ParkingScenes Dataset for Autonomous Parking in Simulation
ParkingScenes, a new multimodal dataset aimed at facilitating end-to-end autonomous parking in simulated settings, has been unveiled by researchers. Utilizing the CARLA simulator, it features structured parking trajectories produced by a Hybrid A* planner alongside a Model Predictive Controller (MPC). The dataset encompasses 16 scenarios for reverse-in parking and 6 for parallel parking, each evaluated under two pedestrian conditions (present and absent), resulting in a total of 704 structured episodes. This initiative seeks to fill the gap in high-quality datasets specifically designed for parking in tight urban environments.
Key facts
- ParkingScenes is a multimodal dataset for end-to-end autonomous parking.
- Built on the CARLA simulator.
- Trajectories generated by Hybrid A* planner and MPC.
- Includes 16 reverse-in and 6 parallel parking scenarios.
- Each scenario executed with and without pedestrians.
- Total of 704 structured episodes.
- Aims to address lack of parking-specific datasets.
- Designed for constrained urban environments.
Entities
—