CU-Multi: A Dataset for Multi-Robot Collaborative Perception
Doncey Albin, Daniel McGann, Miles Mena, Annika Thomas, Harel Biggie, Xuefei Sun, Steve McGuire, Jonathan How, Christoffer Heckman
AI summary
Problem
Benchmarking multi-robot collaborative perception is hindered by a lack of standardized datasets, forcing researchers to rely on artificially split single-robot trajectories that fail to capture realistic multi-robot observational overlap and diversity.
Approach
The authors collected synchronized sensor data from four robots across two large outdoor environments using systematically varied trajectory overlaps, and developed an automated pipeline to fuse LiDAR scans with open geospatial maps for semantic annotation and high-precision ground truth.
Key results
- Released a 16.7 km dataset across two outdoor environments with four synchronized robot trajectories each.
- Introduced systematically varied inter-robot trajectory overlaps and rendezvous points to test different levels of observational redundancy.
- Developed an automated semantic labeling pipeline fusing LiDAR data with open geospatial maps (OSM, DEM, DSM).
- Provided baseline benchmarks for LiDAR place recognition and distributed C-SLAM to demonstrate dataset utility.
Why it matters
It provides researchers and developers with a standardized, realistic benchmark to rigorously evaluate and advance multi-robot collaborative perception and SLAM algorithms.
Abstract
A central challenge for multi-robot systems is fusing independently gathered perception data into a uni- fied 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 re- producible evaluation in multi-robot collaborative perception tasks. The dataset, support code, and updates are available at https://arpg.github.io/cumulti.