Self-Organizing Networks (SONets) with Application to Target Tracking
Abstract
The growing interest in large arrays of deployable sensors is not only the result of recent advances in technology that make cheap expendable sensors readily available, but is also due to the limitations of current large expensive assets in some applications of timely importance such as urban warfare and complex terrain surveillance. Large distributed arrays of deployable configurable sensors cooperating to achieve system-level goals may provide the solution for such problems whether acting as independent networks or as agents gathering localized information to aid large assets. The primary challenge of dynamic allocation of network assets (DANA) is the cost of computation and communication of global optimization and real-time configuration of individual sensors. Scaling of network size generally yields an exponential increase in optimization computation and a prohibitive need for communication bandwidth for scheduling of individual sensors making such approaches of limited real-time use. This paper presents the novel methodology of Self-Organizing Networks (SONets) where small sensors with local decision capabilities and overall system performance knowledge yield an emergent behavior aimed at maximizing system information in a communication-constrained architecture while eliminating (or reducing) the need for sensors to be actively scheduled. Preliminary results demonstrate promising performance in a multi-target/ multi-sensor environment. The SONets methodology is based on sensors making local decisions on which mode to operate in including data collection broadcast, etc. based on perceived value of expected return and thresholding with the capability of adaptively self-organizing Sensors update learning indices (adaptive weights) based on expected return and observation of overall system knowledge. The result is an emergent behavior that may be supervised and altered through general broadcasts from a centralized unit.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 2004
- Accession Number
- ADA432609
Entities
People
- Dana Sinno
Organizations
- Massachusetts Institute of Technology