Predicting operator's cognitive and motion skills from joystick inputs

The skill level of a human operator is crucial in operating a complicated process. In this paper, we pay particular attention to operating a forest harvester. A simple computer game simulates the operation of a harvester as well as collects input data from the player and output data from the simulation model. First, we study the nature of the input and output data and illustrate them using PCA. Then, we proceed to using only input data and train a neural network model from operator inputs to skill level. Results show that the skill can be predicted reasonably well. The model itself is static, but dynamics are captured using specific indicators. Using bare input data simplifies data collection and makes the prediction faster. We do not have to use data that depend on the machine or environment, and the skill level can be predicted soon after the operator grabs the controls. The next phase will be using the skill information for operation support.