UAV Path Planning and Control: Towards a Complete Mission Management System

Unmanned Aerial Vehicles (UAVs) mission management requires methodologies that rely on sophisticated path planning and following capabilities. In this paper, we examine the performance of two different control approaches for achieving flyable paths that are represented by B-Splines. A Proximal Policy Optimization (PPO) approach, which belongs to the Reinforcement Learning (RL) methodologies, is presented and evaluated for the control of the roll angle of a fixedwing UAV in the presence of varying wind conditions. The performance of the PPO approach is examined in parallel with a simple PD-like Fuzzy Logic Controller (FLC) of Mamdani type. It turns out that the trained roll controller (RL) is just slightly better than the heuristically (FLC) designed one, in terms of statistical performance, while being less transparent and generic. Potential future research directions are identified for further refinement of the method and expansion of the application scope.