Reconstructing Simulator Control Inputs: A Machine Learning Approach

Flight simulators are valuable tools for human factors research. However, some simulation platforms fail to record all of the information relevant to the researcher. While the data produced by most simulators includes details about the position and state of the simulated aircraft, some platforms do not record pilots’ control input. Missing control input data make it difficult to evaluate response times, a key behavioral measure in human factors research. Here we describe a technique that uses machine learning to reconstruct aircraft maneuvers using aircraft control surface information, which is typically available in simulator output files. This allows researchers to more accurately estimate the moment at which a pilot initiated a maneuver.