Distributed Sensing and Processing Adaptive Collaboration Environment (D-SPACE)
Abstract
With the proliferation of physical military and commercial sensors, and increasing use of open source social media data for situation understanding, traditional stove-piped exploitation solutions no longer achieve required accuracy in a timely manner. Alternatively, the new cloud technologies are not appropriate for situation understanding in areas of denial, where computation resources are limited, data not easily moved around, and communication highly constrained and unreliable. D-SPACE is a Distributed Sensing and Processing Adaptive Collaborative Environment designed to perform large-scale data queries in resource-constrained domains, where the data comes from diverse sources, modalities, and contains high degree of interdependencies among the variables. D-SPACE represents such data as large-scale graphs, enabling online graph analytics via inexact graph matching process. D-SPACE distributes graph exploitation among a network of autonomous computational resources, designs the collaboration policy among them, and assures that processing is robust to communication and resource failures. D-SPACE achieves its objectives by highly efficient work balancing among the resources, graph data indexing, message compression and prioritization to reduce communication load, and adaptive roll-back, re-planning, and re-biasing to deal with communication delays and resource failures. When complete, D-SPACE will provide increased autonomy, improved analysis accuracy, and fault-tolerance, while decreasing the time required for data processing.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 2014
- Accession Number
- ADA608436
Entities
People
- Andrés Ortíz
- Georgiy Levchuk
- John-collonna Romano
Organizations
- Aptima (United States)