In partially observed environments, it can be useful for a human to provide the robot with declarative information that represents probabilistic relational constraints on properties of objects in the world, augmenting the robot's sensory observations. For instance, a robot tasked with a search-and-rescue mission may be informed by the human that two victims are probably in the same room. An important question arises: how should we represent the robot's internal knowledge so that this information is correctly processed and combined with raw sensory information? In this paper, we provide an efficient belief state representation that dynamically selects an appropriate factoring, combining aspects of the belief when they are correlated through information and separating them when they are not. This strategy works in open domains, in which the set of possible objects is not known in advance, and provides significant improvements in inference time over a static factoring, leading to more efficient planning for complex partially observed tasks.

Goal Recognition Design (GRD) is the task of redesigning environments (either physical or virtual) to allow efficient online goal recognition. In this work we formulate the redesign problem as an optimization problem, aiming at early goal recognition. To this end, we use a measure of worst case distinctiveness (wcd), which represents the maximal number of steps an agent may take before his goal is revealed. With the objective ofminimizing wcd, we construct a search space in which each node in the space is a goal recognition model (one of which is the original model given as input) and one can move from one model to another by applying a model modification, chosen from a set of allowed modifications given as input. Our specific contribution in this work includes the specification of a class of modifications for which we can prune the search space using strong stubborn sets. Such positioning allows reducing the computational overhead of design while preserving completeness. We show that the proposed modification class generalizes previous works in goal recognition design and enriches the state-of-the-art with new modifications for which strong stubborn set pruning is safe. We support our approach by an empirical evaluation that reveals the performance gain brought by the proposed pruning strategy in different goal recognition design settings.

Belief change within the framework of fragments of propositional logic is one of the main and recent challenges in the knowledge representation research area. While previous research works focused on belief revision, belief merging, and belief contraction, the problem of belief update within fragments of classical logic has not been addressed so far. In the context of revision, it has been proposed to refine existing operators so that they operate within propositional fragments, and that the result of revision remains in the fragment under consideration. This approach is not restricted to the Horn fragment but also applicable to other propositional fragments like Krom and affine fragments. We generalize this notion of refinement to any belief change operator. We then focus on a specific belief change operation, namely belief update. We investigate the behavior of the refined update operators with respect to satisfaction of the KM postulates and highlight differences between revision and update in this context.

Recent approaches to goal recognition have progressively relaxed the assumptions about the amount and correctness of domain knowledge and available observations, yielding accurate and efficient algorithms. These approaches, however, assume completeness and correctness of the domain theory against which their algorithms match observations: this is too strong for most real-world domains. In this paper, we develop goal recognition techniques that are capable of recognizing goals using \textit{incomplete} (and possibly incorrect) domain theories. We show the efficiency and accuracy of our approaches empirically against a large dataset of goal and plan recognition problems with incomplete domains.

It is well known that the problems of stochastic planning and probabilistic inference are closely related. This paper makes several contributions in this context for factored spaces where the complexity of solutions is challenging. First, we analyze the recently developed SOGBOFA heuristic, which performs stochastic planning by building an explicit computation graph capturing an approximate aggregate simulation of the dynamics. It is shown that the values computed by this algorithm are identical to the approximation provided by Belief Propagation (BP). Second, as a consequence of this observation, we show how ideas on lifted BP can be used to develop a lifted version of SOGBOFA. Unlike implementations of lifted BP, Lifted SOGBOFA has a very simple implementation as a dynamic programming version of the original graph construction. Third, we show that the idea of graph construction for aggregate simulation can be used to solve marginal MAP (MMAP) problems in Bayesian networks, where MAP variables are constrained to be at roots of the network. This yields a novel algorithm for MMAP for this subclass. An experimental evaluation illustrates the advantage of Lifted SOGBOFA for planning.

Online goal recognition is the problem of recognizing the goal of an agent based on an incomplete sequence of observations with as few observations as possible. Recognizing goals with minimal domain knowledge as an agent executes its plan requires efficient algorithms to sift through a large space of hypotheses. We develop an online approach to recognize goals in both continuous and discrete domains using a combination of goal mirroring and a generalized notion of landmarks adapted from the planning literature. Extensive experiments demonstrate the approach is more efficient and substantially more accurate than the state-of-the-art.

Planning is one of the oldest areas of research within artificial intelligence, studying the selection of actions for accomplishing goals. The more recently established areas of plan, activity, and intent recognition instead study an agent's behavior and task(s) given observations of its chosen actions. While these areas have been independently studied and applied to games in the past for both understanding player behavior and developing game characters, the potential for their integration presents even more opportunities via adaptive interaction with the player. In this manuscript, we discuss recent research on the integration of these areas and investigate potential uses for such integrated systems in games.

Goal recognition is the task of inferring an agent's goals given some or all of the agent’s observed actions. Among different ways of problem formulation, goal recognition can be solved as a model-based planning problem using off-the-shell planners. However, obtaining accurate cost or reward models of an agent and incorporating them into the planning model becomes an issue in real applications. Towards this end, we propose an Inverse Reinforcement Learning (IRL)-based opponent behavior modeling method, and apply it in the goal recognition assisted Dynamic Local Network Interdiction (DLNI) problem. We first introduce the overall framework and the DLNI problem domain of our work. After that, an IRL-based human behavior modeling method and Markov Decision Process-based goal recognition are introduced. Experimental results indicate that our learned behavior model has a higher tracking accuracy and yields better interdiction outcomes than other models.

Recent approaches to goal recognition have progressively relaxed the assumptions about the amount and correctness of domain knowledge and available observations, yielding accurate and efficient algorithms. These approaches, however, assume completeness and correctness of the domain theory against which their algorithms match observations: this is too strong for most real-world domains. In this paper, we develop a goal recognition technique capable of recognizing goals using incomplete (and possibly incorrect) domain theories as well as noisy observations. Such recognition needs to cope with a much larger space of plan hypotheses consistent with observations. We show the efficiency and accuracy of our approach empirically against a large dataset of goal recognition problems with incomplete domains.

Belief revision is concerned with incorporating new information into a pre-existing set of beliefs. When the new information comes from another agent, we must first determine if that agent should be trusted. In this paper, we define trust as a pre-processing step before revision. We emphasize that trust in an agent is often restricted to a particular domain of expertise. We demonstrate that this form of trust can be captured by associating a state partition with each agent, then relativizing all reports to this partition before revising. We position the resulting family of trust-sensitive revision operators within the class of selective revision operators of Ferme and Hansson, and we prove a representation result that characterizes the class of trust-sensitive revision operators in terms of a set of postulates. We also show that trust-sensitive revision is manipulable, in the sense that agents can sometimes have incentive to pass on misleading information.