Incorporating goal recognition into human-machine collaboration

Goal recognition techniques provide a novel way to make human-machine collaboration applicable. After describing a goal recognition algorithm, this paper focuses on how to incorporate goal recognition into practical systems. Specifically, several important properties, namely, partially observing, explanation and the ability to handle ambiguity have been exploited. This method doesn't use a plan library as traditional plan recognition methods do. So it also doesn't suffer the problems such as hand coding of a plan library and searching the plan library space of exponential size. In the last of the paper, theoretical and empirical results of the method have been achieved, which show the accuracy, efficiency and scalability of the method.