Aircraft simulation for pilot training in an immersive environment presents some unique challenges. High fidelity aircraft performance and system models are required to replicate normal and abnormal conditions; human pilots-in-the-loop require high computational frequencies to replicate control feel; aircraft communication busses have seen a dramatic increase in the amount of data. These are just some examples that must be met by a real-time hardware and software architecture to prevent noticeable delays, jitter or other artifacts that could negatively impact training.

Traditionally, individual training devices rely on dedicated hardware for their real-time simulation platform. This approach presents significant limitations in terms of flexibility, scalability, and lifecycle costs. Cloud and on-premise data centers have addressed some of these limitations using high density servers, virtualization and containerization. While these technologies are not new to modeling and simulation, this paper discusses how we have adapted and leveraged them successfully on full flight simulators.

Our approach uses commercially available, enterprise-grade hardware and software, typically used for data centers and integrated as the main computational platform. We will share lessons learned in the process, including performance measurements that precisely describe our real-time requirements and how we were able to achieve them by adapting our chosen hardware/software platform. We will also present the results of our experiments in the use of containers and clusters of containers, without compromising real-time performance. The end result is our first simulator with this new architecture that is expected to be qualified by Transport Canada to the Level D standard in the first half of this year.

Keywords
REAL-TIME;SIMULATORS;VIRTUALIZATION

Additional Keywords
CONTAINERIZATION