An experimental study on the effect of pattern recognition parameters on the accuracy of wireless-based task time estimation

Task time estimation is a core industrial engineering discipline. However, the process to collect the required data is manually intensive and tedious, thus making it expensive to keep the data current. Radio frequency signals have been used to automate the required data collection in some applications. However, such radio frequency data is subject to systemic and random noise, leading to a reduction in the accuracy of the task time estimation. This research investigates the use of a pattern recognition method, the k-nearest-neighbor algorithm, to improve the accuracy of task time estimation in a simulated assembly work area. The results indicate that the parameters of the kNN algorithm can be experimentally tuned to improve the accuracy and to dramatically reduce the necessary computational time and the costs of performing real-time task time estimation.