Mission Profiles and Evidential Reasoning for Estimating Information Relevancy in Multi-Agent Supervisory Control Applications
Abstract
With the increasing use of semi-autonomous systems such as UGVs and UAVs, command and control concepts are proliferating further to the edge. Future network-centric environments such as the Global Information Grid (GIG) hold the potential to bring necessary information to those at the edge; however, providing the information is not enough. The flow of information now risks becoming a flood which can drown individuals as they allocate limited attentional and cognitive resources to filtering out extraneous information and assembling situation awareness from the remaining relevant pieces. One solution to the problem of information overload is to introduce automation that can filter and process information according to the needs of the human user. There has been much interest in this approach in both the civil and military domains. Military approaches with significant resources and a critical need for deterministic computability tend to employ if-then rules constructed by subject matter experts. Civilian information filtering applications, e.g. Internet spam filtering, often use machine learning techniques such as rule-induction or learned Bayesian-networks. Most rule-based systems are based on binary Boolean logic and have difficulties in dealing with trust in the source of the information and the uncertainty inherent in the information. Systems that employ machine-learning techniques do not have explicit goal congruence and may suffer from over-fitting to a particular context as well as a deficiency of trust in the automation on the part of the human operator. This paper introduces a multi-agent architecture that employs an evidential reasoning mechanism and mission profile structures to perform relevancy estimation for information filtering. The incorporation of the mission profile structure provides a means by which operators can easily develop their own profiles or re-use those that have already been developed. Mission profiles provide a means to explicitly connect
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 2010
- Accession Number
- ADA525379
Entities
People
- Brett Walenz
- Nathan Denny