Spiking Neural Controller for Autonomous Robot Navigation in Dynamic Environments

摘要

In this paper, a neural controller based on Spiking Neural Network (SNN) is trained using the Reward-modulated Spike-Timing-Dependent Plasticity (R-STDP) learning approach for the tasks of simultaneous target tracking and obstacle avoidance. The neural controller has two separate layers with a fully connected architecture. A random number vector encodes the sensor data within the network, and its output is obtained by calculating the membrane potential of the output layer. The SNN is connected to a 2 DoF robotic arm with two degrees of freedom and to control the motors. Two moving objects are used as targets and obstacles. The results showed that the network is able to distinguish between two objects in the environment. After learning, the robot found the proper path to reach the target without colliding the obstacle.