Here are some ideas for possible projects. You are welcome to contact us with your own ideas!
Project Ideas
- Mimicking a walking pedestrian: can we program the robot so that its dynamics is similar to the one of a walking pedestrian, when alone and when in dense crowds?
- Multi-agent optimization: developing algorithms for coordinating multiple robots towards a shared goal.
- Embedded programming: low-level software development for robot hardware and real-time control.
- Computer vision: perception systems for robot sensing and environment understanding.
- Machine learning (reinforcement learning) for optimizing robot tasks and behaviors.
Ongoing
RE-FLOW
Funded by EAISI
RE-FLOW develops AI-based decision-making and geometric control methods that let a collective of mobile robots behave as reconfigurable, physical flow-control infrastructure: the robots can assemble into barriers and waypoints, and reconfigure in real-time as the desired flow changes. The project targets a closed-loop flow shaping system, which observes pedestrian (or agent) motion, infers flow descriptors (density, speed, pressure, bottlenecks), plans an updated collective geometry, and executes it safely and quickly with distributed onboard control.
Validation will take place in simulation as well as in the RoboFlow facility using a fleet of ~60 small robots and overhead perception. In the long run, the approach developed in RE-FLOW can be extended to other application areas:
- Mobility: RE-FLOW is the first step towards the use of robotic collectives for traffic and disaster management.
- High-Tech Systems: reconfigurable robotic swarms for distributed sensing and intralogistics.
- Health: foundations for safe, shape-aware micro-robot collectives navigating in blood vessels.
Supervisory Team
Valentina Breschi, Tim Ophelders, Elena Torta (Main PI), Federico Toschi, Bettina Speckmann
Past Projects
L. Liu, J. Werfel, F. Toschi, & L. Mahadevan, Noise-enabled goal attainment in crowded collectives, Proc. Natl. Acad. Sci. U.S.A. 123 (7) e2519032123, https://doi.org/10.1073/pnas.2519032123 (2026).