Kota Itoda

Understanding of human flexible cooperative behavior is important for making social agents and robots, and also construction of cooperative models and simulations are significant for constructive understanding of human group behavior. We have analyzed human group behavior including intention inference in groups and modeled their action selection process using cooperative pattern task so far. As a result, while some methods of intention inferencing and differences of roles in the groups have been revealed, subjects’ behaviors are not enough to explain. Therefore to extend our model, we implement leading behaviors for explicit expression of their roles and intention in the groups observed in subjects experiments in the pattern task.

Flexible and cooperative human group behavior are realized by changing our intentions and behaviors based on dynamic estimation of other participants ’intention, and also adjustment of self and others’intention. We analyze human small group behavior using cooperative pattern task in 2D grid world to clarify an individual action selection process including inference of others ’intention and adjustment of intention among participants. In previous research, we have constructed behavior strategy models based on the human behavioral experiments, implemented the models to cooperative agents, and confirmed the goal achievement in almost the same steps to humans in the agent simulations. In this research, we analyze combinations of human behavior strategies realizing group behavior by comparing agent behavior to subjects behavior.

Understanding of human flexible cooperative behavior in groups is important for constructing intelligent systems cooperate with human. In group behavior, top-down action selection based on sharing the participants' intention and bottom-down intention adjustment based on estimated intention from each behavior are considered to be done. To construct agent models realizing such process, we have analyzed the timing of dynamical change of intention and strategies of adjusting each intention in the group using pattern task which abstracting the group behavior including cooperative selection of intention and action for achieving the particular goal. In this research, we construct an agent model based on the analysis results and check the action selection process of each strategies using model simulations.

We have been modeled based on real soccer data analysis in RoboCup Soccer Simulation 2D. In the case of a soccer game, it is considered that soccer players share cooperative patterns in a pass scene. In this research, we analyze the positional relationship between teammate and opponent team member using Delaunay Triangulation. Furthermore, the key player which receives a pass is extracted from the gaze direction of the ball holder.

In order to understand human flexible cooperative behavior in a group, we have developed a cooperative pattern task requiring switching of intention and behavior. Through the analysis of the behavioral experiment using the task, subjects’ optimal behavior to a shared target in minimal steps and majority voting based action selection are clarified. In this research, agent models including "self- preferred selection" and "estimation of other agents’ intention" are constructed to reveal implicitly shared unobservable strategy under the subjects’ behavior. Moreover, we analyze the subjects’ action selection process by model-based agent simulation and comparison between subjects’ and agents’ behavior.

We have a behavior experiment using pattern task abstracting cooperative behaviors that require intention estimation and action switching to specific goals. And we have analyzed strategies to adjust cooperative intention estimations. In this research, we constructed an agent model that have three strategies of "random selection", "self-priority selection", and "other agent's target pattern estimation". And the decision making process was verified by simulation.

In a goal type ball game such as soccer and handball, a plurality of players who can pass are searched for, and each player's intention is estimated and a player who can pass is selected. Furthermore, it con rms the position and behavior of enemies that exist around passable players, estimates their intentions, and determines teammate players who pass the pass most successfully. In order to realize instantaneous intention estimation and judgment subject to strong temporal and spatial constraints, cooperative patterns shared within the group are considered to exist. Therefore, in this research, we presented to subjects a rst-person perspective of professional soccer players by using virtual environment, and analyzed the gaze behaviors pre- and post-training. Based on the results, we discuss that subjects switch their behavior by estimating the intention of otherplayers who pass and receive to subject by presenting the visual information based on the rst-person perspective.

In recent cooperative behavior study with goal type sports, some devices such as action cams and acceleration sensors are used for measuring personal behaviors. In these measurements, we mainly analyze the output of player’s behavior. But in cooperative behavior, self and other persons confirm mutual actions and share their intention estimation. And they decide their purpose and judge the next own action. In this research we analyze the intention estimation by player’s head movement and model it to simulate the estimation process. We experimented soccer pass behavior and measure player’s head movement by a flying drone. We describe the measurement accuracy of the player's head movement by the drone and the analysis result of the timing when a ball holder shares the intention of other players and passing.

ハンドボールやサッカーといったゴール型ボールゲームでは，敵味方選手が同じフィールドを共有し，お互いの意図の推定に基づき，選手の時空間的な関係を動的に切り替える．本研究はゴール型ボールゲームとしてハンドボールに着目し，隠れマルコフモデル(HMM)及びトランスファー・エントロピー(TE)を用いて選手行動の因果性を分析することで，攻撃や守備といった集団行動における選手の役割と関係を明らかにした．

In recent years, autonomous agents have been developed using statistical and probabilistic machine learnings together with deterministically optimized control. In this paper, a new decision making and motion generation method is proposed for adapting to uncertain environments. In the proposed method, we concentrate on passing in soccer, as a tactical group behavior, and understand how the optimization occurs in a group behavior from individual decision makings. In particular, we have quantified how people pass in plays by analyzing a video and tracking data of real soccer, and have constructed pass models with optimized parameters using logistic regression based on the analysis. As a result, our model predicted the next receiver with a high degree of accuracy by weighting positions of the players around the passer.

本研究では実試合データの映像及びトラッキングデータから，パスが行われているシーンを各フレーム毎に分析し，ロジスティック回帰によりモデルを構築した．さらに本モデルをRobocup2Dシミュレーションエージェントに実装することで，パス行動の予測を行った．