Control of Micro Aerial Vehicles under Aerodynamic and Physical Contact Interactions

点击翻译按钮获取中文摘要

The goal of this project is to make quadrotors and other similar small-scale flying rotorcraft safer and easier to fly. Both recreational and commercial use of these vehicles has recently surged in popularity. However, safety concerns about potentially damaging collisions limit their deployment near people or in close formation, and the current state of the art in vehicle control is insufficient for potential applications involving flight inside of complicated structures such as industrial plants, forests and caves. Specifically, this project will lead to the following innovations: better understanding of the aerodynamic interactions between the environment and the flying vehicle through novel models; creation of a "smart cage" that encloses the vehicle and acts as both a shock-absorber and as a novel type of "touch" sensor; and the derivation of new control strategies to take advantage of the new features to improve performances and ease of maneuvering. Together, these innovations will make small-scale vehicles less likely to cause unintended damage, suitable for use in extreme environments such as caves, and more easily piloted. This will allow in turn the use of these vehicles in new industrial monitoring and search-and-rescue applications, thus bringing the benefits of these platforms to larger segments of society. The technical goals of this project will build on new models of quadrotor aerodynamic, constructed through a combination of simulation and experimental evaluation. These models predict forces and torques due to the aerodynamic interactions of each rotor with the other rotors, as well as with nearby surfaces, such as ground, walls, and ceilings. A "smart cage" surrounds the vehicle, composed of a rigid outer shell to prevent the quadrotor from contacting objects in the environment, and a base, which is rigidly attached to the quadrotor. The cage and the base connect together through a system of springs that are used to absorb impacts. The base includes sensors to measure the relative displacement of the outer shell, thus allowing a partial reconstruction of any impact dynamics. New control strategies are formulated, based on a novel application of contraction theory to Riemannian manifolds. These strategies produce geometric controllers that do not suffer from singularities, that have global exponential convergence guarantees, and that can be automatically tuned to obtain optimal performance. The results will be validated in both simulations and experiments.