammel, S., & Thrun, S. (2010) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/robot.2010.5509799 · Cited by 806 · see citing works
An LSTM Network for Highway Trajectory PredictionPreprint
Altch\'e, F., & Fortelle, A. D. L. (2018) · arXiv (Cornell University) · Cited by 663 · see citing works
A General Safety Framework for Learning-Based Control in Uncertain Robotic Systems
Fisac, J. F., Akametalu, A. K., Zeilinger, M. N., Kaynama, S., Gillula, J., & Tomlin, C. J. (2018) · IEEE Transactions on Automatic Control · doi:10.1109/tac.2018.2876389 · Cited by 460 · see citing works
Online Verification of Automated Road Vehicles Using Reachability Analysis
Althoff, M., & Dolan, J. M. (2014) · IEEE Transactions on Robotics · doi:10.1109/tro.2014.2312453 · Cited by 420 · see citing works
Self-citationNavigating Occluded Intersections with Autonomous Vehicles Using Deep Reinforcement Learning
Isele, D., Rahimi, R., Cosgun, A., Subramanian, K., & Fujimura, K. (2018) · 2018 IEEE International Conference on Robotics and Automation (ICRA) · doi:10.1109/icra.2018.8461233 · Cited by 379 · see citing works
Inevitable collision states — a step towards safer robots?
Fraichard, T., & Asama, H. (2004) · Advanced Robotics · doi:10.1163/1568553042674662 · Cited by 326 · see citing works
Reachability-based safe learning with Gaussian processes
Akametalu, A. K., Fisac, J. F., Gillula, J., Kaynama, S., Zeilinger, M. N., & Tomlin, C. J. (2014) · IEEE Conference on Decision and Control (CDC) · doi:10.1109/cdc.2014.7039601 · Cited by 255 · see citing works
Mirchevska, B., Pek, C., Werling, M., Althoff, M., & Boedecker, J. (2018) · IEEE International Conference on Intelligent Transportation Systems (ITSC) · doi:10.1109/itsc.2018.8569448 · Cited by 207 · see citing works
Self-citationLearning Driver Behavior Models from Traffic Observations for Decision Making and Planning
Gindele, T., Brechtel, S., & Dillmann, R. (2015) · IEEE Intelligent Transportation Systems Magazine · doi:10.1109/mits.2014.2357038 · Cited by 203 · see citing works
Safe Reinforcement Learning With Model Uncertainty Estimates
Lütjens, B., Everett, M., & How, J. P. (2019) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2019.8793611 · Cited by 172 · see citing works
Safe Reinforcement Learning with Scene Decomposition for Navigating Complex Urban Environments
Bouton, M., Nakhaei, A., Fujimura, K., et al. (2019) · IEEE Intelligent Vehicles Symposium · doi:10.1109/ivs.2019.8813803 · Cited by 87 · see citing works
Provably-Correct and Comfortable Adaptive Cruise Control
Althoff, M., Maierhofer, S., & Pek, C. (2020) · IEEE Transactions on Intelligent Vehicles · doi:10.1109/tiv.2020.2991953 · Cited by 85 · see citing works
Self-citationSPOT: A tool for set-based prediction of traffic participants
Koschi, M., & Althoff, M. (2017) · IEEE Intelligent Vehicles Symposium · doi:10.1109/ivs.2017.7995951 · Cited by 83 · see citing works
Self-citationSafe Reinforcement Learning on Autonomous Vehicles
Isele, D., Nakhaei, A., & Fujimura, K. (2018) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2018.8593420 · Cited by 70 · see citing works
Provably safe motion of mobile robots in human environments
Liu, S. B., Roehm, H., Heinzemann, C., et al. (2017) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2017.8202313 · Cited by 66 · see citing works
Verifiably Safe Off-Model Reinforcement Learning
Fulton, N., & Platzer, A. (2019) · Lecture notes in computer science · doi:10.1007/978-3-030-17462-0_28 · Cited by 53 · see citing works
Efficient Computation of Invariably Safe States for Motion Planning of Self-Driving Vehicles
Pek, C., & Althoff, M. (2018) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2018.8593597 · Cited by 28 · see citing works
Self-citationVerification and repair of control policies for safe reinforcement learning
Pathak, S., Pulina, L., & Tacchella, A. (2017) · Applied Intelligence · doi:10.1007/s10489-017-0999-8 · Cited by 24 · see citing works
Can automated road vehicles harmonize with traffic flow while guaranteeing a safe distance?
Althoff, M., & Lösch, R. (2016) · IEEE International Conference on Intelligent Transportation Systems (ITSC) · doi:10.1109/itsc.2016.7795599 · Cited by 23 · see citing works
Self-citationAssured Reinforcement Learning for Safety-Critical Applications
Mason, G. R., Călinescu, R., Kudenko⋆, D., et al. (2017) · White Rose Research Online (University of Leeds, The University of Sheffield, University of York) · Cited by 9 · see citing works
Flagged first, then by citations
Cited by
Sorted by citations
A Review of Trustworthy and Explainable Artificial Intelligence (XAI)
Chamola, V., Hassija, V., Sulthana, A. R., Ghosh, D., Dhingra, D., & Sikdar, B. (2023) · IEEE Access · doi:10.1109/access.2023.3294569 · Cited by 352 · see citing works
“With almost more than half a percent per-class average intersection over union (IoU) and more than 80 per-category average IoU, the methodology surpasses several well-known segmentation models including ENet and SegNet [71], [101].”
Method
- Authors
- Krasowski, H., Wang, X., & Althoff, M.
- Affiliations
- Technical University of Munich (DE)
- Venue
- 2020 IEEE 23rd International Conference on Intelligent Transportation Systems (ITSC)
- Publisher
- IEEE
- Published
- 2020-09-20
- Pages
- 1-7
- OpenAlex
- W3115737827
- Cited by
- 76
- References
- 28
- Citations by year
- 2026: 7 · 2025: 9 · 2024: 13 · 2023: 20 · 2022: 18 · 2021: 8 · 2020: 1
- FWCI
- 9.68
- Type
- conference paper
- Language
- en
- Open access
- green
- mediatum.ub.tum.de
- Licence
- Locations
- IEEE International Conference on Intelligent Transportation Systems (ITSC) (publishedVersion)
- mediaTUM – the media and publications repository of the Technical University Munich (Technical University Munich) (submittedVersion, other-oa)
- Topics
- Physical Sciences › Engineering › Automotive Engineering › Autonomous Vehicle Technology and Safety
- Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
- Physical Sciences › Engineering › Safety, Risk, Reliability and Quality › Safety Systems Engineering in Autonomy
- Keywords
- safe reinforcement learning, safety guarantees, collision avoidance, autonomous driving, highway traffic, motion planning
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
