Path Planning for Cellular-connected UAV using Heuristic Algorithm and Reinforcement Learning

With the development of Unmanned Aerial Vehicle (UAV), a novel technology called cellular-connected UAV has been proposed to make UAV complete its mission more efficiently. We consider a scenario where UAV must take off from a random start point, travel over some specific points (e.g. collecting data from sparce sensors in large area) and reach a random end point while keep connected to the Ground Base Station. One of the major challenges is to plan the flying path of UAV while satisfies all constraints. We abstract the path planning problem into Travel Salesman Problem (TSP) and use A* combine with Genetic Algorithm, Simulated Annealing Algorithm and Reinforcement Learning Model to solve TSP to get the best path for cellular-connected UAV. In addition, we did experiments and recorded the results to analyze the advantages and disadvantages of these algorithms.