Evolutionary Algorithms (EAs) have long been used to provide innovative solutions to complex problems. GAs model their behaviour on biological processes in evolution. This is achieved using a simplification (and sometimes stylisation) of known biological operations incorporated as operators into the algorithm .Two of the most common of these are Recombination and mutation.
Evolvable hardware (EHW) utilises genetic algorithms to create circuit solutions given only limited information, such as the circuits desired functionality. Evolution then determines the circuits structure given the building blocks presented to it (these normally include primitive blocks such as AND gates for digital circuit design). One significant problem currently taxing research in digital EHW is the limited success with which complex digital circuits can be implemented using evolutionary techniques.
Much of the research the group is currently undertaking is aimed at tackling this problem using a functional level approach to evolvable hardware. As a result larger building blocks (such as Full-adders or multipliers), are presented to the system in an attempt to diversify the size, number and complexity of the building block available during the evolution of a circuit. Thus enabling the iterative development of larger and more complex systems in a shorter time period. Much effort is also being spent to develop efficient testing techniques which will enable reliable circuit verification during evolution and provide a means by which the fittest individual (best circuit solution) can be tested with maximum reliability confidence.
The study of new hardware architectures that adapt to environmental changes through use of algorithms that are inspired by the process of natural. Research in the group involves both the design of new algorithms and hardware architectures.
Example projects could include:
- The design of new algorithms and architectures that have inherent fault tolerance. Faults here could be caused by environmental changes such temperature and/or radiation.
- The design of future architectures and algorithms that improve performance in terms of speed and power through exploitation of multidimentionality and cellular structure.
