Towards Multi-Agent Intelligent Systems with Ability of Incremental Learning

Towards Multi-Agent Intelligent Systems with Ability of Incremental Learning

The goal of this project was to establish collaboration between the Artificial Intelligence Laboratory of the Faculty of Computer and Information Sc., Univ. of Ljubljana, and the Department of Cybernetics and AI, Faculty of Electrical Eng. and Informatics, Technology University of Košice. Most of the work was done during two working visits of researchers from Košice in Ljubljana. Two research topics for collaboration in a potential European project were defined:

1. The programming of robots by learning from human demonstration or controlling the robot to carry out a given task, and
2. Robot planning by means-ends technique which would enable the explanation by the robot of how the robot intends to perform a task and what is the purpose of actions the robot is executing.

During their visit, two researchers from Košice were working in Ljubljana on a concrete task of programming a Nao robot through human demonstration, and they were capturing human motion using our motion capture equipment Animazoo. The captured human trajectories were the basis for an adapted imitation by the robot of human demonstration, and later at FRI for the development of a method for automatic learning of compositional hierarchies in sensory-motor systems.