Recent applications of plan recognition face several open challenges: (i) matching observations to the plan library is costly, especially with complex multi-featured observations; (ii) computing recognition hypotheses is expensive. We present techniques for addressing these challenges. First, we show a novel application of machine-learning decision-tree to efficiently map multi-featured observa...

Human users dealing with multiple objectives in a complex environment, e.g., military planners or emergency response operators, are subject to a high level of cognitive workload. When this load becomes an overload, it can severely impair the quality of the plans created. To address these issues, intelligent assistant systems have been rigorously studied in both the AI and the intelligent system...

Deep Convolutional Neural Networks (CNNs) have demonstrated great performance on object detection and classification on still images. However, due to pose change, scaling and other complexities, object detectors and classifiers trained on still images cannot achieve comparable performance when directly applied to videos. In this project, we choose the baseline framework to be T-CNN [1], a deep ...

We propose a formal procedure-based agent model for plan recognition as an alternative to the declarative proposition-based agent model. Our formalism, called the Procedural Agent Model (PAM) allows us to describe procedural plans of agents in a concise and abstract way. Plan recognition uses PAM as a library of plan schemas and applies a graph-based pattern-matching algorithm to the observed e...

Previous work has suggested the use of lexicalized grammars for probabilistic plan recognition. Such grammars allow the domain builder to delay commitment to hypothesizing high level goals in order to reduce computational costs. However this delay has limitations. In the case of only partial observation traces, delaying commitment can prevent such algorithms from forming correct conclusions abo...

In this paper we explore the use of case-based plan recognition to predict a player’s actions in a computer game. The game we work with is the classic Space Invaders game and we show that case-based plan recognition can produce good prediction accuracy in real-time, working with a fairly simple game representation. Our evaluation suggests that a personalized plan library will produce better pre...

Plan recognition systems need a measure to assess the quality of their hypotheses about a user’s goals. Such a measure allows the exclusion of extremely “bad” hypotheses at an early stage or the selection of the “best” one whenever this is required, e.g. when a help system is asked to provide support to the user of an application system. To work properly, such a measure has to take into account...

We describe a general framework for encoding rich domain models and sophisticated plan reasoning capabilities. The approach uses graph-based reasoning to address a wide range of tasks that typically arise in dialogue systems. The graphical plan representation is independent of but connected to the underlying representation of action and time. We describe types of plan recognition that are neede...

This paper presents a probabilistic and abductive theory of plan recognition that handles agents that are actively hostile to the inference of their plans. This focus violates a primary assumption of most previous work, namely complete observability of the executed actions.