Speeding-up digital ecologies evolution using a hardware emulator: Preliminary results

Abstract

For a more than a decade, the idea of applying the biological principle of natural evolution to artificial systems in order to create or to improve digital ecologies has emerged from different laboratories. During the past couple of years, a new trend consists in applying these investigations to hardware design. This concept is called "Evolvable Hardware". For this quest, hardware emulation offers an alternative approach to the development of a generic evolvable system including fitness evaluation. Compared to a software solution, emulation can be on the order of a million times faster which is of higher interest when billion steps of evolution are necessary. A further advantage of emulation is to provide the description of the VLSI to be implemented as well as a validation of its behavior. In this paper, we describe the way followed to implement the system (cellular automata and the surrounding evolutionary control logic) as a hardware description in an emulator. For different examples presented in this paper, reasonable with respect to simulation, processing time of hardware emulation versus software simulation are compared. The time saved by hardware emulation has given the opportunity to increase the complexity of the "evolving organism" by including the selection of intervening neighbors in the parameter selected by evolution.