There are reports that reconfigurable FPGA devices have been made to evolve to perform a useful function by a process to random mutation of the configuration and testing by means of a goodness of fit parameter.
These evolved circuits do not function in simulations in Pspice because they rely on the analogue nature of the digital elements, and the variable gate delays in continuous time (which in ANY digital system is never discrete) for their functionality. Neither do they work when transported (with the same circuit topology) to other forms of the hardware such as discrete CMOS.
We wish to study whether there are any impediments to using such evolved circuits in a real engineering application. Is their behaviour robust? are there unforeseen traps in the evolved systems? It is known that they can be very much more economical in circuit complexity than an engineered version of the same function. It is also known that it can be difficult or impossible to analyse how they work, or even understand why they work.
This experimental electronics has application to the study of neural and cognitive processes in animal brains.
A further project which we would like to follow up involves the coupling together of numbers of low-dimensional chaotic systems to see how the behaviour changes as the complexity is increased.
In the first instance it should be possible to build the individual chaotic elements from discrete components and integrated circuits. The advent of VLSI analogue technology provides a means for making larger arrays, and of controlling the tolerance spreads between the elements.
Much research into chaotic processes has been mathematically based and it is not clear how imperfect real engineering implementations will behave, in view of the sensitivity of chaotic systems to certain types of small error.
There are applications to analogue chaotic neural nets, possible of the form of cellular neural nets (CNNs) which have proven advantages for highly parallel signal and image processing in real time.
There are interesting applications to the analogue mimicry of couple Josephson Junction arrays; these are of interest as they have been proposed as the basis of technology for superfast small scale superconducting computers.
We have demonstrated unforseeable trapping in simple analogue chaotic systems, and in impacting systems. We wish to extend this study of trapping to complex computer systems and complex systems in organisations, as part of a safety-critical system analysis.
There is also interest in modelling electronically, by analogue means, the chaotic motions of coupled impacting systems such as are found in a Newton's Cage arrangement where the individual spheres are not touching in the resting state. This has application to the vibrational failures of fuel rod elements in nuclear reactors, and in other complex chaotic vibrational impacting systems.
Safety-critical computer engineering is being compromised by problems due to arbitration in time, and due to chaotic communications between nodes of a network. The result is that such a system, we believe, is unlikely to behave the same on successive occasions, even though no parameters have changed and the data inputs are also unchanged.
There are various ways in which students and research associates may contribute to this kind of work. The activities will change depending on your particular interests and strengths. People from a strong maths background, with allied programming and electronic constructional skills, are ideal. Two of these three attributes are a minimum requirement. Some prior interest in chaotic systems would help.
A good first degree in Maths, Physics, Electronic Engineering, or Computer Science is needed. A conversion course at MSc level from the first degree speciality into one of these other areas would also help.