Stigmergic reinforcement learning using All-Optical soltionic X-junctions
Ethology has shown that animal groups or colonies can solve complex problems distributing
simple decision-making processes to the group members. This distributed intelligence is very diffuse in nature. It is based on stigmercy: the possibility of spreading information between subjects by modifying the enviroment inside which such subjects move on. For example ant colonies can optimize the trajectories towards the food by performing both a reinforcement (or a cancellation) of the pheromone traces and a switch from one path to another with stronger pheromone. Such ant’s processes can be
implemented in a photonic hardware to reproduce stigmergic signal processing through reinforcement learning. We present innovative, completely integrated X-junctions realized using solitonic waveguides which can provide both ant’s
decision-making processes. The solitonic waveguides act as pheromone traces: IR signals can follow them propagating inside. The intensity of pheromone is here reproduced by the refractive contrast of the waveguides which depends on the intensity of the writing pumping light. Crossing two solitonic waveguides would result in the realization of a X-junction whose switching rate is fixed by the refractive contrasts of the single channels. The realized X-junction can switch from symmetric (50/50) to asymmetric behaviors (80/20) using optical feedbacks, letting unused output channels vanish or the used ones reinforced.