Sequential and parallel processing techniques for real-time adaptive active control are considered in this chapter. Three different algorithms, namely simulation, control and identification, are involved in the adaptive control algorithm. These are implemented on a number of uniprocessor and multiprocessor computing platforms. The interprocessor communication speed and the impact of compiler efficiency on processor performance are investigated. A comparative assessment is provided, on the basis of the real-time communications performance, computation performance and compiler performance, to lead to merits of design of parallel systems incorporating fast processing techniques for real-time active control applications.

15.1 Introduction

The performance demands required in modern control systems have led to complex processes, such as adaptive control, optimisation, failure tolerance, rule-based systems and neural networks, becoming essential control engineering tools. These processes are almost invariably implemented by digital computing systems, and for online control the implementation is required to be in real time. This means that it is required to compute the control algorithm within the loop sample time. Despite the vastly increased computing power which is now available there can still be limitations in the computing capability of digital processors in real-time control applications for two reasons:

1. sample times have become shorter as greater performance demands are imposed on the system

2. algorithms are becoming more complex as the development of control theory leads to an understanding of methods for optimising system performance.

To satisfy these high performance demands, microprocessor technology has developed at a rapid...