Distance Based Triggering and Dynamic Sampling Rate Estimation for Fuzzy Systems in Communication Networks

Abstract

To reduce computational cost in fuzzy systems in communication networks, distance based triggering and sampling rate adaptation probabilities are proposed based on the concept of probability via expectation. The triggering probability, which is calculated by using the square of distance between subsequent input vectors, governs the rate at which the fuzzy system is triggered. The dynamic sampling rate probability, which governs the adaptation of the sampling rate, is computed by using the exponentially weighted moving average (EWMA) of the triggering probability. A stopping criterion, based on convergence tests, is also proposed to ensure that the mechanism switches off when the sampling period has converged. The triggering mechanism reduces the number of computations in the Fuzzy Logic Congestion Detection (FLCD) in wireless Local Area Networks (WLANs) by more than 45%. Performance, in terms of packet loss rate, delay, jitter, and throughput, however, remains virtually the same. On the other hand, the dynamic sampling rate mechanism leads to more than 150% improvement in sampling rate and more than 70% reduction in fuzzy computations while performance in the other key metrics remains virtually the same. As part of future work, the proposed mechanism will be tested in fuzzy systems in wireless sensor/actuator networks.