MOBIUS: MOdelling BIology Using Smart materials.

EOARD contract SPC 024014

OVERVIEW AND AIMS

This project uses methods from computer and materials science to model rarely investigated biological processes. The research employs both computational models and "smart" materials. It is anticipated that the investigation will lead to significant advances in self-designing, self-assembling, self-repairing devices. In more detail, the aims of the research are:

1. To investigate the use of computer models of evolution and development to design and exploit intrinsic properties of novel materials.

2. To investigate the engineering potential of computational biology and novel materials by evolving self-designing, self-assembling devices.

The research will be highly novel and groundbreaking in several respects. It is significant interdisciplinary research, enabling the very latest understandings of nature to be exploited in biologically-inspired technology. It will investigate novel materials capable of sensing, computation and actuation when configured by bio-inspired computational processes. Finally, it will investigate new combinations of recent technologies (e.g., shape memory alloys and stereolithography rapid prototyping), that together should lead to the first ever self-designing, self-assembling, intelligent devices.

BACKGROUND AND MOTIVATION

In recent years, three seemingly unrelated fields of research have begun to change their focus. Traditionally, all explored separate areas, largely unaware of the existence of the others. Today, all have begun to converge in some aspect. The three fields are: Biol-ogy (especially neuroscience, immunobiology and developmental biology), Computer Science (especially neural and evolutionary computation), and Materials Science (particularly the study of "smart materials").

The convergence is coincidental, but highly significant. Take biology and computer science: because of the difficulties in studying and explaining processes of biology, computational models are becoming widely used. Examples include models of im-mune cell responses, neural structure and behaviour, e.g. [Grzeszczuk et al, 1998]. Likewise, because of the difficulty in solving some computational problems, solutions that employ biologically-inspired processes are widely used in computer science [Bentley et al 2001]. Examples include genetic algorithms (based on natural evolution, they find good solutions for difficult problems), and neural networks (based on the workings of neurons, they allow computers to learn, process visual information, and plan) [Bentley 2001]. The convergence is also clear for biology and smart materials: because of the difficulty in visualising some biological or medical computer reconstructions, smart materials are used. For example, 3D printers are used to "print" the form of skulls recon-structed or remodelled by computers, or the form of tumours so that surgeons can see and touch the shape before operating [Evseev and Panchenko 2000]. And most recently, computer science and smart materials are merging. Because of the difficulty in using computers to exploit biological analogies effectively, smart materials are being used. For example, "3D printers" are used to create physical prototypes of evolved products [Bentley and Corne 2001] and materials such as radiation-damaged silicon may permit the computer-evolution and development of neural-like circuitry within them.

Examples of interdisciplinary centres combining two of these three fields are easy to find: the Gatsby Institute, UCL, combines neuroscience with computer science to produce computational neuroscience. The School of Cognitive and Computing Sciences (COGS), University of Sussex, combines biologists of many different disciplines with computer scientists. These are highly suc-cessful interdisciplinary centres: both biologists and computer scientists gain significant and useful knowledge in their respective fields. But to date there is little research combining all three fields.

Their combination is not just convenient, it is essential. As data on natural systems improves, the analyses increasingly rely on better, more realistic modelling and visualisation techniques. As our ambitions to utilise novel materials and develop self-adaptive, self-assembling devices increase, so the need to employ developmental and evolutionary computational processes becomes vital. Nature has solved these problems; her tools are evolution and development. It is now clear that similar processes are necessary to achieve the goal of self-assembling, adapting, moving, thinking devices that will design themselves to be suitable for different environments.

We are seeing the first signs of results from such combinations of fields. For example, an interesting and surprising finding that made a big impact in computer science was in the field of evolvable hardware - a merging of evolutionary processes, computing and electronics [Miller, Job, and Vassilev 2000a,b]. It was shown that by evolving the configuration of logic gates on a Field Programmable Gate Array (FPGA) one could obtain electronic circuits that fulfil the user specification but exploited the underlying physical properties of the silicon substrate [Thompson96]. To this day these circuits are not completely understood. Although the FPGA was designed for the synthesis of digital circuits, the evolved circuits were analogue in nature. Other interesting and important aspects of these circuits is that they were parallel and asynchronous. Another outstanding example of intrinsic (or in-hardware) evolution has been the design of antennae [Bentley and Corne 2001]. In this work wire segment antennae were evolved using reed-switches that outperformed state-of-the-art conventional methods. As yet our mathematics cannot explain these antenna designs.

These findings suggest that when evolution in a computer is allowed to configure systems, the designs produced somehow exploit the physical properties of the apparatus in highly novel ways. Digital evolution, as with natural evolution, can construct amazingly sophisticated systems primarily by exploiting the physical (and chemical) properties of matter.

Other work is no less exciting, with 3D physical robot forms that are evolved and "printed" [Bentley and Corne 2001], or the advent of memory shape alloys to produce "muscle wire", which contracts like a natural muscle when current is passed through it. When combined with a deeper understanding of natural evolution and developmental processes, these technologies look set to transform our capabilities. It will be these techniques that enable the first adaptive, self-designed, self-assembling devices to exist. The gains of such interdisciplinary work is clearly enormous.

