Rezumat articol ediţie STUDIA UNIVERSITATIS BABEŞ-BOLYAI

În partea de jos este prezentat rezumatul articolului selectat. Pentru revenire la cuprinsul ediţiei din care face parte acest articol, se accesează linkul din titlu. Pentru vizualizarea tuturor articolelor din arhivă la care este autor/coautor unul din autorii de mai jos, se accesează linkul din numele autorului.

Modern robots are not build to solve pre-determined tasks, rather they are designed to tackle a wider class of problems. Finding efficient control algorithms for a new problem within the class is not straight- forward. Machine learning techniques, e.g., reinforcement learning (RL) proved to provide suitable methods in finding such control algorithms. Robotic control learning tasks share several common properties, thus, when selecting among RL methods one has to consider these properties. In this paper, we present the state-of-the-art RL algorithms from the perspective of robotic control. We highlight their advantages and drawbacks in conjunction with robotic control, hereby, analyzing their feasibility in this context. Our results are supported by simulated pole balancing control experiments.

Key words and phrases. machine learning, reinforcement learning, robotics, policy gradient.