CU-Multi: A Dataset for Multi-Robot Collaborative Perception

2025-09-23

AI Summary

Multi-robot collaborative perception suffers from a critical bottleneck: the absence of realistic, standardized benchmark datasets. Existing collaborative semantic-aware mapping (C-SAM) evaluations predominantly rely on single-robot trajectory segmentation, failing to capture the intrinsic nature of team coordination and yielding non-comparable results. To address this, we introduce the first large-scale, multi-robot dataset explicitly designed for collaborative perception evaluation. Data were collected synchronously across multiple days and sites using RGB-D cameras, semantic LiDARs, RTK-GPS, and high-precision odometry. We implement a controllable trajectory-overlap mechanism and provide dense, frame-level semantic annotations, achieving millisecond-level temporal synchronization and centimeter-level spatial alignment. Four high-quality synchronized sequences are released, covering diverse dynamic scenarios and including ground-truth trajectories. This dataset bridges a fundamental gap in reproducible, verifiable evaluation for multi-robot collaborative perception, significantly enhancing the fairness and reliability of algorithmic benchmarking.

Technology Category

• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Multiagent Systems: Coordination and Collaboration
• Computer Vision: Multi-modal Vision

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Security and Privacy: Data transparency and provenance

Abstract

A central challenge for multi-robot systems is fusing independently gathered perception data into a unified representation. Despite progress in Collaborative SLAM (C-SLAM), benchmarking remains hindered by the scarcity of dedicated multi-robot datasets. Many evaluations instead partition single-robot trajectories, a practice that may only partially reflect true multi-robot operations and, more critically, lacks standardization, leading to results that are difficult to interpret or compare across studies. While several multi-robot datasets have recently been introduced, they mostly contain short trajectories with limited inter-robot overlap and sparse intra-robot loop closures. To overcome these limitations, we introduce CU-Multi, a dataset collected over multiple days at two large outdoor sites on the University of Colorado Boulder campus. CU-Multi comprises four synchronized runs with aligned start times and controlled trajectory overlap, replicating the distinct perspectives of a robot team. It includes RGB-D sensing, RTK GPS, semantic LiDAR, and refined ground-truth odometry. By combining overlap variation with dense semantic annotations, CU-Multi provides a strong foundation for reproducible evaluation in multi-robot collaborative perception tasks.

Problem

• Fusing independently gathered perception data into unified representation
• Benchmarking hindered by scarcity of dedicated multi-robot datasets
• Existing datasets lack standardization and sufficient trajectory overlap

Innovation

• Multi-day synchronized dataset with controlled overlap
• Combines RGB-D, RTK GPS, and semantic LiDAR sensing
• Provides dense semantic annotations for reproducible evaluation