SPACE-TIME GRAPH PATH PLANNING FOR UAS TRAFFIC MANAGEMENT SYSTEMS

The unmanned aerial vehicle (UAV) technology has evolved considerably in recent years and the global demand for package delivery is expected to grow even more during COVID-19 and the social distance era. The low cost of acquisition, payload capacity, maneuverability, and the ability to y at low-altitude with a very low cost of operation, make UAVs a perfect fit to revolutionize the payload transportation of small items. The large-scale adoption of drone package delivery in high-density urban areas can be challenging and the Unmanned Aircraft Systems (UAS) operators must ensure safety, security, efficiency and equity of the airspace system. In order to address some of these challenges, FAA and NASA have developed a new architecture that will support a set of services to enable cooperative management of low-altitude operations between UAS operators. The architecture is still in its conceptual stage and designing a mechanism that ensures the fair distribution of the available airspace to commercial applications has become increasingly important. Considering that, the path planning is one of the most important problems to be explored. The objective is not only to find an optimal and shortest path but also to provide a collision-free environment to the UAVs. Taking into consideration all these important aspects and others such as serving on-demand requests, flight duration limitation due to energy constraints, maintaining the safety distance to avoid collisions, and using warehouses as starting and ending points in parcel delivery, this dissertation proposes: (i) an energy-constrained scheduling mechanism using a multi-source A* algorithm variant, and (ii) a generalized path planning mechanism using a space-time graph with multi-source multi-destination BFS generalization to ensure pre-flight UAV collision-free trajectories. This dissertation also uses the generalized path planning mechanism to solve the energy-constrained drone delivery problem. The experimental results show that the proposed algorithms are computationally efficient and scalable with the number of requests and graph size.