Motion planning with dynamics and environment awareness for aerial robotic manipulation in inspection and maintenance

Abstract

Unmanned Aerial Vehicles (UAVs) endowed with robotic manipulation capabilities, also known as Aerial Robotic Manipulators (ARMs), have demonstrated a promising future in their application for Inspection and Maintenance (I&M) activities. However, their associated capabilities still need to be extended to reach higher levels of autonomy, reliability, accuracy, safety and e ciency, among others. In addition to important improvements in human safety, this will lead to signi cant cost savings, making ARMs an e ective solution to be exploited in real conditions. Motivated by the previous statement, this thesis has found in motion planning a means to endow ARMs with enhanced functionalities. Thus, the presented research has been focused on the design, development and validation of motion planning methods for aerial robotic manipulation in I&M. More in detail, the need of planning has been identi ed for three main topics, which are manipulation using ARMs endowed with robotic arms, manipulation with hybrid-locomotion robots and multiARM manipulation. For each of them, a motion planning method has been formulated and then, several extensions have been introduced to increase its capabilities. Concerning manipulation using ARMs endowed with robotic arms, a motion planner specially oriented to this kind of ARMs has been formulated for both navigation and manipulation phases in cluttered environments. This planner considers the joint operation of the aerial platform and the manipulation system within the planning process. Over the fundamentals of the previous method, three extensions have been proposed. Due to the complex dynamics existing in ARMs, the rst extension introduces Dynamics Awareness (DA) in the planner operation for robust obstacle avoidance. Complementing this DA extension, a new Velocity Adaptation (VA) mechanism allows a better optimisation of the execution time of the planned trajectories but without increasing the computational burden considerably. Alternatively, accounting for the ARM aerodynamics, the DA approach has also been extended with Aerodynamics Awareness (ADA) to face aerodynamic phenomena that may address robots to collisions. Switching to manipulation with hybrid-locomotion robots, the motion planner is devoted to the generation of e cient plans along sequences of manipulation points, involving this kind of ARMs in cluttered environments. For that, the method takes advantage of the ying and rolling capabilities o ered by these robots. Finally, with the focus on multiARM manipulation, a motion planner for multiARM systems subject to limited payload capacities and dynamic constraints has been presented. This method gives response to missions that require visiting e ciently a set of target regions where loads on board the robots are deployed. Numerical results, realistic simulations and real-world indoor and outdoor ight experiments have demonstrated the bene ts of the planning methods to compute trajectories that lead a wide variety of ARMs to ful l real I&M operations in di erent scenarios. These application scenarios range from the installation of sensors in industrial environments to the inspection of pipe arrays in oil and gas re neries or chemical plants and even the installation of bird ight diverters on electric power lines.