Alan Garvey

A survey of research in deliberative real-time artificial intelligence

This paper surveys recent research in deliberative real-time artificial intelligence (AI). Major areas of study have beenanytime algorithms, approximate processing, and large system architectures. We describe several systems in each of these areas, focusing both on progress within the field, and the costs, benefits and interactions among different problem and algorithm complexity limitations used in the surveyed work.

Criteria-directed task scheduling

Abstract Scheduling complex problem solving tasks, where tasks are interrelated and there are multiple different ways to go about achieving a particular task, is an imprecise science and the justification for this lies soundly in the combinatorics of the scheduling problem. Intractable problems require specialized solutions, perhaps even a cadre of different specialized techniques. We have developed a new domain-independent approach to task scheduling called Design-to-Criteria that controls the combinatorics via a satisficing methodology and custom designs, schedules to meet a particular clients goal criteria. In Design-to-Criteria, criteria directed focusing, approximation, and heuristics, in conjunction with soft goal criteria are used to make the scheduling problem tractable. We describe the interesting facets of the Design-to-Criteria approach and give examples of its power at reducing the complexity of the scheduling task while designing custom satisficing schedules.

A REAL-TIME CONTROL ARCHITECTURE FOR AN APPROXIMATE PROCESSING BLACKBOARD SYSTEM

Approximate processing is an approach to real-time AI problem solving in domains in which compromise is possible between the resources required to generate a solution and the quality of that solution. It is a satisficing approach in which the goal is to produce acceptable solutions within the available time and computational resource constraints. Previous work has shown how to integrate approximate processing knowledge sources within the blackboard architecture. However, in order to solve real-time problems with hard deadlines using a blackboard system, we need to have: (1) a predictable blackboard execution loop, (2) a representation of the set of current and future tasks and their estimated durations, and (3) a model of how to modify those tasks when their deadlines are projected to be missed, and how the modifications will affect the task durations and results. This paper describes four components for achieving this goal in an approximate processing blackboard system. A parameterized low-level control loop allows predictable knowledge source execution, multiple execution channels allow dynamic control over the computation involved in each task, a meta-controller allows a representation of the set of current and future tasks and their estimated durations and results, and a real-time blackboard scheduler monitors and modifies tasks during execution so that deadlines are met. An example is given that illustrates how these components work together to construct a satisficing solution to a time-constrained problem in the Distributed Vehicle Monitoring Testbed (DVMT).

Control Heuristics for Scheduling in a Parallel Blackboard System

This paper investigates the effects of parallelism on blackboard system scheduling heuristics. A parallel blackboard system is described that allows multiple knowledge source instantiations (KSIs) to execute in parallel using a shared-memory blackboard approach. New classes of control knowledge are defined that order the agenda by using information about the relationships between the goals of the KSIs. This control knowledge is implemented and tested in the DVMT application on a Sequent multiprocessor using BB1-style control heuristics. The usefulness of the heuristics is examined by comparing the effectiveness of problem-solving with and without the heuristics (as a group and individually). Problem solving with the new control knowledge results in improved system performance.

Design-to-time scheduling and anytime algorithms

Design-to-time real-time scheduling is an approach to solving time-sensitive problems where multiple methods are available for many subproblems. It is an alternative to the anytime algorithm approach, scheduling discrete methods rather than anytime algorithms with the goal of maximizing the value of the scheduled computation. In this paper we briefly introduce the design-to-time approach, describe how design-to-time can be used to schedule anytime algorithms including some experimental results, and examine anytime characteristics of our design-to-time scheduling algorithm.

Writing in an upper-level CS course

This paper discusses the use of writing as a teaching approach for an upper level computer science course. In describing my experiences, I hope to encourage those schools/teachers who are considering incorporating writing into such a course. Many different kinds of writing are compared and contrasted. Attention is paid to practicalities of the writing process and of the demands made of faculty in these writing-based courses.

Representing and Scheduling Satisficing Tasks

A satisficing solution to a problem is one that is “good enough” or satisfactory in a particular situation. Because of the lack of task predictability, and interdependences among tasks it is desirable to use both approximate solutions for tasks and approximate scheduling algorithms for scheduling task execution. Iterative refinement and the use of multiple methods are two approaches that achieve satisficing behavior. This paper examines these approaches including their effects on task monitoring and on sharing intermediate results among tasks. The design-to-time approach to scheduling satisficing tasks is then discussed.

A blackboard system for real-time control of approximate processing

Approximate processing is an approach to real-time AI problem solving in domains in which compromise is possible between the resources required to generate a solution and the quality of that solution. It is a satisficing approach in which the goal is to produce acceptable solutions within the available time and computational resource constraints. This paper describes four components for achieving this goal in an approximate processing blackboard system. A parametrized low-level control loop allows predictable knowledge source execution, multiple execution channels allow dynamic control over the computation involved in each task, a meta-controller allows a representation of the set of current and future tasks and their estimated durations and results, and a real-time blackboard scheduler monitors and modifies tasks during execution so that deadlines are met. An example is given that illustrates how these components work together to construct a satisficing solution to a time-constrained problem in the Distributed Vehicle Monitoring Testbed.<<ETX>>