A Simple Aplysia-Like Spiking Neural Network to Generate Adaptive Behavior in Autonomous Robots

Abstract

In this article, we describe an adaptive controller for an autonomous mobile robot with a simple structure. Sensorimotor connections were made using a three-layered spiking neural network (SNN) with only one hidden-layer neuron and synapses with spike timing-dependent plasticity (STDP). In the SNN controller, synapses from the hidden-layer neuron to the motor neurons received presynaptic modulation signals from sensory neurons, a mechanism similar to that of the withdrawal reflex circuit of the sea slug, Aplysia. The synaptic weights were modified dependent on the firing rates of the presynaptic modulation signal and that of the hidden-layer neuron by STDP. In experiments using a real robot, which uses a similar simple SNN controller, the robot adapted quickly to the given environment in a single trial by organizing the weights, acquired navigation and obstacle-avoidance behavior. In addition, it followed dynamical changes in the environment. This associative learning scheme can be a new strategy for constructing adaptive agents with minimal structures, and may be utilized as an essential mechanism of an SNN ensemble that binds multiple sensory inputs and generates multiple motor outputs.