Goal Recognition Via Linear Programming

Goal Recognition is the task by which an observer aims to discern the goalsthat correspond to plans that comply with the perceived behavior of subjectagents given as a sequence of observations. Research on Goal Recognition asPlanning encompasses reasoning about the model of a planning task, theobservations, and the goals using planning techniques, resulting in veryefficient recognition approaches. In this article, we design novel recognitionapproaches that rely on the Operator-Counting framework, proposing newconstraints, and analyze their constraints' properties both theoretically andempirically. The Operator-Counting framework is a technique that efficientlycomputes heuristic estimates of cost-to-goal using Integer/Linear Programming(IP/LP). In the realm of theory, we prove that the new constraints providelower bounds on the cost of plans that comply with observations. We alsoprovide an extensive empirical evaluation to assess how the new constraintsimprove the quality of the solution, and we found that they are especiallyinformed in deciding which goals are unlikely to be part of the solution. Ournovel recognition approaches have two pivotal advantages: first, they employnew IP/LP constraints for efficiently recognizing goals; second, we show howthe new IP/LP constraints can improve the recognition of goals under bothpartial and noisy observability.