Experimental dataset of video and radar detection for cooperative perception in urban environment

Abstract

Cooperative perception is an emerging concept in intelligent transportation systems that enhances situational awareness by allowing vehicles and infrastructure nodes to share sensor information. By extending the sensing range beyond the line of sight of a single agent, cooperative perception enables safer and more informed decision-making in complex traffic situations. To support research in this area, especially from the perspective of infrastructure-based sensing, high-quality datasets are essential. This article presents a dataset that combines radar and camera-based object detection data in standardized Collective Perception Messages (CPMs), collected in a real vehicular environment. The dataset includes object-level information such as unique tracking identifiers, spatial position, speed, heading, and classification. In addition to the raw sensor detections, it provides message-level CPMs generated in real time by the infrastructure node, following the European Telecommunications Standards Institute (ETSI) Collective Perception Service (CPS) specification and applying its object inclusion rules. All data is timestamped and spatially referenced, enabling the reconstruction of object trajectories and behavior over time. The dataset is suitable for developing and evaluating cooperative perception algorithms, as well as applications like trajectory prediction, object classification refinement, and multi-sensor fusion benchmarking. Its accessibility aims to support the research community in advancing perception and prediction models for autonomous driving and intelligent transportation systems.