During the first eight months, the consortium completed the first phase of the project, providing state of the arts about cognitive computing technologies and learning from rare events and states.
As a first step, the Co2Team consortium produced a state of the art (SoA) of cognitive computing and its use in different application domains and described the difference with a normal AI. Both use the same machine learning or deep learning technologies and can solve very complex problems in a short period of time, but their approach and objective are very different. In addition to focusing on natural communication, CCT supports human decision making while AI aims to solve problems without human intervention. Thus, the pilot's skills are increased rather than adding more AI-based automation. The advantage of CCT is that it can process a large amount of data, monitor and cross-reference it, to provide the pilot with reliable and understandable information. To learn, these technologies need a large amount of data, but the intelligent teammate must be able to help the pilot in abnormal situations and therefore learn with little data. A deliverable is therefore dedicated to the learning of rare events and intelligibility of Machine Learning.
The integration of an intelligent teammate in a cockpit requires a thorough analysis of the human factors involved. The Co2team consortium worked for eight months on Human Factors (HF) involved in a Human intelligent Machine Team, in parallel with the SoA on cognitive computing. We looked at past and present research and attempts in different fields of application. The knowledge acquired, with SoA in parallel, on CCTs has made it possible to extend research and to move away from the state of the art on human factors related to system automation. We provided a state of the art of the human factors involved in a hybrid team (Human-Cognitive Computing Teammate) and how to form this team that communicates, shares knowledge, information, collaborates and trusts each other to ensure flight safety at the highest level. As we had the objective of flight safety, we reviewed human factors recommendations and obligations and how they could be updated and implemented for SPO with a CCTeammate. In addition we interviewed pilots to make an in-depth analysis of all the tasks and knowledge required by pilots to carry out their duties and get their views on the implementation and interactions with the CCT and their ideas and recommendations for better acceptability.
After submitting those four deliverables, Co2Team consortium completed the second milestone of the project and launched a review of Cognitive technologies for bidirectional Human-System collaboration and the definition of the use case.
Co2Team consortium identified a specific use case and a very complex and risky flight scenario designed with pilots and human factors experts to challenge the HiMT. The scenario includes several flight phases (descent, approach, Go Around, and landing) the most accident prone according to Airbus and Boeing figures. We submitted and discussed the use case scenario with Airbus, Dassault and Thales who found it full of interest, useful and in relation to their expectations.
