Content area
Abstract
We address control systems deployed over communication networks, referred to as a “networked control” systems. This dissertation details work towards developing policies, in the form of theoretical results, and mechanisms, in the form of implemented system architecture, for such networked control systems.
We are mainly concerned with the effect that an unreliable communication channel has on overall system performance. The first policy result we present derives bounds for the optimal achievable performance in a networked control system or sensor network with packet loss. The results address both where to locate controller logic within a network, as well as what logic to use. We present upper and lower bounds on the optimal cost which are close in several example systems. This is significant because this type of problem has previously been considered intractable [1].
We then turn to the concept of linear temporal coding. We show that in response to a packet loss it is in fact better to transmit a linear combination of dropped measurements than to simply send the most recent observation—which is commonly done. This result has implications for the design of systems with limited communication and computation capacity, as well as systems where sampling rate is much higher than transmission bandwidth. It also raises the possibility of creating a transport protocol specifically for networked control. The protocol could modify untransmitted packet contents based on previous message disposition status.
We then consider the effects of delay on system stability. As in the packet drop case, we address both the issues of controller location as well as control law. We obtain a necessary condition on the packet delay probability distribution for system stability. We then present several sub-optimal estimator schemes and obtain sufficient conditions for stability. This allows us to study the impact of delay probability on distribution stability.
We next consider the issue of mechanisms in networked control. We address two topics related to enabling collaborative collision avoidance between vehicles by using wireless communication. The first is a collision avoidance system for the model vehicles in the Information Technology Convergence Laboratory. We address the problem of system architecture. We study advantages and disadvantages of several alternative system architectures, leading to the choice of one which has been deployed.
The second mechanism we present is a “Message Dispatcher” (MD) layer for information dissemination in wireless vehicle-to-vehicle communication. The MD enables efficient and flexible transmission of data elements between vehicles. It has been implemented in several vehicles at Toyota, and has been adopted as basis for a standard by the Society of Automotive Engineers for wireless safety application communication. We present its architecture. We also present a potential extension to illustrate the utility of using linear predictive coding to reduce bandwidth usage.
