Policy gradient learning for quadruped soccer robots |
| |
Authors: | A Cherubini F Giannone L Iocchi D Nardi PF Palamara |
| |
Affiliation: | 1. Aix Marseille Université, CNRS, EFS-AM, ADES UMR7268, 13344 Marseille, France;2. Université Paul Sabatier, CNRS, AMIS UMR5288, 31073 Toulouse, France;3. Service de Parasitologie-Mycologie, Centre Hospitalier Universitaire de Toulouse/INSERM UMR1043/CNRS UMR5282/Université de Toulouse UPS, Centre de Physiopathiologie de Toulouse Purpan (CPTP), 31300 Toulouse, France;4. CIC-EC Antilles Guyane CIE 802 Inserm, Centre Hospitalier Andrée Rosemon, France;5. Equipe EPaT EA 3593, Université des Antilles et de la Guyane, Cayenne, Guyane française, France;6. Établissement Français du Sang Alpes Méditerranée, 13005 Marseille, France |
| |
Abstract: | In real-world robotic applications, many factors, both at low level (e.g., vision, motion control and behaviors) and at high level (e.g., plans and strategies) determine the quality of the robot performance. Consequently, fine tuning of the parameters, in the implementation of the basic functionalities, as well as in the strategic decisions, is a key issue in robot software development. In recent years, machine learning techniques have been successfully used to find optimal parameters for typical robotic functionalities. However, one major drawback of learning techniques is time consumption: in practical applications, methods designed for physical robots must be effective with small amounts of data. In this paper, we present a method for concurrent learning of best strategy and optimal parameters using policy gradient reinforcement learning algorithm. The results of our experimental work in a simulated environment and on a real robot show a very high convergence rate. |
| |
Keywords: | |
本文献已被 ScienceDirect 等数据库收录! |
|