Scientific Interests and Approach
Generally speaking, I am interested in learning from biological systems and applying this knowledge in technological domains. After all, nature has been developing solutions to real world problems for billions of years, so there is much to be gained by borrowing successful concepts and design principles. To succeed in this "industrial espionage", one must first understand the system well enough to separate key principles from the unimportant details peculiar to the biological implementation.
Success in this challenging task is likely to require a collaborative, multi-disciplinary effort involving, among others, biologists and engineers. But this introduces its own challenges. Indeed, scientists from either side of this divide often struggle to understand where the other is coming from, due to fundamental differences in outlook and skill set. What is needed, therefore, is a community of truly inter-disciplinary scientists who can bridge this gap. By bringing an engineering approach to the study of biological systems, I hope to contribute not only to the development of better technologies, but also to our fundamental understanding of the mechanisms of life.
Initially, my interest in robotics led me to take a bachelors in electrical engineering at the University of Cape Town, but I became increasingly fascinated with biological systems after realising the extent to which even simple animals outperform robots in terms of robustness and flexibility. I therefore went on to do an MSc in the same department, but on a topic closer to computational neuroscience. During this two year research project I explored the field of neural modelling, and wrote a thesis on synaptic plasticity in the Aplysia californica gill-syphon withdrawal reflex. Having been bitten by the bio-bug, I went on to do a PhD here at Leeds, studying the locomotion of the nematode worm C. elegans (see below).
In early 2010 I completed a PhD entitled "C. elegans locomotion: an integrated approach" (PDF), as part of a highly cross-disciplinary project involving biology, mathematics, computer science and engineering. I was supervised by Professor Netta Cohen and was part of the BioSystems research group in the School of Computing. My work also involved extensive collaboration with the Hope Laboratory in the Faculty of Biological Sciences.
Introducing C. elegans
Caenorhabditis elegans is a small nematode worm which is an ideal model organism for both genetics and systems biology. In addition to the huge wealth of genetic information available about the worm, we even know the structure and connectivity of its small and invariant 302-neuron nervous system. Thus C. elegans offers the best possible opportunity to understand and model the inner workings of a complete organism at the single cell level. Furthermore, despite the relative simplicity of its nervous system, the worm is capable of a remarkably rich repertoire of behaviours, almost all of which are mediated by locomotion.
|M1: C. elegans locomotion in increasingly (from left) resistive media|
Deciphering the Locomotion System
The goal of our project was to apply an integrated approach to further our understanding of C. elegans locomotion. Forward locomotion involves smooth propagation of undulations from head to tail, but the properties of the locomotion wave (i.e. frequency, wavelength etc.) are dramatically different when the worm swims in water versus when crawling on a firm gel (agar) substrate. In fact, it has generally been thought that these are two fundamentally distinct forms of locomotion. But how does a creature with such a limited neural circuit generate both of these behaviours?
To address this question, we used a combination of biological experiments, computational data analysis and computer simulations, with each of these components feeding back and complementing the others. Our first key contribution was in developing a new experimental approach that allowed us to demonstrate that the worm's ``swimming'' and ``crawling'' are not, in fact, distinct. Rather, they are snapshots from a single behaviour that is strongly modulated by the physics of the environment. Intermediate behaviours can be revealed by placing worms in appropriate (intermediate) environments, as shown in movie M1.
|M2: Locomotion of the virtual worm in increasingly (from left) resistive media|
In light of this finding, the next step was to account for the generation and modulation of the locomotion wave with the worm's limited neural resources. To this end I developed an integrated neuro-mechanical model - including representations of the neurons, muscles, body and environment - that accounts for the full range of forwards locomotion behaviours (see movie M2). The model suggests that C. elegans may be exploiting some unusual design principles (including bistable, non-spiking neurons and sensor driven oscillations) to achieve robust locomotion with minimal processing. In fact, the model adapts to external constraints without explicitly sensing anything about the environment, as shown in movie M3.
From Worm to WormBot
|M4: Prototype wormbot|
|M3: Locomotion with external constraints|
Worm- or snake-like robots have the potential to excel in many applications where wheeled, tracked or even legged robots would struggle, but the control of such robots is challenging. Having seen how easily the model adapts to external constraints, I became interested in the idea of using the virtual neural circuit as a controller for a serpentine robot. Particularly appealing is the way the model passively accommodates obstructions without direct sensing, representation or planning. Clearly such a control scheme would need to be complemented by some higher level planning and control, but as a low level behaviour it seemed promising.
As a first test of this idea, I undertook a project with Jon Tapson and Tony Lewis at the 2009 Telluride Neuromorphic Engineering Workshop in which we developed a crude robotic implementation over a couple of weeks. The robot, shown in M4, has a rigid, articulated body much like some previous lamprey robots, but is controlled by an on board electronic implementation of my C. elegans neural model. While the behaviour is clearly a bit clumsy, I was encouraged by the relative ease with which forwards propulsion was achieved.
Recently I have been able to pursue this idea further thanks to a 12 months fellowship from the EPSRC. My goal was to develop a more serious WormBot prototype aimed at future search and rescue applications in difficult terrain, like the rubble of collapsed buildings.
Jordan H. Boyle and Netta Cohen Physics matters: an integrated model of C. elegans locomotion (in preparation)
Jordan H. Boyle, Stefano Berri, Ian A. Hope and Netta Cohen It's not a choice it's a reflex! Front. Behav. Neurosci. (accepted)
Netta Cohen and Jordan H. Boyle Swimming at low Reynolds numbers: a beginners guide to undulatory locomotion Contemp. Phys. Vol. 51 pp. 103-123 2010 (PDF)
Bao Kha Nguyen, Jordan H. Boyle, Abbas A. Dehghani and Netta Cohen A C. elegans-Inspired Micro-Wormbot with Polymeric Actuators Proc. ROBIO 2010
Stefano Berri, Jordan H. Boyle, Manlio Tassieri, Ian A. Hope and Netta Cohen Forward locomotion of the nematode C. elegans is achieved through modulation of a single gait HFSP J. Vol. 3 pp. 186-193 2009 (PDF)
Jordan H. Boyle and Netta Cohen Caenorhabditis elegans body wall muscles are simple actuators Biosystems Vol. 94 pp. 170-181 2008 (PDF)
Jordan H. Boyle, John A. Bryden and Netta Cohen An integrated neuro-mechanical model of C. elegans forward locomotion LNCS Vol. 4984 pp. 37-47. Springer-Verlag Berlin 2008 (PDF)
Jordan H. Boyle and Netta Cohen The role of body wall muscles in C. elegans locomotion Proc. IPCAT 7 pp. 363-375 2007 (PDF)