Biological systems can adapt excellently to the demands of a dynamic world and changing tasks. What kind of information processing and reasoning do they use? There are numerous studies in psychology, cognitive neuroscience and artificial intelligence which complement each other and help in getting a better understanding of this riddle. Our paper presents a biologically inspired architecture for a spatiotemporal learning system. Multiple interconnected memory structures are used to incorporate different learning paradigms. Concurrent inherent learning processes complete the functionality of corresponding memory types. Our architecture has been evaluated in the context of mobile rescue robots: The task consists of searching objects while navigating in an unknown maze.

The presented work aims at developing an autonomous rescue system with flexible behavior. The embodied cognitive system is based on the interrelated multiple memory structures whose functionality is modeled at a high level of abstraction. The biologically motivated learning processes are incorporated in the system corresponding to the kind of memory they belong to. The system was tested and evaluated in an unknown maze in a simulator by searching and navigating a set of specified objects in a given order.

This paper describes a cognitive architecture that is predominantly substantiated by data processing principles of biologically inspired multiple memory systems. The current memory configuration involves episodic, semantic, procedural and working memory units. We introduce the reader to essentials of the mammalian memory system and explain our highly abstracted software realization of it as a distributed, concurrently communicating node structure. We also show our first results of its implementation for cognitive function as novelty detection and exploration of a maze in the case of simulated autonomous mobile robot.