Traffic Control based on CARMA Platform for Maximal Traffic Mobility and Safety
archive: archived pipeline: cataloged verified
Get this paper ↗ (full text — opens at the source; we link to it, we don't host it)
Summary
This research addresses the persistent trade-off between safety and mobility in connected and automated vehicle (CAV) platooning, a challenge exacerbated by dynamic real-world conditions such as heterogeneous traffic, environmental disturbances, and communication imperfections. Traditional control methods rely either on physics-based models, which offer stability but struggle with nonlinearities, or learning-based approaches, which adapt well but lack interpretability and require extensive data. To overcome these limitations, the authors developed the Physics Enhanced Residual Learning (PERL) framework, designed to integrate with the U.S. DOT’s CARMA platform as a tactical-level cooperative longitudinal control component. The goal is to enhance Transportation System Management and Operations (TSMO) by enabling adaptive gap regulation and disturbance mitigation while maintaining system stability. The PERL framework employs a hybrid control architecture consisting of three modules: a vehicle platoon disturbance module, an inherent physical model, and a residual learning model. The physical model serves as the foundational controller, using Model Predictive Control (MPC) to generate reference trajectories based on vehicle dynamics and constraints, ensuring interpretability and baseline stability. The residual learning module utilizes a neural network trained on limited dynamic data to predict and correct residual errors caused by unmodeled dynamics and external disturbances in real time. Unlike Physics-Informed Neural Networks (PINNs) that embed physical laws as training regularizers, PERL treats the physical model as a fixed base and learns residuals separately, allowing for online adaptation without offline retraining. The system was evaluated through high-fidelity simulations and reduced-scale robot car experiments to assess performance under diverse traffic scenarios and disturbances. Results demonstrate that the PERL framework significantly outperforms both pure physics-based and pure learning-based controllers. In simulations, PERL reduced cumulative position and speed errors by more than 50% compared to baseline methods. In scaled platform tests involving robot cars, the framework achieved up to a 99% reduction in errors, exhibiting rapid convergence following external disturbances and robust platoon stability. The neural network successfully compensated for discrepancies between ideal and actual vehicle behaviors, allowing the platoon to maintain precise speed regulation and trajectory adherence despite nonlinearities and friction. The study concludes that PERL effectively mitigates the safety–mobility trade-off by combining the reliability of physics-based control with the adaptability of data-driven learning. This approach supports the deployment of safe, efficient CAV platooning in mixed traffic environments by improving roadway throughput and cooperative driving efficiency. The findings suggest that PERL can strengthen CARMA’s cooperative driving capabilities, offering a scalable solution for future standards in adaptive platoon control systems. By providing a transparent yet adaptive control mechanism, the framework facilitates smoother integration of platooning strategies into broader transportation management operations.
Key finding
The Physics Enhanced Residual Learning framework significantly outperforms pure physics-based and pure learning-based controllers by reducing cumulative position and speed errors by more than 50% in simulations and up to 99% in scaled platform tests.
Methodology
simulation_modeling
Provenance
The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via author_sweep_intake on 2026-05-28.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | author_sweep | — | — | 2 | 2026-05-28 |
| archive | success | canonical_url | — | — | 1 | 2026-06-06 |
| extract | success | cached | — | — | 3 | 2026-06-10 |
| clean | success | clean | — | — | 1 | 2026-06-04 |
| chunk | success | chunk | — | — | 1 | 2026-06-04 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-04 |
| enrich | success | — | — | — | 1 | 2026-05-28 |
| promote | success | — | — | — | 1 | 2026-06-04 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 2 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 15 | 2026-06-11 |
| verify | success | — | — | — | 2 | 2026-06-10 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.