Cross-Layer Optimization and Information Assurance in Decentralized Detection over Wireless Sensor Networks

Wireless sensors generate a vast amount of data, which must be processed rapidly for the purpose of decision making. Stringent delay constraints typical of sensor networks suggest that a classical capacity/throughput analysis of the communication infrastructure associated with a sensor network may not offer an accurate assessment of overall performance. Rather, the allocation of system resources should be carefully crafted to optimize overall system performance. This work explores the potential benefits of a cross-layer design in the context of detection over sensor networks. The allocation of system resources is considered and overall performance for sensor systems subject to hard delay constraints is analyzed. A continuous-time Markov model is introduced to capture the unreliable nature of wireless connections. Based on this model, a sensor equivalent single-server queue is introduced and an overall performance metric is proposed. Numerical results suggest that substantial gains are possible through cross-layer optimization. A queueing-aware architecture outperforms both the instantaneous transmission model common to the decentralized detection literature and the throughput maximization approach often favored by the sensor network community. The performance evaluation methods presented in this paper provides an elegant framework to quantify the amount of physical resources necessary to achieve a desired performance level at the fusion center. More importantly, this research underscores the importance of a global system-theoretic view in the design of real-time applications over sensor networks.