Evolvable Hardware Group
Evolvable Hardware in Computational Biology
In this work we consider biological systems and consider how an understanding of these systems might help us towards better techniques for developing complex reliable designs. In addition, we see the need for computational help to further biologists understanding of these complex systems. To model these systems in hardware allows us to explore the possibilities of hardware modelling using evolutionary techniques. One of the challenges modelling these systems provides us with, is modelling based on incomplete information. This is also a challenge in the design of complex circuits. Another challenge is to find a good method for representing the components of these systems. Representation is again a challenge in evolvable hardware.[top]
Pauline Haddow, Assoc. Professor, Department of Computer and Information Science, NTNU
Astrid Lægreid, Assoc. Professor, Department of Physiology and Biomedicine Technology, NTNU
Berit Johanson, Assoc Professor, Department of Botany, NTNU
Piet Van Remortel, PhD student,
Gunnar Tufte, PhD student; Department of Computer and Information Science, NTNU
NN (position not filled as yet)
Applying Artificial Evolution to Reconfigurable Hardware to Achieve Hardware Modelling of Signal Transduction
In this project we seek to develop a new direction for evolvable hardware that is hardware modelling in addition to improving the
technique evolvable hardware itself.
One important challenge in such a project is for the two fields to really understand one other. A model is only asgood as it's specification. Can the biologists specify the problem in such a way such that technologists can understand it? On the other hand, can technologists present their model in such a way as to ensure that the biologists can understand the model and help us to achieve realism in the model?
To model a biological process we need to find a method to represent the different components of the system and then our intention is to use evolutionary techniques i.e. genetic algorithms or genetic programs to evolve or even grow the systems. In this work it is hoped to model biological processes in a way so as to allow for a better understanding of the modelling technique by the biologists themselves by providing a method to physically represent components of complex processes and their interactions in hardware. A physical representation is aimed at giving a close one-to-one mapping between the actual biological processes and the abstract technological representation. As such, it is hoped that abstraction is at a level where both the model and it's development may be closer followed and understood by biologists to ensure a closer understanding between the two fields.
Through this modelling process we hope to learn a lot more about representation. To enable the design technique evolvable hardware to be able to design complex systems we need to turn to biology and learn from nature better methods of representation and develop the evolution technique itself.
A further challenge is the lack of information available from biologists. That is, we will be modelling based on incomplete information attempting to come to hypothesis that can be further tested out using biological experimentation. As such this provides
both an exciting modelling challenge and a very useful study for design purposes. In designing complex designs we know that today's representations will require unrealistically large individuals which means that evolution of complex designs will not be viable. Learning to evolve based on incomplete information will allow us to reduce the amount of information to be stored within the individuals of the evolution process and hopefully still manage to evolve a reliable complex design.
Shrinking the Genotype: L-systems
by Pauline C Haddow, Gunnar Tufte and Piet van Remortel, submitted to ICES'2001