Planejamento de movimento agressivo para um sistema de quadricóptero com carga suspensa baseado em RRT

Abstract

The application of Unmanned Aerial Vehicles (UAVs), also called drones, has been growing lately. That is due to advances in technology, which has made them more accessible, versatile, and has also reduced production costs. Drone technology has been used in many commercial applications, including the transport segment. In this segment, drones can bring benefits to society as load vehicles. For this purpose, a cable is used to suspend and transport the load with the drone. However, the cable allows oscillations in the load, which requires the development of motion planning and control architectures capable of handling the oscillation efficiently. Many algorithms can be used to solve such problems, including reinforcement learning or quadratic programming to generate feasible and optimal collision-free trajectories. Although these techniques provide optimal motion planning, the success of finding a solution depends on geometric constraints for simple and convex obstacles. This dissertation proposes a motion planner that generates feasible trajectories and does not constrain the obstacle geometry. To this end, a variant of the algorithm Rapidly-exploring Random Tree (RRT) is proposed, which is a randomized motion planning algorithm. Results show that the proposed algorithm is able to solve the motion planning problem in many situations.