Decision And Control In Distributed Cooperative Systems

This dissertation presents novel matrix-based approaches for decision and control in distributed cooperative systems such as wireless sensor networks. A novel matrix-based Discrete Event Controller has been implemented for task planning and resource dispatching in a network consisting of stationary ground sensors and mobile agents also known in the literature as mobile wireless sensor networks. The use of shared resources in such systems can sometimes cause a phenomenon called system deadlock, where all the processes in the system come to a standstill. This work presents a new matrix-based algorithm that has been implemented for deadlock avoidance in a system with shared resources and dynamic resource assignment. The analysis of deadlock is based on certain Petri Net objects in such systems called critical siphons and critical subsystems. The analysis of deadlock avoidance becomes even more difficult when routing of tasks and resources are involved. The critical siphons and critical subsystems have to be redefined. This dissertation presents a new matrix-based approach for deadlock avoidance in such systems. This is a generalized approach that can be used for systems with or without routing. This work also presents a method for tackling a certain pathological case called second order deadlocks. In routing systems, dynamic decisions have to be made. In the presence of numerous agents which act as resources, a collective decision can be made based on the individual decisions of agents. This method is called data fusion. Dempster Shafer (DS) theory has been extensively used in the past for data fusion since it provides an excellent framework for conditions involving uncertainty. But the mathematics of computation involving DS belief functions is difficult to fathom because of the many summations over set inclusions and intersections. The equations are often difficult to comprehend and discourage readers due to their complexity, and are often difficult to implement using software. This dissertation provides a new matrix formulation for updating evidence and computing beliefs and plausibilities in DS theory. The work also shows how evidence theory can be used in routing systems for Condition Based Maintenance. Finally, this work presents a framework for trust propagation and maintenance in a network of nodes or mobile agents that yields global consensus of trust under rich enough communication structure graphs. This work considers the case where the graph structure is a time-varying function of the trusts based on the graph connectivity. This makes the trust consensus scheme bilinear. This trust consensus is incorporated into cooperative control laws that depend on local information from neighboring nodes, yet yield team-wide desired behavior such as flocking and formations.