Sim-to-real transfer and reality gap modeling in model predictive control for autonomous driving
DOI: 10.1007/s10489-022-04148-1
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 "reality gap" in autonomous driving development, specifically the discrepancy between simulation results and real-world performance. While simulators like CARLA are essential for validating autonomous systems due to the infinite variability of real-world scenarios, the stochastic nature of physical environments often leads to errors when transferring control algorithms from simulation to reality. The authors propose a framework to quantify this sim-to-real gap for path tracking systems, aiming to validate simulation suites as reliable tools for controller design without requiring extensive real-world prototyping during early development stages. The study employs a Model Predictive Control (MPC) steering controller based on adaptive artificial potential fields. The system was developed and tested using the CARLA simulator and a real-world prototype: a Citroën C4 equipped with an IMU, GPS, and a DC motor for steering actuation. To ensure comparability, the simulated vehicle (a Toyota Prius) was parameterized to match the real car’s geometry and localization precision. The MPC utilizes a kinematic bicycle motion model and a Gaussian potential field that narrows with increasing vehicle speed to optimize trajectory tracking. The controller was evaluated on diverse road structures, including straights, curves, and roundabouts. The reality gap was quantified by comparing lateral error and control action variations between simulation and real-world data using the Pearson correlation coefficient (PCC) and max normalized cross-correlation (MNCC). The results demonstrate that the proposed statistical framework effectively assesses the correlation between simulated and real-world behaviors. The MPC controller, designed with a non-uniform spatial prediction horizon to improve performance, successfully operated in real-time conditions. The analysis revealed that the simulation suite could accurately predict the controller's performance in real traffic, provided the reality gap remained within critical limits. The study also compared the proposed MPC against an Iterative Linear Quadratic Regulator (ILQR)-based controller, further validating the approach. By statistically analyzing data streams from both environments, the authors identified that the simulation closely resembled real-world counterparts, allowing for a full software-based design process. The significance of this work lies in providing a systematic method to quantify sim-to-real transferability for autonomous vehicle control tasks. This approach allows developers to rely on simulation for the majority of the design and validation cycle, reducing the risks, costs, and logistical challenges associated with real-world testing. The framework enables the identification of poorly handled driving scenarios and supports parameter optimization within the simulator. The authors conclude that this methodology can be transferred to other simulators and control implementations, offering a robust tool for ensuring safety and efficiency in the development of autonomous driving systems.
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.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | Crossref | — | — | 1 | 2026-06-25 |
| archive | success | canonical_url | — | — | 1 | 2026-06-26 |
| extract | success | cached | — | — | 2 | 2026-06-26 |
| clean | success | clean | — | — | 1 | 2026-06-25 |
| chunk | success | chunk | — | — | 1 | 2026-06-25 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-25 |
| promote | success | — | — | — | 1 | 2026-06-25 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 1 | 2026-06-26 |
| tag | success | vector_similarity | — | — | 6 | 2026-06-25 |
| verify | success | — | — | — | 1 | 2026-06-26 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.
Information type
What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).
- Methodological Resource: tool software, validation psychometrics
- Theoretical Contribution: computational model