Path Planning for a Fleet of Drones: Exploration and Multi-Object Target Tracking

The thesis revolves around the path planning of a fleet of drones. This planning is specifically geared towards enabling efficient exploration with partial knowledge of regions with static obstacles, with the ultimate goal of multiple target identification. Once a target has been successfully identified, a designated sub-set of the drone fleet is tasked with tracking it, while the remaining drones continue their search for any remaining targets. This research is based on state-of-the-art methodologies for both the exploration process and the tracking of identified targets, all of which take place within realistic simulation environments. A good portion of this work is devoted to the setup and implementation of a reliable platform for the fast deployment of a fleet of UAVs for Coverage Path Planning (CPP) and successive Target Tracking (TT). All the software developed is mainly thought for the PX4/ROS2/Gazebo development stack, but a modular approach has been used in order to allow code portability. The exploration of the Region Of Interest (ROI) is based on a recently proposed planner called DARP (Divide Areas based on Robots' Initial Positions). Its main characteristics are: guarantee of complete coverage, non-backtracking required, minimum coverage path assigned to each UAV, no preparatory stage needed before the optimization, possibility to assign different coverage percentages. Finally, it has the key feature of taking in input ROIs with non-convex shapes and non-convex obstacles. The TT relies on a Connectivity Maintenance approach. The agents assigned to TT, together with the target itself, are modeled as a graph. The position of an agent (or target) corresponds to a node in the graph and the quality of the connectivity between two nodes is modeled with a weighted edge. With this method, obstacle avoidance and inter-agent collision avoidance are taken into account. Finally, drone-agnostic and realistic simulations in Gazebo are carried out to assess the effectiveness of the developed platform.