Objectives: This project addresses fundamental issues that arise in information representation architectures for autonomous reasoning and learning, decentralized planning, and decision-making in multiagent systems. The overall goal of the project is to develop efficient and adaptive strategies to process, represent, exchange, and act upon relevant information from massive data collections, much of which can be irrelevant, imprecise, and contradictory.
Technical Approaches: This project takes an ambitious approach to handle the collection, representation, and organization of information. Minimalism is at the core of the technical approach: this idea concerns the proper identification of the required information needed to achieve a given task with a desired performance level and provable performance guarantees. Minimal representations involve how appropriate models should be selected, how uncertainty should be managed, and how information should be represented, decomposed and communicated. Key role within this framework play set based approaches in order to perform information decomposition and synchronization for distributed filtering, and representation for meta-reasoning and coordination.
Expected Outcome: We expect the proposed research to lead to fundamental principles,
theory, and algorithms that will enable distributed teams of autonomous vehicles to coordinate robustly in highly dynamic environments and under information uncertainty and to achieve a mission within specified success criteria. The applicability of the expected research outcomes is broad, ranging from sensor coverage and surveillance problems, to formation control and synchronization.