Abstract

Recent years have witnessed a strong interest and intensive research activities in the area of networks of autonomous unmanned vehicles such as spacecraft formation flight, unmanned aerial vehicles, autonomous underwater vehicles, automated highway systems and multiple mobile robots. The envisaged networked architecture can provide surpassing performance capabilities and enhanced reliability; however, it requires extending the traditional theories of control, estimation and Fault Detection and Isolation (FDI). One of the many challenges for these systems is development of autonomous cooperative control which can maintain the group behavior and mission performance in the presence of undesirable events such as failures in the vehicles. In order to achieve this goal, the team should have the capability to detect and isolate vehicles faults and reconfigure the cooperative control algorithms to compensate for them. This dissertation deals with the design and development of fault detection and isolation algorithms for a network of unmanned vehicles. Addressing this problem is the main step towards the design of autonomous fault tolerant cooperative control of network of unmanned systems. We first formulate the FDI problem by considering ideal communication channels among the vehicles and solve this problem corresponding to three different architectures, namely centralized, decentralized, and semi-decentralized. The necessary and sufficient solvability conditions for each architecture are also derived based on geometric FDI approach. The effects of large environmental disturbances are subsequently taken into account in the design of FDI algorithms and robust hybrid FDI schemes for both linear and nonlinear systems are developed. Our proposed robust FDI algorithms are applied to a network of unmanned vehicles as well as Almost-Lighter-Than-Air-Vehicle (ALTAV). The effects of communication channels on fault detection and isolation performance are then investigated. A packet erasure channel model is considered for incorporating stochastic packet dropout of communication channels. Combining vehicle dynamics and communication links yields a discrete-time Markovian Jump System (MJS) mathematical model representation. This motivates development of a geometric FDI framework for both discrete-time and continuous-time Markovian jump systems. Our proposed FDI algorithm is then applied to a formation flight of satellites and a Vertical Take-Off and Landing (VTOL) helicopter problem. Finally, we investigate the problem of fault detection and isolation for time-delay systems as well as linear impulsive systems. The main motivation behind considering these two problems is that our developed geometric framework for Markovian jump systems can readily be applied to other class of systems. Broad classes of time-delay systems, namely, retarded, neutral, distributed and stochastic time-delay systems are investigated in this dissertation and a robust FDI algorithm is developed for each class of these systems. Moreover, it is shown that our proposed FDI algorithms for retarded and stochastic time-delay systems can potentially be applied in an integrated design of FDI/controller for a network of unmanned vehicles. Necessary and sufficient conditions for solvability of the fundamental problem of residual generation for linear impulsive systems are derived to conclude this dissertation.