CARLA-Round: A Multi-Factor Simulation Dataset for Roundabout Trajectory Prediction

TL;DR

CARLA-Round addresses the challenge of roundabout trajectory prediction by introducing a systematically designed multi-factor simulation dataset built in CARLA. It explores a $5\times5$ factorial design over weather conditions and traffic density, with a mixed driving-behavior population, yielding 25 controlled scenarios and 449 trajectories with $93{,}803$ frames and rich semantic annotations. Baseline experiments show that traffic density dominates prediction difficulty in a monotonic fashion, while adverse weather has non-linear effects; real-world validation yields $\mathrm{ADE}=0.312$ m on rounD Type 0 with substantially less training data, demonstrating meaningful sim-to-real transfer. The dataset is publicly released to enable rigorous factor analyses and reproducible research in roundabout trajectory prediction.

Abstract

Accurate trajectory prediction of vehicles at roundabouts is critical for reducing traffic accidents, yet it remains highly challenging due to their circular road geometry, continuous merging and yielding interactions, and absence of traffic signals. Developing accurate prediction algorithms relies on reliable, multimodal, and realistic datasets; however, such datasets for roundabout scenarios are scarce, as real-world data collection is often limited by incomplete observations and entangled factors that are difficult to isolate. We present CARLA-Round, a systematically designed simulation dataset for roundabout trajectory prediction. The dataset varies weather conditions (five types) and traffic density levels (spanning Level-of-Service A-E) in a structured manner, resulting in 25 controlled scenarios. Each scenario incorporates realistic mixtures of driving behaviors and provides explicit annotations that are largely absent from existing datasets. Unlike randomly sampled simulation data, this structured design enables precise analysis of how different conditions influence trajectory prediction performance. Validation experiments using standard baselines (LSTM, GCN, GRU+GCN) reveal traffic density dominates prediction difficulty with strong monotonic effects, while weather shows non-linear impacts. The best model achieves 0.312m ADE on real-world rounD dataset, demonstrating effective sim-to-real transfer. This systematic approach quantifies factor impacts impossible to isolate in confounded real-world datasets. Our CARLA-Round dataset is available at https://github.com/Rebecca689/CARLA-Round.

Figures (4)

• Figure 1: Representative scenarios from CARLA-Round dataset showing Town03 roundabout under different conditions: (a) Clear Noon: baseline conditions with full visibility, (b) Wet Noon: post-rain wet surface, (c) Clear Sunset: low sun angle creating glare effects, (d) Hard Rain: heavy precipitation with reduced visibility.
• Figure 2: Overview of the CARLA-Round pipeline.
• Figure 3: Data distribution across weather conditions and traffic density levels.
• Figure 4: Factor impact on prediction difficulty.