A Motivation Filter Scheme for Behavior Sequence Learning in Virtual Environment
Wei Song, Beibei Zhang, Kyungeun Cho, Kyhyun Um · 한국멀티미디어학회 국제학술대회 · 2009
This paper proposes a motivation filter scheme for behavior sequence learning system which does not require predefined probability of the states' transition. When interacting with unknown environments, a virtual agent needs to learn how to generate a behavior sequence to achieve a goal and determine the transition probability based on the current state and the action taken. In a sequence learning process, the sensed states are grouped by a set of proposed motivation filters in order to reduce the learning computation of the large state space. The proposed motivation filters work motivated by the change in the agent's internal variables. We simulate a virtual environment to elucidate the process of the system.