The Berkeley Unmanned Aerial Vehicle (UAV) research aims to design, implement, and analyze a group of autonomous intelligent UAVs and UGVs (Unmanned Ground Vehicles). The goal of this dissertation is to provide a comprehensive procedural methodology to design, implement, and test rotorcraft-based unmanned aerial vehicles (RUAVs). We choose the rotorcraft as the base platform for our aerial agents because it offers ideal maneuverability for our target scenarios such as the pursuit-evasion game. Aided by many enabling technologies such as lightweight and powerful computers, high-accuracy navigation sensors and communication devices, it is now possible to construct RUAVs capable of precise navigation and intelligent behavior by the decentralized onboard control system. Building a fully functioning RUAV requires a deep understanding of aeronautics, control theory and computer science as well as a tremendous effort for implementation. These two aspects are often inseparable and therefore equally highlighted throughout this research.; The problem of multiple vehicle coordination is approached through the notion of a hierarchical system. The idea behind the proposed architecture is to build a hierarchical multiple-layer system that gradually decomposes Each RUAV incorporated into this system performs the given tasks and reports the results through the hierarchical communication channel back to the higher-level coordinator.; In our research, we provide a theoretical and practical approach to build a number of RUAVs based on commercially available navigation sensors, computer systems, and radio-controlled helicopters. For the controller design, the dynamic model of the helicopter is first built. The helicopter exhibits a very complicated multi-input multi-output, nonlinear, time-varying and coupled dynamics, which is exposed to severe exogenous disturbances. This poses considerable difficulties for the identification, control and general operation. A high-fidelity helicopter model is established with the lumped-parameter approach. With the lift and torque aerodynamic model of the main and tail rotors, a nonlinear simulation model is first constructed. The control models of the RUAVs used in our research are derived by the application of a time-domain parametric identification method to the flight data of target RUAVs. Two distinct control theories, namely classical control theory and modern linear robust control theory, are applied to the identified model. The proposed controllers are validated in a nonlinear simulation environment and tested in a series of test flights.; With the successful implementation of the low-level vehicle controller, the guidance layer is designed. The waypoint navigator, which decides the adequate flight mode and the associated reference trajectory, serves as an intermediary between the low-level vehicle control layer and the high-level to commands that are compatible with the low-level structure, a novel framework called Vehicle Control Language (VCL) is developed. The key idea of VCL is to provide a mission-independent methodology to describe given flight patterns. The VCL processor and vehicle control layer are integrated into the hierarchical control structure, which is the backbone of our intelligent UAV system. The proposed idea is validated in the simulation environment and then fully tested in a series of flight tests.
展开▼
