Abstract

It is generally agreed that communication is a critical technological factor in designing networked unmanned systems (NUS) that consist of a large number of heterogeneous assets/nodes that may be configured in ad-hoc fashion and that incorporate intricate architectures. In order to successfully carry out the NUS missions, communication among assets need to be accomplished efficiently. In contrast with conventional networks, NUSs have specific features that may render communication more complex. The main distinct characteristics of NUS are as follows: (a) heterogeneity of assets in terms of resources, (b) multiple topologies that can be fully-connected, (c) real-time requirements imposed by delivery timeliness of messages under evolving and uncertain environments, (d) unknown and random time-delays that may degrade the closed-loop dynamics performance, (e) bandwidth constraints reflecting differences in assets behavior and dynamics, and (f) protocol limitations for complying with the wireless features of these networks. The NUS system consists of clusters each having three nodes, namely, a sensor, a decision-maker, and an actuator. Inspired by networked control systems (NCS), we introduced a generic framework for NUSs. Using the fluid flow model (FFM), the overall dynamical model of our network cluster is derived as a time-delay dependent system. The following three main issues are investigated in this thesis, bandwidth allocation, an integrated bandwidth allocation and flow rate control, and congestion control. To demonstrate the difficulty of addressing the bandwidth allocation control problem, a standard PID is implemented for our network cluster. It is shown that in presence of feedback loops and time-delays in the network, this controller induces flow oscillations and consequently, in the worst-case scenario, network instability. To address this problem, nonlinear control strategies are proposed instead. These strategies are evaluated subject to presence of unknown delays and measurable/estimated input traffic. For different network configurations, the error dynamics of the entire controlled cluster is derived and sufficient stability conditions are obtained. In addition, our proposed bandwidth allocation control strategy is evaluated when the NUS assets are assumed to be mobile. The bandwidth allocation problem is often studied in an integrated fashion with the flow rate control and the connection admission control (CAC). In fact, due to importance of interaction of various components, design of the entire control system is often more promising than optimization of individual components. In this thesis, several robust integrated bandwidth allocation and flow rate control strategies are proposed. The third issue that is investigated in this thesis is the congestion control for differentiated-services (DiffServ) networks. In our proposed congestion control strategies, the buffer queue length is used as a feedback information to control locally the queue length of each buffer by acting on the bandwidth and simultaneously a feedback signaling notifies the ordinary sources regarding the allowed maximum rate. Using sliding mode generalized variable structure control techniques (SM-GVSC), two congestion control approaches are proposed, namely, the non degenerate and degenerate GVS control approaches. By adopting decentralized end-to-end, semi-decentralized end-to-end, and distributed hop-by-hop control approaches, our proposed congestion control strategies are investigated for a DiffServ loopless mesh network (Internet) and a DiffServ fully-connected NUS. Contrary to the semi-decentralized end-to-end congestion control strategy, in the distributed hop-by-hop congestion control strategy, each output port controller communicates the maximum allowed flow rate only to its immediate upstream node(s) and/or source(s). This approach reduces the required amount of information in the flow control when Compared to other approaches in which the allowed flow rate is sent to all the upstream sources communicating through an output port.