Robots that imagine – can hippocampal replay be utilized for robotic mnemonics?

Neurophysiological studies on hippocampal replay, which was a phenomenon first shown in rodents as the reactivation of previously active hippocampal cells, has shown it to be potentially important for mnemonic functions such as memory consolidation/recall, learning and planning. Since its discovery, a small number of neuronal models have been developed to attempt to describe the workings of this phenomenon. But it may be possible to utilize hippocampal replay to help solve some of the difficult challenges that face robotic cognition, learning and memory, and/or be used for the development of biomimetic robotics. Here we review these models in the hope of learning their workings, and see that their neural network structures may be integrated into current neural network based algorithms for robotic spatial memory, and perhaps are particularly suited for reinforcement learning paradigms.