Development of a System for the Training Assessment and Mental Workload Evaluation

Several studies have demonstrated that the main cause of accidents are due to Human Factor (HF) failures. Humans are the least and last controllable factor in the activity workflows, and the availability of tools able to provide objective information about the user’s cognitive state should be very helpful in maintain proper levels of safety. To overcome these issues, the objectives of the PhD covered three topics. The first phase was focused on the study of machine-learning techniques to evaluate the user’s mental workload during the execution of a task. In particular, the methodology was developed to address two important limitations: i) over-time reliability (no recalibration of the algorithm); ii) automatic brain features selection to avoid both the underfitting and overfitting problems. The second phase was dedicated to the study of the training assessment. In fact, the standard training evaluation methods do not provide any objective information about the amount of brain activation\resources required by the user, neither during the execution of the task, nor across the training sessions. Therefore, the aim of this phase was to define a neurophysiological methodology able to address such limitation. The third phase of the PhD consisted in overcoming the lack of neurophysiological studies regarding the evaluation of the cognitive control behaviour under which the user performs a task. The model introduced by Rasmussen was selected to seek neurometrics to characterize the skill, rule and knowledge behaviours by means of the user’s brain activity. The experiments were initially ran in controlled environments, whilst the final tests were carried out in realistic environments. The results demonstrated the validity of the developed algorithm and methodologies (2 patents pending) in solving the issues quoted initially. In addition, such results brought to the submission of a H2020-SMEINST project, for the realization of a device based on such results.