Abstract
Recent research has found situations where the identification of agent goals could be purposefully controlled, either by changing the underlying environment to make it easier, or exploiting it during agent planning to delay the opponent’s goal recognition. The paper tries to answer the following questions: what kinds of actions contain less information and more uncertainty about the agent’s real goal, and how to describe this uncertainty; what is the best way to control the process of goal identification. Our contribution is the introduction of a new measure we call relative goal uncertainty (rgu) with which we assess the goal-related information that each action contains. The rgu is a relative value associated with each action and represents the goal uncertainty quantified by information entropy after the action is taken compared to other executable ones in each state. After that, we show how goal vagueness could be controlled either for one side or for both confronting sides, and formulate this goal identification control problem as a mixed-integer programming problem. Empirical evaluation shows the effectiveness of the proposed solution in controlling goal identification process.
Highlights
Goal recognition—the ability to recognize the plans and goals of other agents—enables humans, AI agents, or command and control systems to reason about what the others are doing, why they are doing it, and what they will do [1]
The following sections present the methods we developed for calculating rgu and controlling the goal uncertainty of a given goal identification control problem
We present a new perspective of goal identification control using off-the-shelf probabilistic goal recognizer, and introduce the relative goal uncertainty value for actions in a goal recognition task
Summary
Goal recognition—the ability to recognize the plans and goals of other agents—enables humans, AI agents, or command and control systems to reason about what the others are doing, why they are doing it, and what they will do [1]. Goal recognition system has worked well in many applications such as human–robot interaction [2], intelligent tutoring [3], system-intrusion detection [4] and security applications [5] Though this technique has been successfully applied to many application domains, new problems arise when goal recognition encounters a goal uncertainty discovered in certain environmental settings. One focuses on constructing a suitable library of plans or policies [16,17], while the other one takes the domain theory as an input and use planning algorithms to generate problem solutions [21,23] Before reviewing these two formulations, we would first give the definition of probabilistic goal recognition (or plan recognition) as follows: Definition 1.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
