ion-Partitioned Evaluator (APE) architecture which has been tested in a simulated, single-agent, indoor navigation domain [SH90]. The APE architecture is composed of a number of concurrent, hierarchically abstract action control layers, each representing and reasoning about some particular aspect of the agent's task domain. Implemented as a parallel blackboard-based planner, the five layers --- sensor/motor, spatial, temporal, causal, and conventional (general knowledge) --- effectively partition the agent's data processing duties along a number of dimensions including temporal granularity, information/resource use, and functional abstraction. Perceptual information flows strictly from the agent sensors (connected to the sensor /motor level) toward the higher levels, while command or goal-achievement information flows strictly downward towards the agent's effectors (also connected to the sensor/motor level). Besides mechanisms for communicating with other layers, each layer in the AP...

Closed-world inference is the process of determining that a logical sentence is false based on its absence from a knowledge base, or the inability to derive it. This process is essential for planning with incomplete information. We describe a novel method for closed-world inference and update over the first-order theories of action used by planning algorithms such as NONLIN, TWEAK, and UCPOP. We show the method to be sound and efficient, but incomplete. In our experiments, closed-world inference consistently averaged about 2 milliseconds, while updates averaged approximately 1.2 milliseconds. We incorporated the method into the XII planner, which supports our Internet Softbot (software robot). The method cut the number of actions executed by the Softbot bya factor of one hundred, and resulted in a corresponding speedup to XII. 1

Complex, real-world domains require a rethinking of traditional approaches to AI planning. Planning and executing the resulting plans in a dynamic environment requires a continual approachinwhich planning and execution are interleaved, there may be uncertaintyin the current and projected world state, and replanning may be required when the situation changes or planned actions fail. Furthermore, complex planning and execution problems may require multiple computational agents and human planners to collaborate on a solution. In this article, we describe a new paradigm for planning in complex, dynamic environments, whichweterm distributed,continual planning (DCP). We argue that developing DCP systems will be necessary in order for planning applications to be successful in these environments. We give a historical overview of research leading up to the current state of the art in DCP, and describe research in distributed and continual planning. The increasing emphasis on r...

THE aim of this thesis is to investigate logical formalisms for describing, reasoning about, specifying, and perhaps ultimately verifying the properties of systems composed of multiple intelligent computational agents. There are two obvious resources available for this task. The first is the (largely AI) tradition of reasoning about the intentional notions (belief, desire, etc.). The second is the (mainstream computer science) tradition of temporal logics for reasoning about reactive systems. Unfortunately, neither resource is ideally suited to the task: most intentional logics have little to say on the subject of agent architecture, and tend to assume that agents are perfect reasoners, whereas models of concurrent systems from mainstream computer science typically deal with the execution of individual program instructions. This thesis proposes a solution which draws upon both resources. It defines a model of agents and multi-agent systems, and then defines two execution models, which ...

Abstract not found

With the vast number of information resources available today, a critical problem is how to locatc retrieve and process information. It is impracticaJ to build a single unified system that combines all of these information resources. A more modular approach is to build specialized information agents where each agent provides access to a subset of these resources and can serve as an information source to other agents. In this paper wc present the architecture of the individual information agents and describe how this architecture supports a network of cooperating information agents. Wc describe how these information agents represent their knowlcdgc communicate with other agcnts dynamicaJly construct information rctricvaJ plans and learn about other agents to improve their accuracy and efficiency. Wc have aJrcady built a smaJl network of agents that have these capabilities and provide access to information for logistics planning.

Virtual environments are a promising milieu for education and training, because they allow students to practice their skills in 3D simulations of work settings. Autonomous agents can improve the effectiveness of such environments by assisting and collaborating with students as appropriate. This paper describes an autonomous pedagogical agent called Steve that can support the training of procedural skills such as operating or repairing complex equipment. Steve's architecture permits him to sense and manipulate dynamic virtual worlds. The architecture also enables Steve to assume alternative realizations, either as a full, articulated, human figure or as abstract pointers and disembodied hands. Steve employs a combination of intelligent capabilities in his interactions with students and the environment: plan revision and execution, explanation, and student monitoring. These capabilities are employed in multiple ways in order to support alternative pedagogical styles. Steve's knowledge re...

In order for a virtual environment to be effective as a training tool, it is not enough to concentrate on the fidelity of the renderings and the accuracy of the simulated behaviors. The environment should help trainees develop an understanding of the task being trained, and should provide guidance and assistance as needed. This paper describes a system for developing virtual environments in which pedagogical capabilities are incorporated into autonomous agents that interact with trainees. These pedagogical agents can monitor trainees' progress and provide guidance and assistance. The agents interact with simulations of objects in the environment, and with trainees. The paper describes the architectural features of the environment and of the agents that permit the agents to meet instructional objectives within the virtual environment. It also discusses how agent-based instruction is combined with other methods of delivering instruction. 1. Introduction Training is an important applicat...

Classical planners have traditionally made the closed world assumption --- facts absent from the planner's world model are false. Incompleteinformation planners make the open world assumption --- the truth value of a fact absent from the planner's model is unknown, and must be sensed. The open world assumption leads to two difficulties: (1) How can the planner determine the scope of a universally quantified goal? (2) When is a sensory action redundant, yielding information already known to the planner? This paper describes the fully-implemented xii planner, which solves both problems by representing and reasoning about local closed world information (LCW). We report on experiments utilizing our UNIX softbot (software robot) which demonstrate that LCW can substantially improve the softbot 's performance by eliminating redundant information gathering. Introduction Classical planners (e.g., (Chapman 1987)) presuppose correct and complete information about the world. Although recent wo...

Current specialized planners for query processing are designed to work in local, reliable, and predictable environments. However, a number of problems arise in gathering information from large networks of distributed information. In this environment, the same information may reside in multiple places, actions can be executed in parallel to exploit distributed resources, new goals come into the system during execution, actions may fail due to problems with remote databases or networks, and sensing may need to be interleaved with planning in order to formulate efficient queries. Wehavedeveloped a planner called Sage that addresses the issues that arise in this environment. This system integrates previous work on planning, execution, replanning, and sensing and extends this work to support simultaneous and interleaved planning and execution. Sage has been applied to the problem of information gathering to provide a flexible and efficient system for integrating heterog...