COORDINATING PLANNING ACTIVITY AND INFORMATION
FLOW IN A DISTRIBUTED PLANNING SYSTEM

Marie desJardins
Senior Computer Scientist
SRI International

2:00pm Friday October 8, 1999
Lecture Hall V, ECS

In this talk, I will present our research on Distributed SIPE-2 (DSIPE), a distributed version of SRI's SIPE-2 hierarchical task network (HTN) planning system. DSIPE provides decision support to human planners in a collaborative planning environment, in which human planners are geographically dispersed. In this environment, each user is responsible for generating a subplan within the overall planning process. The goal of DSIPE is to provide intelligent support within this environment by identifying dependencies among the subplans early in the planning process, in order to avoid potential conflicts when the subplans are merged. DSIPE's support for distributed planning centers around a set of methods for automatically identifying and sharing potentially relevant information among distributed planning cells. These methods rely on DSIPE's subplan representation, which provides constraint-based, consistent local views of the global plan, which provide each planner a view of how other planners' subplans relate to their local planning decisions. The overall plan is produced by a subplan merging step that leverages the shared structure of the subplans to generate a complete, final plan. In the talk, I will give an overview of the system, then describe DSIPE's distributed planning process in more detail, using an example from the military maritime campaign planning domain. The paper on which this talk is based is scheduled to be published in the Winter 2000 issue of AI Magazine. A prepublication draft is available at http://www.ai.sri.com/~marie/papers/dsipe-aimag.ps.Z .

Dr. Marie desJardins is a Senior Computer Scientist at SRI International in Menlo Park, CA. She was awarded a Ph.D. in artificial intelligence from the Department of Computer Science at the University of California at Berkeley in 1992. Her dissertation presented a model for autonomous machine learning in probabilistic domains. She has technical interests in planning, machine learning, knowledge representation, probability theory and decision theory. Since joining SRI in 1991, she has led or participated in numerous applied research projects, primarily in the areas of planning and machine learning. Planning projects have included distributed planning techniques, hybrid generative/case-based planning methods, a planning architecture to support distributed planning and scheduling agents, and task-based information distribution in collaborative environments. Machine learning projects have included mixed-initiative methods for knowledge acquisition, partially automating the development of planning knowledge for AI planning systems, prediction of biochemical function of enzymes, and modeling student learning behavior in intelligent tutoring systems. Dr. desJardins has presented papers and chaired several workshops in the areas of planning and machine learning; edited special issues of the Machine Learning Journal and AI Magazine; has reviewed papers for numerous journals, conferences, and workshops; has served on program committees including ICMAS-2000, AAAI-98, AAAI-96, ICML-94, and AAAI-93; and is Chair of the AAAI-2000 Workshop Program.