Simulation of distributed control systems

Abstract

A wireless sensor network (WSN) consists of a set of nodes powered by batteries and collaborating to perform sensing tasks in a given environment. It may contain one or more sink nodes (base stations) to collect sensed data and relay it to a central processing and storage system. A sensor node can be divided into three main functional units: a sensing unit, a communication unit and a computing unit. A distributed control system consists of a set of nodes having sensing, control and actuation capabilities and interacting using an overlapping network (Branicky et al. 2003). In such a system, any of the three main control loop tasks of sensing, control and actuation can be performed in a distributed manner. A wireless implementation of these networks is discussed in this chapter. The use of wireless networks for distributed sensing and control offers several benefits. It allows cost reduction and eliminates the need for wiring. Wiring could become costly and difficult in the case of a large number of control nodes needed for the sensing and control of a large process. However, the use of wireless networks in a control system introduces new challenges. In fact, these networks can suffer from several problems such as unbounded delays and packet losses. This chapter describes the different architectures that allow the implementation of distributed control systems using wireless sensor networks. These architectures differ according to whether the sensing and/or actuation are performed centrally or in a distributed fashion. In order to analyse the architectures, the Georgia Tech Sensor Network Simulator (GTSNetS) (Ould-Ahmed-Vall et al. 2005c) is extended to simulate distributed control using sensor networks. One of the main features of this simulator is its scalability. In fact, GTSNetS was shown to simulate networks of up to few hundred thousand nodes. The simulator is implemented in a modular way and the user is allowed to choose from different architectural implementations. If a specific approach or algorithm is not available, the user can easily implement it by extending the simulator. This simulator is demonstrated using an existing Bayesian fault tolerance algorithm. Fault-tolerance of individual sensing nodes is also considered in this chapter. The Bayesian fault tolerance algorithm is presented and extended to adapt to dynamic failure rates. The enhanced algorithm allows nodes to learn dynamically about the operational conditions of their neighbours. Each node then gives different weight factors (confidence levels) to the information received from each of its neighbours. The weight factors are function of the failure probability of the specific neighbour. The probability of failure is computed from the reliability of the information received from the specific neighbour compared to other neighbours. The simulator facilitates the implementation and evaluation of the algorithm even for a network containing a large number of nodes. The remainder of the chapter is as follows. Section 2 presents distributed control systems and their applications. Section 3 describes the different architectures that can be used to implement distributed control systems with sensor networks. Section 4 discusses the simulation of distributed control systems using sensor networks. Section 5 presents and compares the original and the enhanced fault tolerance algorithms. Section 6 gives an example simulation scenario. Section 7 concludes the chapter. © Springer-Verlag Berlin Heidelberg 2007.