CoopUQ forecasts occupancy uncertainty for multi-robot navigation
Original title: CoopUQ: Robust Multi-Robot Social Navigation via Uncertainty-Aware Cooperative Forecasting
Unity3D and real-robot tests report safer navigation than baselines under occlusion and out-of-distribution crowds.
CoopUQ is a multi-robot social navigation framework that combines uncertainty-aware cooperative occupancy forecasting with constrained multi-agent reinforcement learning. Robots fuse spatiotemporal features to predict future occupancy grids, and the policy cost penalizes travel through regions predicted as occupied or highly uncertain. Experiments in multiple Unity3D environments cover in-distribution and out-of-distribution layouts, pedestrian densities, and rushing behavior. The authors report better prediction and safety than representative baselines, with additional real-robot tests; code and videos are at coopuq.github.io.
Unconfirmed: Safety and prediction gains are author-reported against unnamed representative baselines.
- Forecasts come from cross-robot spatiotemporal feature fusion.
- The cost penalizes occupied or highly uncertain predicted regions.
- Tests include diverse layouts, pedestrian densities, and rushing behavior.
- Real-robot experiments are reported in addition to Unity3D simulation.
- Products
- Unity3D
From the original article
1 min read at arxiv.orgReinforcement learning enables mobile robots to make fast and scalable navigation decisions in constrained human-shared environments. However, directly extending single-robot approaches to multi-robot settings often fails to ensure safety and efficiency, especially under occlusions and out-of-distribution (OOD) scenarios, where individual robots may overlook potential risks or make inaccurate predictions.
By Zehao Wang · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.