A Sensor Management Model Using Simulation-Based Approximate Dynamic Programming
Abstract
This paper presents a new approach to sensor management of distributed sensor networks (DSNs). Given the current proliferation of remote sensors and their inherent resource constraints, DSN managers face a growing problem of managing the tradeoff between DSN performance and resource consumption. Our model, the Sensor Network Optimal OPerations Simulator, or SNOOPS, addresses this tradeoff by identifying a DSN control strategy that reaches an acceptably certain representation of the search region while minimizing operating costs. The core of the SNOOPS model is an approximate dynamic programming (ADP) process that uses simulation-based policy Iteration to identify an efficient DSN control strategy. Results indicate that the SNOOPS-recommended DSN control strategy improves the efficiency of DSN operations by up to 47 percent over the Base Policy of activating all sensors. In addition to determining efficient DSN control strategies, our model also provides a research base to (1) investigate the fusion of observations from disparate sensors, (2) demonstrate the use of non-imaging sensors to provide adequate situational awareness where precision emplacement of more-capable sensors is not possible, and (3) develop operational concepts to integrate DSN operations with user needs.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 31, 2003
- Accession Number
- ADA418276
Entities
People
- Sandor Z. Der
- Stephen D. Patek
- William S. Bland
Organizations
- United States Army Research Laboratory