Dynamic learning in cognitive robotics through a procedural long term memory

Brain-like robotic approaches aim to reproduce the complex processes occurring within the biological brains to achieve a higher level of autonomy. One of the key aspects of these approaches is dynamic learning, that is, how to provide the cognitive architectures that control de robot with adaptive learning capabilities. Several options have been considered in this line in the field of Cognitive Robotics, although the development of a proper memory system has provided the best practical results up to now. This work also follows this approach, seeking to show the advantages of using a Long-Term Memory (LTM) for optimizing the adaptive learning capabilities of a cognitive robot in dynamic environments. Specifically, a procedural LTM that stores basic models and behaviours is included in the evolutionary-based Multilevel Darwinist Brain (MDB) cognitive architecture. The LTM management system that has been developed to control when a model must be stored or replaced is presented here in detail. Moreover, a Short-Term Memory (STM) sub-system included in the MDB is also explained due to its strong relationship with the operation of the LTM. The LTM elements are tested in theoretical functions and in a simulated example using the AIBO robot in a dynamic context with successful adaptive learning results.