Learning decision making for Soccer Robots: A crowdsourcing-based approach

From the early days of Soccer Robotics, they were not supposed to be only another funny game and entertaining hobby. They were shaped around a specific idea in the mind, to beat FIFA World Cup champions by 2050. In this regard and based on the prospective roadmap, soccer robot field has become a fertile testbed for developing Artificial Intelligence (AI) and machine learning algorithms and methods as well as mechanical equipments design. Despite the advancements achieved over the years, it seems there is a long way ahead to meet the determined goal, at least from the “soft” perspectives, including decision making, action selection, etc. In other words, lack of human-level intelligence to provide robots with capabilities to compete with human counterparts is a great drawback. To cope with such issues, a working solution may be to train soccer agents by humans. Following this idea, in this paper a conceptual view of a novel crowdsourcing-based framework for soccer robots to learn (human-like) decision making is proposed. Also, different aspects of the system are studied. It is believed that such an approach can greatly improve soccer robots' ability to make proper and effective decisions - and subsequently selecting best actions- in different situations. Moreover, it could be regarded as a first step towards equipping soccer agents with necessary human-level requirements for the ultimate goal.