Creation of a Neuromorphic Collective

The focus of my research has been in emergent behavior, especially as it relates to robotic systems. Within this work I discuss three aspects of this research, simulation, hardware, and embedded systems, each of which had use in swarm behavior, and the latter two have applications outside of this niche as well. The simulation was a swarming application on the TENNlab framework. The best networks were manually reviewed to determine how accurately they completed the task and therefore how faithfully the fitness function captured the intent of the program. Testing those functions was one of the largest parts of this project. Due to being such an open-ended project the performance is difficult to summarize, but the networks developed well in simulation and translated fairly easily into hardware. The hardware began as a dedicated platform for the simulation but became its own project with the goal of being easy to use and repurpose. Its success was measured primarily in binaries of whether or not certain features worked as intended. The robot was found capable of precise movement and easily applicable to a range of projects. The embedded system was not integrated into the other two projects but came about as a result of working with the tools available in the lab. It is a variation of the neuromorphic kit where the encoding and decoding units are combined onto a single processor. I measured the speed, power, energy, and variability of use when scoring the results of this variation. In each of these categories my kit matched or exceeded the baseline set by the original. Some of the changes could also be implemented into the original with little effort.