Error performance analysis of sequential distributed detection using level-triggered sampling

In applications of wireless sensor networks (WSNs), sequential distributed detection using level-triggered sampling (LTS) is a powerful scheme, whose advantages include short average decision delay and enabling asynchronous low-rate communication in network. In existing works, asymptotic optimality properties are analyzed only in terms of average decision delay, while in this paper we consider the global error performance of detection under the assumption of symmetric hypotheses test and i.i.d. observations across time and sensors. The LLR information bits from all local sensors are first ranked according to their transmitting time and produce a new sequence at the fusion center (FC), then global error probability are analyzed by method of exhaustion based on the new sequence. With local error probability computed via simulations and some coefficients computed recursively beforehand, the global error probability can be calculated by derived close-form expressions. Numerical results of derived expressions are compared to MonteCarlo simulation results to validate our performance analysis, which shows that numerical results approximately coincide with Monte-Carlo experimental results.