Spike-Timing Dependent Plasticity Learning for Visual-Based Obstacles Avoidance
Hedi Soula and Guillaume Beslon
Simulation of Adaptive Behavior 2006 (SAB 2006)
Rome, Italy, 25-29 September 2006
Summary
In this paper, a recurrent spiking neural network is trained on a robot to learn to avoid obstacles relying only on its visual flow. The training is conducted using 'spike-time dependent plasticity'-like (STDP) learning rule during the "life" of the agent. The task was designed to compel the robot to make use of its controller's dynamics. As such, the temporal integration needed to perform it will be supported by the dynamics of the neural network. The behaviors of avoidance we obtain are homogenous and elegant. In addition, we observe the emergence of a neural selectivity to the distance after the learning process.