Robotic episodes learning for building cognitive experience map

This paper is about incremental online learning of robotic experience under uncertainty based on biology-inspired episodic memory. The episodic memory-driving Markov decision process (EM-MDP) with a novel episode model is proposed by imitating the firing patterns of hippocampal CA1 neurons, to simulate the organization process of episodic memory. Episodes learning method is presented by utilizing sparse distributed memory (SDM) for building experience map, and the uncertain information of system is considered. The experience map has the bionic cognitive abilities of self-organizing, self-learning and adaptivity based on episodic memory. Experimental results show that the approach has robustness to uncertainty and can realize memory real-time storage, incremental accumulation, integration and updating.