Analytic Prediction of Emergent Dynamics for Autonomous Negotiating Team (ANT) Systems
Abstract
In the work on analytic prediction and dynamics at USU, three distinct perspectives have been brought to bear. The computer science perspective was closely tied to real-time scheduling and planning questions, leading to case-based negotiation. In this work, autonomous negotiating systems are composed of logically separated software agents that control resources that altruistically seek to perform useful work in a cooperative manner. This study examines the negotiation strategy that improves over time by gained experience. A case-based negotiation strategy is presented that allows self-organized scheduling of the tasks. The mathematics perspective examines task completion under a general resource allocation model, where the allocation problem as a nonlinear differential equation, which was used to predict completion ability. This predictive model was then compared with simulation models. The electrical engineering perspective examined the praxeic decision theory approach to multiple agent coordination. Inference, the problem of estimating the goals of other agents in the arena, is discussed in the praxiec context. Another viewpoint toward multiple agents systems is also presented using catastrophe theory. In this analysis it is determined that a phase transition" behavior is to be expected.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 2003
- Accession Number
- ADA419765
Entities
People
- Daniel Watson
- James Powell
- Todd K Moon
Organizations
- Utah State University