Temporal Planning for Automatic Service Composition
Abstract
This project created, implemented, and evaluated distributed computational mechanisms for automating composition and scheduling of computer network-based services in applications where timing is important. For example, a commander might need intelligence to be gathered, processed, summarized, and visualized, but with timing preferences, such as that the results are available by a deadline but should be as up-to-date as possible. Preferences and constraints on timing impact which service providers are selected (depending on availability), which services they will provide (depending on the timing characteristics for different levels of service provision), and when each should begin and complete delivery of its promised service(s). Decision making is complicated by factors such as uncertainty over how long some services might take, competing service requests for scarce services, and inherent distribution of authority and private knowledge across service providers and requesters. The project's technical accomplishments included novel extensions to temporal reasoning algorithms that allow computational agents to cooperatively uncover and reconcile timing constraints between their schedules quickly and with minimal loss of privacy. In addition, the researchers devised new distributed techniques for allocating tasks to service-providing agents that scale to larger service networks and avoid trying to formulate schedules in impossible (over-subscribed cases). Finally, the project defined a novel abstraction for coordinating services with uncertain durations, and has developed algorithms that have been shown to solve problems orders of magnitude faster than the prior state of the art, and to solve larger problems than has previously been possible.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 28, 2011
- Accession Number
- ADA563674
Entities
People
- Edmund H. Durfee
Organizations
- University of Michigan