Controller analysis and synthesis for wireless servo systems

Closed loop control over wireless networks has become an increasingly popular research topic recently. While wireless network control loops may be currently infeasible for high bandwidth or critical systems, they are very attractive for a wide variety of applications due to the affordability and ubiquity of standardized wireless communication hardware, and the compatibility of wireless networks with controller and actuator mobility. The convergence of communication in a wireless environment and control has great potential as an enabling technology for many emerging and future applications.; This research develops models for servo control over wireless networks and assembles associated analysis and design tools to address many of the challenges posed by introducing the network. A communication framework utilizing hybrid timer-based sensing/actuation with event-based control is presented that simplifies the scope of the analysis. In this scheme, problems of variable transmission delays and data dropout are analyzed as losses in a discrete-time setting with a fixed step of delay. A discrete-time Markovian jump linear system forms the basic modeling framework. A two-state Markov chain describes the condition of the network and begins to capture natural burstiness of network transmissions. Switched dynamics characterize different modes of operation that depend on the success of a full transmission loop. Two different control approaches are examined in detail. The first approach examines how a state feedback controller designed for a nominal servo system without considering the network can be analyzed when wireless communications are introduced. Strategies are presented, analyzed, and experimentally tested to improve the control performance by applying loss compensation or by using sample time variation to indirectly regulate the network. The second approach develops robust state feedback control synthesis techniques that simultaneously incorporate polytopic plant and network condition uncertainty. This method provides sufficient LMI based conditions to directly synthesize state feedback controllers with guaranteed performance bounds when the true plant and network conditions are unknown, but lie within known convex regions. Controllers designed using this technique are implemented on an experimental wireless inverted pendulum system and are shown to be robust to the designed uncertainty while providing good regulation performance.