Parallel Processing in Digital Control is a volume to be published in the new Advances in Industrial Control series, edited by Professor M.J. Grimble and Dr. M.A. Johnson of the Industrial Control Unit, University of Strathclyde. The growing complexity of digital control systems in such areas as robotics, flight control and engine control has created a demand for faster and more reliable systems. This book examines how parallel processing can satisfy these requirements. Following a survey of parallel computer architectures, MIMD (Multiple Instruction Multiple Data) machines are identified as suitable systems for digital control problems, which are characterised by a mixture of regular and irregular algorithmic tasks. An example of a typical MIMD architecture, suitable for real-time control, (the Inmos Transputer) is introduced together with its associated parallel programming language (Occam). The key problem in implementing parallel software is associated with mapping parallel tasks onto physical processors. In this book a variety of schemes are described and assessed to help illustrate potential areas of difficulty for the real-time control software engineer. Solutions are proposed and tested on a flight control case study example. Recognising the widespread acceptance of MATLAB and its derivatives for computer aided control system design, this book demonstrates how mapping strategies can be realised in this environment and integrated with a transputer development system for on-line performance evaluation. A case study example demonstrates the power of this approach and important issues are highlighted. Readers will experience the advantages of parallel processing in digital control while being made aware of the key factors to be considered in the development of an effective solution. Practising control engineers and graduate/post-graduate students will find the book of particular interest and benefit.

Inhalt

1 Introduction.- 1.1 The Need for Parallel Processing.- 1.2 Outline of the Book.- 2 Parallel Processing and the Transputer.- 2.1 Introduction.- 2.2 Parallel Processing Computer Architectures.- 2.3 Why use Occam and Transputer ?.- 2.4 The Transputer.- 2.5 Parallel Programming and Occam Language.- 2.6 Real-Time Control Transputer Applications.- 2.7 Summary.- 3 Parallel Algorithm Strategies.- 3.1 Introduction.- 3.2 Partitioning and Task Allocation.- 3.3 Vap Control Law.- 3.4 Mapping Strategies.- 3.5 Summary.- 4 Epicas - An Environment for Parallel Implementation of Control Algorithms and Simulation.- 4.1 Introduction.- 4.2 Epicas.- 4.3 Matlab Tools for Parallel Partitioning.- 4.4 Occam Tools for Tasks Allocation.- 4.5 Using Epicas for Simulation.- 4.6 Summary.- 5 Performance Issues: Granularity, Topology, Mapping Strategies.- 5.1 Introduction.- 5.2 Measuring Parallel Processor Performance.- 5.3 Processor Farm Performance.- 5.4 Processor Star Performance.- 5.5 Granularity Issues.- 5.6 Summary.- 6 Concluding Remarks.- 6.1 General Review.- 6.2 Future Work.- 6.3 Summary.- References.