Quantifying the Accuracy of Hardware-in-the-Loop Simulations

Abstract

In this paper we present a methodology for quantifying the accuracy of hardware-in-the-loop (HWIL) simulators. HWIL simulation refers to the process of interfacing a subset of system hardware to a controller containing a numerical model of the rest of the system. Quantification of prediction accuracy is fundamentally important in all simulations, and presents unique challenges in the HWIL context. Hardware included in HWIL is often uncertain, non-linear, time varying, or generally unpredictable. Therefore, it is impractical to expect rich sources of information about this hardware's behaviour (e.g. probability distributions) as a basis for accuracy quantification. The methodology in this paper is able to produce <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">a</i> <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">priori</i> hard guaranteed bounds on simulation error based on a minimal set of assumptions regarding the behaviour of the hardware under test. The method is applicable to simulators with uncertain, non-linear, time varying hardware, provided the software subsystem, sensors and actuators are linear. A convergence proof is given. The method has been successfully demonstrated on a real experimental HWIL simulator.