Enhancing real-time embedded systems development using artificial immune systems

The Consumer Electronics industry produces a large number of systems which exhibit both real-time and embedded properties (RTES). Frequently problems are encountered during the development of these systems due to a mismatch between the requirements of traditional real-time development techniques and the restrictions imposed on the development process by the retail market. There is therefore a requirement for a method which can be used to enhance the development process and improve the reliability of these systems, yet which does not require the intensive formal analysis often required by established real-time development techniques.

Observing that organisms in the natural world often exhibit characteristics which would be desirable in computer systems, such as the ability to adapt to changing environments and protect themselves against attack from external agents, this thesis examines the potential for techniques inspired by biological observations to be employed as a means to detect common problems encountered in the development of RTES.

The Dendritic Cell Algorithm (DCA), an algorithm derived from the observed operation of the innate immune system which has previously been applied to a number of anomaly-detection problems, is included in a simulated RTES to detect deadline overruns in a task scheduling system, and its effectiveness investigated over a variety of sample systems of varying sizes and complexities.

In conjunction with this, the issues surrounding the incorporation of an immune-inspired technique into embedded systems are examined in detail, with focus on the optimisation of the DCA for maximum effectiveness, whilst still providing flexibility. Attention is paid towards ensuring that the DCA is able to function satisfactorily in a constrained environment, by establishing the level of system resources required and proposing further enhancements to the algorithm to enhance its suitability for constrained-resource implementation.

The work makes contributions to the real-time community by proposing an original method by which software development for real-time embedded systems can be enhanced, to enable the production of reliable software without the need for the formal analysis procedures normally needed.

There are also significant contributions made to the AIS community, by way of the introduction of an application domain in which the application of AIS techniques has not previously been proposed, and the enhancements made to the DCA to make it better suited to this domain than previously.