BIOLOGICALLY INSPIRED TRANSPORTATION: A DISTRIBUTED INTELLIGENT CONVEYOR (EPSRC EP/H023631/1)
A parallel manipulator is a massive array of simple individual actuators with a small power density that collectively transport and position objects with masses considerably higher than the force generated by a single actuator alone. This design is inspired by the biological phenomena of cilia, small hair-like structures on the surface of cells which can either sense local properties such as in the rod photoreceptors for vision or in olfactory neurons for smell, or can move in coordinated wave action to move liquid over their surface, as in the trachea and kidneys. Employing these capabilities in an analogous array of micro-actuators will produce a conveyor of parallel intelligent manipulation able to sense object properties, move them in different directions and effectively sort objects according to their properties. Crucially, the actuator array will be capable of communicating local information about objects to other parts of the array to enable coordinated action.
LEARNING AND COMPUTATION IN DISORDERD NETWORKS OF MEMRISTORS: THEORY AND EXPERIMENT (EPSRC EP/H014381/1)
Memristor (memory resistor) is a device whose resistance changes depending on the polarity and magnitude of a voltage applied to the device's terminals and the duration of this voltage's application. The memristor is a non-volatile memory because the specific resistance is retained until the application of another voltage. A memristor implements a material version of Boolean logic and thus any logical circuit can be constructed from memristors. We propose to fabricate in laboratory experiments an adaptive, self-organized disordered network of memristors. This practical fabrication will be backed up by rigorous computer simulation experiments. The memristor network is comprised of a conglomerate of conductive polymer fibres interspersed with particles of solid electrolyte. The conglomerate is placed on a matrix of micro-electrodes capable of recording voltage and generating current sources and sinks. Machine learning techniques will be applied in order to design logical schemes and basic arithmetical circuits.
MOULD INTELLIGENCE: BIOLOGICAL AMORPHOUS ROBOTS (The Leverhulme Trust Research Proposal F00577/I)
We propose to combine our unique experience in designing theoretical and laboratory prototypes of reaction-diffusion chemical computers, methods of non-linear medium based computing and our techniques in self-organized and biologically inspired robotics to design experimental prototypes of an amorphous biological robot. The proposed design of the amorphous robot will be based on vegetative state of unicellular organism Physarum polycephalum. We aim to experimentally demonstrate that the plasmodium robot senses data-objects represented by sources of nutrients, selectively spans them in optimal proximity graphs by protoplasmic veins, calculates the shortest distance between any given data-objects, and is capable of intelligent distributed sorting, transporting and assembling of light-weight objects. See Jeff Jones and Soichiro Tsuda.
REPRESENTATION OF PARTICLE DYNAMICS IN CELLULAR AUTOMATA BY REGULAR LANGUAGES (EPSRC EP/F054343/1)
Cellular automata are spatially extended discrete systems, where local sites simultaneously update their states depending on states of their immediate neighbours. We are dealing with cellular automata that exhibit dynamics of travelling self-localizations, or particles, compact mobile patterns composed of non-quiescent states. Studies of cellular-automaton particles bring significant values in a range of disciplines, from non-linear sciences to computer engineering. The automaton particles are discrete analogues of breathers, solitons, excitons, kinks, defects and other localizations observed in natural systems. The discrete travelling self-localizations are used as signals and modulators in collision-based unconventional computing architectures. We propose a novel way to formally typify, and dynamically classify, and to estimate computational potential of interacting particle systems. We represent dynamics of interacting automaton particles by sets of regular expressions, and construct deterministic finite-state machines generating the expressions derived from the localization dynamics. See Genaro Martinez.
DYNAMICAL LOGICAL CIRCUITS IN SUB-EXCITABLE CHEMICAL MEDIA ( EPSRC EP/E016839/1)
Pilot project: PhD Studentship in Collision-Based Computing in Excitable Media
Two PhD studentship positions aimed for interdisciplinary research, knowledge transfer and training in non-classical computation, the field of science emerging at the boundary between novel and nature-inspired computing paradigms and architectures, and non-linear chemistry and physics. We will explore complex behaviour and interaction dynamics of localised mobile excitation patterns to design working prototypes of architectureless computing devices. We experimentally demonstrate that excitation wave-fragments in a Belousov-Zhabotinsky (BZ) medium with immobilised catalyst can be used to build elementary logical gates and circuits, and to simulate certain classes of mathematical machines. We represent information values of variables by the presence/absence or structure of wave-fragments. The wave-fragments may annihilate, fuse, split and change their velocity vectors as a result of collision with other wave-fragments --- thus the values of variables represented by the wave-fragments change and certain logical operations and information processing operations are implemented. We aim to theoretically design and experimentally build a universal collision-based massively-parallel processor based on excitable chemical reactions, where the space-time dynamics of travelling localised excitations realise embedded logical and arithmetical circuits and mathematical machines. See Liang Zhang.
