Ergodic coverage for drone teams
Drone teams that keep exploring and monitoring a field that changes over time, with no central coordinator. Flown on three drones.
Read more →MattiaMantovani
Robotics researcher. I build autonomous systems that work outside the lab.
About.
I’m a robotics PhD at UNIMORE, working on guidance, navigation and coordination for multi-robot teams, especially drones, and flying them outside the lab. Most of my PhD was about coverage and exploration under uncertainty: robot teams that move through an unknown environment, gather information and model it with Gaussian Processes. Ergodic and Voronoi coverage tell them where to look, and sampling-based MPC (MPPI) turns that into trajectories they can actually fly. The best moment is seeing it all fly on real drone teams.
Flying in real-world conditions also pulled me into state estimation, and I now supervise master’s theses on navigation and localization in GNSS-denied environments.
Lately I’ve been bringing learning into the mix, from diffusion-based planning to exploring vision-language-action models (VLAs) for multi-robot teams. I’m open to industry roles from January 2027, in Zurich, Munich, Paris or London.
- Visiting PhD at Harvard’s REACT Lab (2025): GP-based trajectory estimation for 3D reconstruction of a moving object, published at RSS 2026.
- Papers in T-RO, RA-L, RSS and ICRA, from convergence proofs to real flight.
01 Selected work
Mullet: ground control station for drone swarms
Plan, launch and monitor multi-drone missions from a 3D map. Used in field operations at Skymule.
Read more →Learning while covering: distributed GP coverage
Robot teams that map an unknown, changing field while deploying optimally over it, with no central computer. Up to 74% less compute per robot.
Read more →Seeing moving objects: next-best-view planning
A camera robot that predicts where a moving object will be and picks views that reconstruct it completely. Published at RSS 2026, with Harvard.
Read more →02 Selected publications
- RSS2026Motion-Uncertainty-Aware Next-Best-View Planning for Moving Object Reconstruction
- RA-L2026Distributed multi-robot ergodic coverage control for estimating time-varying spatial processes
- T-RO2026Online Constrained Ergodic Coverage with Constant-Memory Potential Feedback
- T-RO2025Distributed coverage control for time-varying spatial processes
- ICRA2025Online multi-robot federated learning for distributed coverage control of unknown spatial processes
- RA-L2024Distributed coverage control for spatial processes estimation with noisy observations