Suivi du système sensorimoteur du pilote d'hélicoptère et application à la prédiction de la charge mentale

Human factors are the human contributions to the accomplishment of a task, in terms of interactions with the available technological tools, the environment in which the operator works, the applicable procedures, and with other operators. It is one of the key elements of aviation safety, and comes into play during pilot training as well as in day-to-day tasks, once the operator is in the work environment. This thesis focuses on the sensorimotor system of helicopter pilots. In other words, on the one hand, we studied the behavior of the vestibular system of professional pilots during a psychophysical experiment conducted in a flight simulator; and on the other hand, we established a predictive model of mental workload during realistic flights conducted on the same simulator. In the first experiment, which focused on the sensory system, nine professional helicopter pilots were placed in a Level-D full-flight simulator (the highest level of certification available) and instructed to return the simulator to the neutral position with all lights off after a passive movement beyond their control. The results showed that the pilots were significantly more accurate in roll (i.e., in the frontal plane) than in pitch (i.e., in the sagittal plane). Conversely, we found no difference between left and right roll, nor between anterior and posterior pitch. Finally, there is a bias towards the initial angle: the greater the passive displacement, the greater the pilot error; for the angles used, the relationship found is linear. In the second experiment, the pilots were placed in the same simulator and subjected to two classic realistic scenarios (a reconnaissance mission, and a medical evacuation mission). Both scenarios were designed to induce large variations in mental workload, with some very quiet moments and others very intense. The pilots were asked to subjectively assess their level of mental workload at key moments identified by aviation experts. In parallel, a number of parameters of different origins were recorded (physiological parameters, machine parameters, and human-machine interface parameters). We then used a machine learning model to try to predict the level of mental workload experienced by pilots at any given time. The main result of this study is that the most relevant set of parameters is that of the interface, which reflects the actions of the pilot's motor system. Finally, this work is also an opportunity for us to propose a general framework for thinking about the human-machine interfaces used by operators of complex machines (aircraft, nuclear power plants, etc.). Indeed, we are in the midst of the Fourth Industrial Revolution, driven by artificial intelligence and the Internet of Things; in light of the above-mentioned results, we believe it is relevant to outline two new technologies that could improve aviation safety: on the one hand, the individual longitudinal monitoring of an operator; and on the other hand, his or her intelligent personal assistant. These tools could be used for the initial and continuous training of aircraft pilots.