Publications

In this paper, we present a method for area balancing by a team of robotic agents when the inter-agent communication is discrete and opportunistic. The motivation for this problem comes from a setting where a team of robots works in a distributed manner to achieve persistent coverage in an environment where long-range or global communication is not feasible. Our area balancing method switches between a greedy rule and an altruistic heuristic to obtain fast convergence to an equitable partition of the environment. As a stepping stone to experiments in realistic settings, we perform software-in-the-loop tests of the proposed algorithm, including using a novel swarm-aware ground control station, and demonstrate the performance of our algorithm through a Gazebo-based simulation of a multi-drone mission to extinguish forest fires.

The paper studies UAV–UGV tracking in which a single downward-facing UAV camera is the UGV's only exteroceptive sensor, supplying both its lane-following reference and its localization measurement. The shared camera couples navigation and estimation through the UAV altitude z: the measurement-noise covariance grows as R ∝ z², while the camera footprint, and with it lane coverage, shrinks as the UAV descends. The estimation-optimal altitude is the lowest one that keeps the searched lane region inside the footprint, and it rises with path curvature when the searched lane region grows in turns. Adapting altitude along the path beats every fixed altitude only when this coverage floor binds and varies along the path; otherwise a fixed altitude is already optimal. We implement the policy, COAST (Coupled-Altitude Shared-camera Tracking), with a belief-space ground controller and an event-triggered aerial planner. A ROS 2/Gazebo study confirms the z² scaling of the position error, a monotone predicted estimation cost and the predicted tie in the non-binding regime, and event triggering reduces aerial-planner computation at matched estimation quality.