論文

In this paper, we aim to build an agent model that enables cooperative behaviors by estimation human intention. Human changes their action decision process by other's behavior. So, we analyze human behavior with switching action decision process. We have targeted collision avoidance as an example of a simple cooperative behavior. We have set two agents with Meta-Strategy model to a virtual environment SIGVerse. We have analyzed subject's walking trajectories, when two agents have different behavior strategies. It was confirmed that subjects switch their avoidance behaviors by agents' strategies.

In goal-type ball games, such as handball, basketball, hockey or soccer, teammates and opponents share the same field. They switch dynamically their behaviors and relationships based on other players' behaviors or intentions. Interactions between players are highly complicated and hard to comprehend, but recent technological developments have enabled us to acquire positions or velocities of their behaviors. We focus on handball as an example of goal-type ball games and analyze causality between teammates' behaviors from tracking data with Hidden Semi-Markov Model (HSMM) and delayed Transfer Entropy (dTE). Although 'off-the-ball' behaviors are a crucial component of cooperation, most research tends to focus on 'on-the-ball' behaviors, and relations of behaviors are only known as tacit knowledge of coaches or players. In contrast, our approach quantitatively reveals player's relationships of 'off-the-ball' behaviors. The extracted causal models are compared to the corresponding video scenes, and we claim that our approach extracts causal relationships between teammates' behaviors or intentions and clarifies roles of the players in both attacking and defending phase.

We present a novel approach for detecting, classifying, and linking entities from Twitter posts (tweets). The task is challenging because of the noisy, short, and informal nature of tweets. Consequently, the proposed approach introduces several methods that robustly facilitate successful realization of the task with enhanced performance in several measures.

Realizing flexible cooperative group behavior of human and social robots or agents needs a mutual understanding of each intention or behaviors of participants. To understand cooperative intelligence in group behavior, we must clarify the decision-making process with intention estimation in multiple persons. Multi-people decision-making process have top-down intention sharing and bottom-up decision making based on the intention inference and amendment based on the each participants’ behavior. This study suggests the cooperative pattern task focusing on the selection process of others whom to be noticed and balancing process of each intention to achieve the shared purpose. In the 2D grid world of an abstract cooperative environment with restricted modal-ity of subjects, they communicate with each other in a nonverbal way and infer their intention based on their behavior to achieve the purpose. We analyzed the human subjects’ behavior and clarified their policy of behavior and concepts which assumed to be shared by each subject for preventing misunderstanding of each intention. Two main results were obtained through the experiment. First, optimal behavior based on the purpose in minimal steps prevent the misunderstanding of each intention. Second, the narrowing down the number of subjects who change their policy assumed to reduce the burden of intention inference.