Project abstract

TitleOrganization of Semantic and Episodic Memory in Motivated Learning of Robots

Keywords: Semantic memory, episodic memory, motivated learning, neural networks.

Research objective:
The  main  goal of the project is  to  extend the current state-of-the-art  in  design mechanisms  for creation and organization of semantic and episodic memory in motivated learning of robots. Based on these  mechanisms  one can build the memory of autonomous  systems operating  in a changing complex environment. The episodic memory will be created in interaction with  dynamically constructed semantic memory. This is an innovative approach beyond the current understanding of organization of both types of memories.

The semantic memory  represents knowledge through  a set of concepts and associations between them. The knowledge increases gradually,  with  cumulated experiences. The project  will  develop  the semantic memory structure and algorithms for its self-organization.

The episodic memory records  sequences of episodes that are  relevant  to  the  system operation.  The main tasks in the construction of episodic memory systems is creating episodes, recalling memorized  episodes, and gradual  forgetting  of less useful  episodes. The project aims to implement these tasks. So far, there were no computational models integrating the both memory types.
Research method
The theoretical  foundation isthe motivated learning and goal creation system   developed by the principal  investigator  (PI), . Such systems  control an embodied intelligent agent  that  learns  how to effectively interact with the environment. It  has been shown that these systems learn better and faster than traditional reinforcement learning systems   considered  so far  as the most effective machine learning methods.

Under this project, the developed by the  PI  approach to the organization and interaction of cognitive memory systems  areas,  will be used to record behaviors and events relevant  to the system. They  in turn will be important  for  organization of  the  semantic memory.  It is assumed that  autonomous  systems will gain essential knowledge and learn proper  actions  through interactions with the environment. Therefore, construction of mechanisms and  self-organization of the  semantic and  episodic memory are  essential  for  the development of autonomous systems.

Structures of the semantic and episodic memory exploit long-term memory cells developed by the PI. These cells exhibit desirable properties such as fast learning and  tolerance  in recognition  of similar  spatio-temporal sequences,  and the ability to register the episodes’ significance.
Results impact
This work will  further develop  the pioneering research  of the PI  in motivated learning of autonomous systems.  The project will enhance the knowledge about the role of agent’s interaction with the environment in the development of perception, memory organization, prediction,  and use of memory in the development of motor skills and machine  intelligence. The project will develop machine learning methods and improve agent’s ability for autonomous operation.

The developed methods and algorithms will be disseminated in the scientific community through conferences, organization of a workshop, and a web page containing all the publications and the source code of programs.

One of the  expected  results is scientific  development  of  the faculty  and  involvement of  students  in the research process.  Another expected project outcome is  achieving  a  unique expertise in the construction and development of advanced control of  autonomous machines operating in a real environment and supporting human activities. less