STALL: A Simple Model for Workload Analysis in Early System Development

Abstract

The Saturation of Tactical Aviator Load Limits (STALL) is defined as the intersection of asymptotically high and low load limits. In a closed queuing system consisting of M homogeneous demand generators, it has been shown that response time becomes asymptotically linear as M increases. This provides a quantitative basis for specifying the saturation point if one knows both arrival rate and service rate (the inverse of task duration). Early in system development, one can typically estimate arrival rates based on mission analyses. But task durations cannot be estimated until procedures have been defined, based on system design. At this stage, it is useful to determine the design requirements. Given the imposed load, how fast must servicing be to keep up with demand? Logically, service rates must exceed arrival rates, but the question is: by how much? Two related criteria can apply: the number of backlogged demands, or the system response time. STALL computes statistics for both. Preliminary model validation has been accomplished, using simulation runs to study model robustness to systematic violations of assumptions. Predictive validity depends on being able to demonstrate that the assumptions are valid in a particular application. The simulations demonstrate what can happen when that match is not successfully achieved. These studies demonstrated that the predictions will typically be most robust for an over-saturated system. The model is least sensitive to violations of the servicing assumptions. Furthermore, it is easy to relax the assumption of homogeneous demand generators by developing planned model extensions.

Human Performance Models for Computer-Aided Engineering

“…The characteristics of task selection on the basis of expected utilities and costs also lie at the core of the concurrent performance assumptions made by many of the predictive models of complex task performance (Pew, Baron, Feehrer, and Miller, 1977), such as the human operator simulator (HOS) (Harris, Iavecchia, Ross, and Shaffer, 1987;Strieb, Lane, Glenn, and Wherry, 1981;Wherry, 1976), SAINT' (Laughery, Drews, and Archer, 1986;Wortman, Duket, Seifert, Hann, and Chubb, 1978;), PROCRU (Zacharias, Baron, and Muralidharan,1981), STALL (saturation of tactical aviator load limits; Chubb, Stodolsky, Fleming, and Hassoun, 1987), and those models developed by Siegel and Wolf (1969), , Chu and Rouse (1979), and Tulga and Sheridan (1980). Essentially these models assume that when two (or more) tasks compete for attention (call for completion at the same time), an algorithm assesses the order in which the tasks are to be performed.…”