Abstract Concept Learning Approach Based on Behavioural Feature Extraction

In this paper, we propose a novel approach in which an intelligent agent can learn complex concepts in abstract forms. This approach provides a useful tool for non-episodic problems, where agent must search the environment to find special concepts; in addition, yielded abstract representation of the concepts can be used in further high level planning tasks. In order to perform concept learning process in this framework, agent utilizes its own actions according to limitations of sensory data and complexity of related analysis. It extracts required features from environment according to complexity of concepts and their distinctions. These features are composed of sequences of agent’s primitive actions. The proposed method is tested on a mobile robot benchmark, and learned concepts are used for a path planning problem. The simulation results demonstrate the capability of our approach in abstracting concepts.