Aerial refueling consists of transferring fuel from one aircraft to another during flight. Currently, the U.S. military employs two main aerial refueling technologies: the flying boom and the probe-and-drogue system. Both Air-to-Air Refueling (AAR) techniques are complex and require significant training. AAR training can be conducted using flight simulators because they provide several advantages, including a safe training environment and a significant reduction in the cost of training. However, to conduct effective Virtual Aerial Refueling (VAR) training, the simulators must be designed to satisfy several elements of fidelity. Normally, VAR training is conducted using a synthetic environment with the capability of generating constructive entities that perform the role of either aircraft tanker or aircraft receiver. When the simulator is used as the receiver aircraft, a constructive aircraft is simulated to be used as the tanker, or vice-versa. However, the utilization of a constructive aircraft tanker equipped with a flying boom refueling system will be limited if there is a lack of realism of the flying boom movement during the pre-contact and disconnect phases. Consequently, this type of training does not provide the required fidelity for a high level of training transfer. To enhance the realism and consequently maximize the training transfer of VAR training with a flying boom refueling system, this paper proposes using an Artificial Intelligence (AI) approach to emulate the actions and procedures that flying boom operators would perform during AAR missions. To do so, nine KC-135 boom operator experts were interviewed to gather their expert knowledge, and then a rule-based expert system that emulates the action of a boom operator was subsequently derived. This paper will present the algorithm to emulate the action of a boom operator, the results of its implementation to a KC-135 constructive tanker, the validation of the algorithm, and the crucial lessons learned.
Training
An Emulation of a Flying Boom Operator Using a Rule-Based Expert System
22 views