Traffic Engineering of Combined Mobile Telephone and Dispatch Systems

If a server group carrying short-holding-time queued traffic at high blocking probability is combined with another server group carrying normal blocked-calls-cleared traffic at low blocking probability, the resulting mixed group will in many cases operate at an intermediate blocking probability. Thus the mixed group provides better service to the queued traffic, while degrading the service of the cleared traffic. If the random arrival characteristics of the short-holding-time traffic are suitably modified by delaying calls an appropriate time before service, then the grades of service for both types of traffic may be engineered independently. An example of a possible use of this technique of induced delay, the combining of dispatch and mobile-telephone traffic on the same set of radio channels in the High-Capacity Mobile Telecommunications System, is discussed.     

Correction to "Statistical Estimation of Mean Signal Strength in a Rayleigh-Fading Environment"

An error in calculating the first and second moments of the "exponential-Rayleigh" distribution appears in a recent paper; this correspondence presents the correct calculations.     

Statistical Estimation of Mean Signal Strength in a Rayleigh-Fading Environment

A commonly used model for signal fading in many types of communication channels is that the amplitude of the received signal at a given time is a Rayleigh-distributed random variable. In this paper we show how classical statistical techniques may be applied to the problem of estimating the Rayleigh distribution parameter (i.e., the mean), given samples from the distribution. In particular, we first consider the problem of estimating the population mean, given a sequence of independent samples. We derive an unbiased maximum-likelihood estimator. We show that this estimator is unique, and since it is based on a sufficient statistic, it is therefore "best" in the Blackwell-Rao sense of minimizing expected loss. Using this estimator, we then develop confidence intervals whose length can be used as a guide in selecting the required sample size. We then consider the same estimation problem when the signal samples are obtained from the output of a logarithmic receiver. We derive an interval estimator which does not require taking the antilogs of the log samples, and which is not appreciably worse than the "best" estimator.