Closed Loop Control Systems Using Neural Learning Populations

Abstract

We begin by describing the photon neuron project, an effort to create a new photonic computational primitive based upon neural processing, and its potential to open new application domains. In particular, we focus on ultrafast control systems for high speed avionics. Then we discuss the strengths of Nengo, a neural simulator, which are its similarities to the physical photonic neuron implementation and its integration of control theory. Finally, we simulate basic control systems via function estimation using neural learning populations in Nengo. With these simulations, we show that artificial neural networks, like the photonic neuron, have the potential to be used in the construction of ultrafast control systems.