GENERAL PURPOSE SPATIAL COMPUTATION ( EPSRC EP/F003811/1)
To address inherent limitations in current research in non-classical computing in unstructured non-linear media, to develop formalism of amorphous computing and to make nature-inspired computing devices scaleable and programmable we aim to develop and advance novel paradigm, called blob computing. A blob is a generic primitive used to structure a uniform computing medium into an easier-to-program parallel virtual machine: a self-developing and self-mapping network of automata. Blob machines mimic cellular biological development. Blobs acts like amorphous media contained in membranes. A blob can divide and produce a network of blobs able to perform computation in a parallel way. Blobs are designed to be as simple computational building blocks as possible, so they can run on an arbitrary uniform computing medium. Blob instructions include: division, encapsulation and deletion. Blobs define an autonomous region, structuring the computing medium into a higher level, easier to program virtual machine. While the problem of programming this virtual blob machine has been studied in depth, the implementation of the virtual machine on a computing medium has only been partially performed. Funding is sought to study and implement the blob representation and develop minimal set of the blob instructions on a hexagonal lattice. The project will address a theoretical challenging problem of discretization of a membrane-line object (we will exploit findings from mathematical physics and numerical analysis). The project also poses a precise experimental framework: well defined formal specification of blob computers, so that parallel algorithms using these membranes as building blocks exhibit competitive performance in time and space. Results of the project can be deployed in design and manufacturing of future architecture-less high-performance computers. See Fred Gruau.
EMERGENCE OF COMPUTATION IN CELLULAR AUTOMATA: PHENOMENOLOGY OF MOBILE SELF-LOCALISATIONS DYNAMICS ( EPSRC EP/D066174/1)
We deal with emerging computational paradigm of structureless, or collision-based, computing in discrete spatially extended systems. The computation is based on spatio-temporal interaction of self-localisation mobile patterns (defects, gliders, quasi-particles). Data and results of the computation are to be represented by initial and final configuration of the localisations (generators of localisation or bound states); algorithm of computation is represented by initial trajectories and self-localisations and configurations of reflectors (stationary patterns that change trajectories of the localisations). The pilot project is the first phase of designing structureless computers in non-linear system: we aim to study cellular automata with binary and ternary cell states, discover all types of mobile localisations, characterise their interaction dynamics and develop analytical tools for automatic search of localisaton-supporting discrete media. See Genaro Martinez pages for more info.
Evolutionary Design of Collision-based Computing Schemes in Two-Dimensional Cellular Automata ( EPSRC EP/E005241/1)
We wish to evolve novel types of unstructured computational devices, where computation is based on collision between gliders (cellular-automaton analogues of breathers, solitons, defects) travelling in uniform space free from heterogeneity imposed by traditional computing architectures. A significant contribution --- nature-inspired design solutions and computing architectures --- will be made to the high-potential field of novel and emerging computational paradigms and architectures, and non-classical computation. Results of the project will be deployed in future automatic design and manufacturing of architecture-less high-performance computers. See Emanuel Sapin page for more info.
Non-linear Media based Computers: Chemical & Neuronal Networks through Machine Learning ( EPSRC GR/T11029/01)
There is growing interest in research into the development of hybrid wetware-silicon devices focused on exploiting their potential for 'non-linear computing'. The aim is to harness the as yet only partially understood intricate dynamics of non-linear media to perform complex 'computations' (potentially) more effectively than with traditional architectures and to further the understanding of how such systems function. The area provides the prospect of radically new forms of machines and is enabled by improving capabilities in wetware-silicon interfacing. The research proposed here will present an approach by which networks of non-linear media - neurons and reaction-diffusion systems - can be produced to achieve a user-defined computation (or behaviour) in a way that allows control of the media used and the substrate in which they exist. Simulated evolutionary algorithms will be used to design the appropriate network structures by searching a defined space of preparation procedures to create a computing resource capable of satisfying a given objective(s). See Larry Bull page for more info.
Smart Table: Massively Parallel Actuator Array Controlled By Non-Linear Medium ( EPSRC GR/S79640/01)
We will develop theoretical foundation for designing and programming parallel non-linear medium based manipulator, which intelligently transports, filters, orients and positions several objects at a time. The parallel manipulator will have a modular structure; its control units and effectors will be arranged in a regular locally connected network. All actuators, sensors and elements of the non-linear medium will work in parallel. In designing the computational model of manipulator we will realise novel ideas of actuation in a non-linear medium; manipulating objects by the employment of spreading and growing patterns of diffusive wave-fronts and waves of excitation. The control of the actuator will be based on advances and findings in the fields of non-linear physics, smart materials and structures, space-time dynamics in disordered media, phenomenology of pattern formation and emergence of computation in optical, reaction-diffusion chemical and biological excitable media. See Sergey Skachek page for more info.