No abstract

In this paper we present a biologically-inspired model of spatio-temporal learning in the hippocampus and prefrontal cortex which can be used in tasks requiring the behavior of the robot to be constrained by sensory and temporal information. In this model chains of sensory events are learned and associated with motor actions. The temporality of these sequences is also learned and can be used to predict the timing of upcoming events. The neural network acts as a novelty detector and can modulate the behavior of the robot in case its actions do not have the expected consequences. The system is used to solve two different robotic navigation tasks involving an alternation between random exploration, goaldirected navigation and waiting periods of various lengths.