CommonRoad: Composable benchmarks for motion planning on roads
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 introduces CommonRoad, a standardized, composable benchmark collection designed to address the lack of reproducibility and comparability in motion planning research for road vehicles. The authors argue that existing literature often omits critical experimental details—such as vehicle dynamics, road networks, and obstacle behaviors—due to space constraints or the absence of a unified data format. While datasets like NGSIM provide traffic recordings, they lack the necessary components for benchmarking, such as defined initial states, goal regions, and dynamic vehicle models. CommonRoad aims to fill this gap by providing a repository where every benchmark is fully defined by a unique ID, allowing researchers to reconstruct experiments precisely without searching for realistic parameters. The methodology centers on a modular composition of benchmarks consisting of three interchangeable components: vehicle models, cost functions, and scenarios. Vehicle models range from simple point-mass approximations to complex multi-body dynamics, including kinematic and standard single-track models that account for tire slip and load transfer. Cost functions are constructed as weighted sums of partial costs, such as time, acceleration, jerk, and lane center offset, allowing for flexible optimization objectives. Scenarios are described using an XML format based on "lanelets," which represent road networks as directed graphs of drivable segments. These scenarios include static and dynamic obstacles, with data sourced from both real-world traffic recordings and hand-crafted dangerous situations to ensure representativeness. The system supports scalability, from simple static environments to complex dynamic interactions, and provides executable implementations in MATLAB and Python to ensure portability. The paper presents numerical experiments comparing the different vehicle models using a BMW 320i parameter set. Results demonstrate significant differences in behavior: the kinematic single-track model produces the tightest turns by ignoring tire slip, while the single-track model accounts for slip, resulting in wider paths. The multi-body model further captures saturation of tire forces and vertical load transfers, exhibiting phenomena like understeering, oversteering, and pitch changes during braking and acceleration. These findings highlight the importance of model fidelity depending on the planning task. The benchmarks are designed to be independent of specific planning libraries, serving as an interchange format that facilitates fair comparison across different algorithms. The significance of CommonRoad lies in its potential to standardize evaluation in the automotive robotics field. By providing open, unambiguous, and composable benchmarks, it enables researchers to reproduce results and compare planning methods objectively. The authors emphasize that while the collection facilitates benchmarking, specific performance metrics should be determined through community consensus. The work establishes a foundation for reproducible science in motion planning, offering a scalable and portable infrastructure that saves researchers time and ensures that experimental conditions are clearly defined and accessible.
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 | OpenAlex-citations | — | — | 1 | 2026-06-20 |
| archive | success | unpaywall | — | — | 2 | 2026-06-26 |
| extract | success | cached | — | — | 2 | 2026-06-26 |
| clean | success | clean | — | — | 1 | 2026-06-20 |
| chunk | success | chunk | — | — | 1 | 2026-06-20 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-20 |
| promote | success | — | — | — | 1 | 2026-06-20 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 1 | 2026-06-26 |
| tag | success | vector_similarity | — | — | 6 | 2026-06-20 |
| 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: dataset resource, tool software