Aerial Robotic Chain: Modeling, Control, Shape and Motion Planning
Loading...
Authors
Nguyen, Dinh Huan
Issue Date
2020
Type
Thesis
Language
Keywords
Model Predictive Control , Path Planning , Reconfigurable Aerial Robots
Alternative Title
Abstract
Research in aerial robotics is pushing the frontier of overall autonomy, sensing, processing and endurance characteristics of Micro Aerial Vehicles (MAVs). Flying robots are now being integrated in an ever increasing set of applications including those of infrastructure inspection, mine exploration, surveillance, entertainment and more. However, despite this unprecedented progress there still exist challenges and mission profiles that are rendered impossible for current aerial robotic technology. Largely, existing MAV designs are monolithic and thus present certain trade-offs, for example between payload and endurance or size. A traditional quadrotor design for example cannot simultaneously have the ability to navigate through narrow corridors and have the capacity to integrate a significant payload or present long endurance. Aiming to overcome these limitations, this work investigates the advanced potential of a reconfigurable and multilinked system-of-systems of aerial robots in order to simultaneously achieve the ability to cross narrow sections, morph shape, ferry significant payloads, offer distributed sensing and computing, and enable redundancy alongside system extendability. The proposed design, dubbed the Aerial Robotic Chain (ARC), corresponds to a reconfigurable system of systems with individual quadrotor MAVs (ARC-units) connected through rigid links and 3-Degree of Freedom (DoF) joints. The proposed design is generic, applicable to N-connected MAVs, while in this specific work we have realized an experimental prototype consisting of two connected quadrotors (ARC-Alpha). In this work we contribute the design, modeling, control design, shape and motion planning for the aerial robotic chain. In particular, a parallel control design consisting of individual SO(3) controller for angular control of each rigid link, alongside Model Predictive Control for the position control of the system-of-systems is proposed. The shape and motion planner for ARC exploits a library of Aerial Robotic Chain configurations, optimized either for cross-section size or sensor coverage, alongside a probabilistic strategy to sample random shape configurations that may be needed to facilitate continued collision-free navigation. Experiment and simulation studies demonstrate performance of the proposed controller and motion planner.