Properly estimating the performance of a wireless link remains a fundamental challenge that has yet to be fully solved. In large part, this is due to the very complex nature of wireless links, which are affected by a large number of phenomena. Despite this difficulty, this task arises in several aspects of wireless networks, such as routing, handoff decisions, and dynamic rate adaptation. Because the performance of a wireless link is time-varying, estimates must be periodically updated. This is often achieved using moving averages. Recently, an alternative called hypothesis test estimator (HTE) was proposed targeting reducing the variability of the estimates, which can be beneficial in certain scenarios, such as for routing protocols. In this paper, we provide an in-depth evaluation of HTE’s performance in comparison to the more traditional moving averages. To that end, we provide mathematical models for the variability and accuracy of the estimates output by HTE, exponentially weighted moving average, and an simple moving average. We then validate those models and use them to compare each approach and to understand HTE’s sensibility to each of its parameters—the window size and significance level. Our results support that HTE is indeed able to significantly reduce estimate variability in comparison to the moving averages, but they also point toward possible shortcomings, such as the difficulty in properly choosing suitable parameters.

正确估计无线链路的性能仍然是一个尚未完全解决的基本挑战。这在很大程度上是由于无线链路的非常复杂的性质，受到大量现象的影响。尽管有这个困难，但这项任务出现在无线网络的几个方面，例如路由、切换决策和动态速率自适应。由于无线链路的性能随时间变化，因此必须定期更新估计值。这通常是使用移动平均线来实现的。最近，提出了一种称为假设检验估计器 (HTE) 的替代方案，旨在减少估计的可变性，这在某些情况下可能是有益的，例如路由协议。在本文中，与更传统的移动平均线相比，我们对 HTE 的表现进行了深入评估。为此，我们为 HTE、指数加权移动平均和简单移动平均输出的估计值的可变性和准确性提供了数学模型。然后我们验证这些模型并使用它们来比较每种方法并了解 HTE 对其每个参数（窗口大小和显着性水平）的敏感性。我们的结果支持，与移动平均线相比，HTE 确实能够显着降低估计变异性，但它们也指出了可能存在的缺点，例如难以正确选择合适的参数。和一个简单的移动平均线。然后我们验证这些模型并使用它们来比较每种方法并了解 HTE 对其每个参数（窗口大小和显着性水平）的敏感性。我们的结果支持，与移动平均线相比，HTE 确实能够显着降低估计变异性，但它们也指出了可能存在的缺点，例如难以正确选择合适的参数。和一个简单的移动平均线。然后我们验证这些模型并使用它们来比较每种方法并了解 HTE 对其每个参数（窗口大小和显着性水平）的敏感性。我们的结果支持，与移动平均线相比，HTE 确实能够显着降低估计变异性，但它们也指出了可能存在的缺点，例如难以正确选择合适的参数。