BIO-INSPIRED CELLULAR SYSTEMS WITH CYCLIC METAMORPHIC MEMORY M. Samie Electronic Department Shiraz University Shiraz, IRAN G. Dragffy Computing, Engineering and Mathematical Sciences University of the West of England Bristol, United Kingdom E. Farjah Electronic Department Shiraz University Shiraz, IRAN Abstract - The unique characteristic of a biological system in nature is described by its DNA through the transcription of its genes. This is like a memory map that each cell of an organism contains. An artificial embryonic cell contains a similar memory map where the specific ‘gene’ it executes determines the functionality of the cell. An electronic system is then constructed by a large number of identical cells where each possesses a different behaviour. The cells collectively however determine the characteristic of the target system. This paper proposes a new, variable size memory map based on a novel gene selection algorithm that no longer uses the hitherto common address decoding approach to access the cell’s gene it will execute. Instead it applies the principle of cyclic metamorphic gene selection of the artificial DNA memory. A further benefit of the approach is that through genetic operators or variable memory space environment for enhanced behaviour the functionality of the system can also be easily altered. Keywords: Bio-Inspired Systems, Embryonics (embryonic electronics), Artificial Life, Artificial DNA, Self-Repair, Fault Tolerance. 1 Introduction The design of reliable electronic systems and ensuring their long-term fault free operation is one of the major challenges we are facing today. How can we design such complex but reliable systems? Nature offers some remarkable examples. One such important process is the development of the individual from a single fertilised egg (zygote) through its repeated division and differentiation. Embryonics (embryonic electronics) [1-5] tries to adapt and transpose the development of such processes and living characteristics of organisms to the world of silicon integrated circuits. Systems are built by a homogenous array of identical cells similarly to that of commercial FPGAs, but they posses self-replication, self-repair and fault-tolerant properties [6]. Unlike in nature however, if fault in a cell of a bio-inspired system develops or it dies, the silicon on which it is fabricated cannot grow new healthy cells to replace the damaged ones. In order to maintain fault free operation, in Embryonics spare standby cells are employed [7, 8]. These, as shown in Fig. 1, are usually distributed in rows and/or in columns around functional cells. Functional behaviour of an embryonic system is defined, analogously to biological systems, by its DNA stored in the memory of every one of its cells. The size of this memory and the required decoding circuit that selects the appropriate gene, that specialises the cell for a specific behaviour, can be large and consume a disproportionate area of the cell. Research to date [5, 9] tries to simplify and optimise this memory but little attention has been paid so far to more efficient memory address generation techniques. This paper attempt to address the later problem by introducing a novel gene selection technique that eliminates the need for address generation during cellular division and cellular differentiation as well as during a fault initiated system repair process. The efficiency of our metamorphic cell-memory based cyclic gene selection algorithms is demonstrated by the implementation of a frequency divider using a Xilinx XC9500 CPLD device. Sections 2 of this paper will give a brief overview work done to date in Embryonics within the international research community. Section 3 introduces our proposed cyclic gene implemented memory and how it aids the processes of cellular division and cellular differentiation. In Section 4 a new metamorphic memory map based artificial DNA is presented. Finally, implementation and simulation results of a frequency divider example are detailed in section 5, followed by a brief conclusion in Section 6.