Learning-Based Geometric Leader-Follower Control for Cooperative Rigid-Payload Transport with Aerial Manipulators

Researchers Omayra Yago Nieto and Leonardo Colombo have published a paper introducing a novel learning-based geometric control framework that enables multiple aerial manipulators (drones with robotic arms) to cooperatively transport rigid payloads. The system adopts a leader-follower architecture where one designated drone generates the desired payload wrench based on geometric tracking errors, while the remaining follower drones realize this wrench through constraint-consistent force allocation. This approach explicitly models contact forces, payload dynamics, and internal force redundancy within a unified geometric model, creating a coupled agent-payload system that handles the complex physics of multi-point lifting.

The technical core uses Gaussian Process (GP) regression to compensate for unknown disturbances and modeling uncertainties, with high-probability bounds on learning error explicitly incorporated into the control design. This combines GP feedforward compensation with geometric feedback control. Through Lyapunov analysis, the researchers established uniform ultimate boundedness of payload tracking errors with high probability, where the ultimate error bound scales with the GP predictive uncertainty. This represents a significant advance in cooperative aerial manipulation, moving beyond single-drone systems to enable teams of drones to perform precise transportation tasks that were previously impossible due to coordination challenges and environmental uncertainties.