작업 예들로부터 관계적 행동 모델의 학습
박찬영, 김인철 · 제어로봇시스템학회 합동학술대회 논문집 · 2010
In order to behave robustly in dynamic environments, robots should be able to learn compact models of their actions. In this paper, we present an algorithm for learning STRIPS-like relational action models from task examples. The models can represent the conditional effects of actions in addition to the primary conditions and effects. Because they are relational, they can generalize over the identities of the objects in the environment. Once action models have been learned, automated planning approaches can be directly applicable to generate plans to achieve the given tasks.
Cited by 0 works.