Drift Cornering Control and Real-Vehicle Deployment for Electric Vehicles
archive: archived pipeline: cataloged verified
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This paper addresses the challenge of deploying reinforcement learning (RL) controllers for high-speed drift cornering on real-world autonomous electric vehicles. While RL has shown promise in simulations, significant discrepancies between simulated and real-world conditions—such as unpredictable road surfaces, sensor noise, and actuator constraints—have hindered practical application. The authors propose a hybrid control framework that integrates trajectory optimization with an RL-based controller and a Model Predictive Controller (MPC) to bridge the simulation-to-reality gap. This approach aims to minimize cornering time while ensuring stability and safety during extreme maneuvers. The methodology consists of two main phases: controller design and real-vehicle deployment. For controller design, the authors employ Bézier-based pre-trajectory optimization to generate smooth paths with minimum curvature, which serves as a reference for reward shaping in the RL training process. They utilize the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm, training it in a simulated environment using CarSim for high-fidelity vehicle dynamics. The reward function is designed to encourage adherence to the pre-optimized trajectory, maximize side-slip angles for drift intensity, and minimize total cornering time. For real-world deployment, the authors implement an RL-MPC fusion mechanism. The TD3-derived maneuvers act as primary inputs, while a two-step forward MPC provides corrective inputs to track the optimal trajectory precisely, compensating for real-world uncertainties and actuator limitations. The study validates the proposed framework through real-vehicle tests on consumer-grade electric vehicles, focusing on U-turns and right-angle turns. Simulation results demonstrate that Bézier-based trajectory optimization reduces cornering times compared to using road centerlines; for instance, cornering times for 90-degree, 135-degree, and U-turns were reduced from 2.94s to 2.81s, 4.59s to 4.09s, and 4.96s to 4.52s, respectively. The real-vehicle tests successfully executed these drift maneuvers, confirming the controller's ability to handle transient drift states and complex curves in unpredictable environments. The RL-MPC fusion effectively mitigated the simulation-to-reality gap, allowing for stable and precise trajectory tracking despite environmental variability and actuator constraints. This work is significant as it represents the first deployment of a scenario-to-action RL-based transient drift cornering algorithm on consumer-grade electric vehicles. By successfully bridging the gap between simulation and reality, the study advances the practical application of reinforcement learning in autonomous driving, particularly for high-performance scenarios like autonomous racing. The proposed hybrid control framework offers a viable solution for enhancing vehicle handling and reducing lap times, contributing to the development of L4 and L5 autonomous vehicles capable of executing extreme driving maneuvers safely and efficiently.
Key finding
A hybrid control framework integrating reinforcement learning with model predictive control successfully enables stable drift cornering on consumer-grade electric vehicles by bridging the simulation-to-reality gap.
Methodology
on_road
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.