METHODOLOGY

The research will focus on two goals:

(1) The creation of a new biologically-inspired algorithm that exploits both evolution and processes of development to enable computers to automatically find solutions to large-scale, complex problems, overcoming current scalability problems in evolutionary computation.

(2) The demonstration of this algorithm on a novel combination of "smart materials" (such as memory shape alloys and rapid prototyping technologies) for the design of self-adapting forms.

In more detail: Current research with leading developmental biologists (Prof. Lewis Wolpert, Prof. Anne Warner) and computer scientists (Dr Bentley and his PhD student Sanjeev Kumar) at UCL has created one of the first detailed computer models of natural developmental processes. Learning from this, and with the continued support of the same (and numerous other) biologists at UCL, this research will examine fundamental issues concerning development. These include: how do developmental processes enable natural evolution to find such remarkably complex solutions to large-scale problems, how does development enable these solutions to repair themselves should they be damaged during development, how does development enable the creation of solutions that are able to change and adapt themselves to different environments. The results of this investigation (in addition to publications) should be an extension to a standard evolutionary algorithm, which will enable computers to evolve genotypes that are then used to define the development of solutions (and hence not be a direct mapping to the solutions). Such an algorithm will be a specialised form of an evolutionary algorithm, extending its capabilities of optimization and exploration, and enabling computers to evolve solutions of greater complexity than is currently possible, and also potentially solutions capable of self-adaptation and repair.

The second goal involves collaboration between UCL's computer science, evolutionary biology departments and external collaborators to investigate the use of computers to evolve novel physical (robot) forms capable of moving (and potentially sensing objects) within their environments. This work will employ memory shape alloys and 3D printer technology available through Kings’ Engineering Dept and BAE Systems. The research will test the capabilities of the evolutionary development algorithm by using it to generate partially self-building forms capable of locomotion, "shape-shifting" and sensing, suited to different environments. Key features to be examined will include the ability of computer-evolution and development to exploit intrinsic properties of the materials and environment, using them to create novel solutions, the ability of the solutions to develop differently in different environments, and the ability of the solutions to be robust and capable of self-repair during development. Some 14 research centres internationally have expressed their interest in this research, including NASA Jet Propulsion Laboratory, Science Applications International Corp. and BAE Systems. Dr Julian Miller of Birmingham University is also keen to act as an academic collaborator, testing the same ideas for a different application: electronic circuit design in radiation-damaged silicon and other materials.

It is anticipated that the research will pave the way to considerable advances in future technology, for example:

- new computational development algorithms capable of finding solutions to large or complex problems

- new methods to create autonomous, self-adapting forms tailored to different environments, capable of movement and sensing (for example, a device able to redesign and alter its morphology to satisfy different aerodynamic or radar requirements like a chameleon changing its color to avoid detection, or a device able to reconfigure itself to maximise solar power gathering efficiency like a plant growing towards light).

- new forms of electronics that exploit intrinsic properties of novel materials, resulting in lighter, smaller, more fault-tolerant, more complex circuits that may work in radiation-harsh environments.

REFERENCES

[Bentley 2001] "Digital Biology". Hodder Headline Press, London.

[Bentley and Corne 2001] (Eds) "Creative Evolutionary Systems" Morgan Kaufmann Pub., San Fran.

[Bentley,1999] (Ed) "Evolutionary Design by Computers" Morgan Kaufman Pub., San Fran.

[Bentley, Gordon, Kim and Kumar 2001] "New trends in Evolutionary Computation" Proceedings of the 2001 Congress on Evolutionary Computation (IEEE Cat.No.01TH8546).IEEE, Piscataway, NJ, USA; 2001; 2 vol. (xii+1441) pp. p.162-9 vol. 1.

[de Lacy Costello 1996] B. P. J. de Lacy Costello et al. Novel composite organic-inorganic semiconductor sensors for the quantitative detection of target organic vapours, J. Mater Chem., 1996, 6, 289-294.

[Grzeszczuk, Terzopoulos, and Hinton 1998] "NeuroAnimator: fast neural network emulation and control of physics-based models" Computer Graphics.Proceedings. SIGGRAPH 98 Conference Proceedings. ACM, New York, NY, USA; p.9-20.

[Evseev and Panchenko 2000] "Laser stereolithography for medical applications" In Proceedings-of-the-SPIE The-International-Society-for-Optical-Engineering. vol.4070; p.401-10.

[Miller, Job, and Vassilev 2000b] "Principles in the Evolutionary Design of Digital Circuits -- Part I", Journal of Genetic Programming and Evolvable Machines, Vol. 1, No. 1, pp. 8-35

[Pentecost, Icardo, and Thornburg 1999] "3D computer modeling of human cardiogenesis" In Computerized-Medical-Imaging-and-Graphics.vol.23, no.1; Jan.-Feb.; p.45-9.

[Thompson 1996] "An evolved circuit, intrinsic in silicon, entwined with physics", Proceedings of 1st Int. Conf. on Evolvable Systems:From Biology to Hardware (ICES96) Lecture notes in Computer Science, vol 1259, Springer-Verlag, pp. 390-405.