An Application of Quantitative Methods for Motor Ability Level Classification, Performance Prediction and Training Protocol Selection

Over the past four decades, the application of automation technology in industrial operations has expanded dramatically, primarily due to the development of advanced computer technology. However, manual work is still required in manufacturing, especially in assembly processes. To achieve high production rates, assembly operators are required to develop specialized motor skills. Numerous research studies have been conducted to define training protocols for novice operators to support individual motor skill development to maximum achievable levels. The effectiveness of these training approaches has also been validated in prior laboratory research. However, in real-world applications, such as manual assembly on a production line, it is more commonly expected that operators are trained to a uniform performance level (i.e., "normal" or 100% performance) rather than each operator achieving his/her highest skill level in order to prevent bottlenecks or work-in-process inventory accumulation in production operations. Therefore, there is a need to identify appropriate methods to classify novice operators based on their initial motor performance and to assign them to suitable training protocols facilitating different levels of skill development towards 100% performance.;The objective of this research was to develop an algorithm for classification of operator motor ability on the basis of baseline performance in a simple motor-control test. The research was also to specify appropriate virtual reality (VR)-based training methods, on the basis of skill classification, such that operators might achieve desired levels of motor performance in a real-world design assembly task using a computer-mediated environment.;The study followed a two-phase experimental approach. In the first phase, a batch of 21 right-handed participants was recruited for the computer-based motor test performance along with completion of a standardized psychomotor test (in physical form). The results of the standardized test (Purdue Pegboard) were used as a "gold standard" to validate the computer-based motor test for automatically assessing participant motor ability. A statistics-based model was then developed to classify participant motor ability level using as inputs a set of features based on kinematic parameters generated through the computerized motor test. The finalized model achieved a classification accuracy of ~98% using a 75/25 cross validation approach.;In the second phase, a different batch of 36 right-handed participants was recruited to perform the same computer-based motor test. Performance results were input to the classification model to predict motor ability level. Based on the classification results ("high", "medium" and "low" ability level), each participant received one-of-three pre-designed haptic- VR training protocols, including consistent haptic guidance in motor movements, resistive haptic forces counter to movements, and random haptic disturbances. Prior research has demonstrated varying effects of these methods on training performance with some superiority of haptic disturbances. Results revealed participants identified as "medium" or "low" were able to achieve levels of motor performance comparable to "high" participants through 1-hour training with a VR-based 2D pattern assembly task. Results further validated the accuracy of the motor ability classification algorithm.;The findings of this study verified a quantitative motor skill classification algorithm and an approach to designing and assigning proper training protocols for novice operators to achieve comparable levels of motor performance. This methodology could be applied to operator training for real-world manual assembly operations and to ensure that a group of workers achieve uniform performance levels.