Accelerating the design of flight simulators for the aviation industry
A380 flight simulator - © Airbus SAS - Photo P. Masclet
Over the past ten years, the Coprin team has worked with the flight simulation department at Airbus to automate the generation of code on the basis of equations written by physicists. A demonstration program of the Mosela environment was completed in December 2012 and should be brought to market soon.
An interview with Laurent Farenc, an Airbus flight simulation department engineer and the manager of the Mosela project.
What was the aim behind your work with Inria?
Laurent Farenc - © Laurent Farenc
Our department carries out flight simulations in order to, for example, specify and verify the control laws of an aircraft. Unlike simulations for the design of the aircraft's shape or structure, we do not need to simulate physical phenomena, such as aerodynamics, very accurately. We use simplified equations that are not usually found in commercial tools such as Mathlab Simulink. We therefore write our own equations and choose our discretisation algorithms to produce a high-performance real-time simulation.
Approximately ten years ago, engineers still wrote their equations on paper, which was then sent to subcontractors who would turn the equations into usable IT code, such as C or Fortran, to produce models that the simulator could run rapidly. The work undertaken with Yves Papegay and the Coprin team let us gradually give up the paper stage by using an environment that lets us write equations in a high-level language that is adapted to our working methods and transcribe the equations into a low-level language that can be understood by the machine.
What benefits has the Mosela environment brought?
As Mosela automatically interprets the equations written with the editor and translates them into C code, it lets us bypass the programming stage. The development cycles are thereby shortened by a factor of five or six, although, in practice, it is currently far less as rewriting existing equations with this new technology takes time. The biggest benefit resides in the fact that Mosela greatly reduces the number of interpreting errors which occur when paper-based equations are translated into C code. It guarantees that the equations entered by the engineer with the editor and the corresponding code are coherent. It is also easier to check the code as it is accessed using high-level language. Finally, Mosela has let us become far more responsive when we have to assess the impact of a change in a configuration quickly, particularly as not having to resort to subcontractors means more efficient and faster optimisation processes.
Will the work be continued?
We have decided to bring the program to market. The prototype has been approved using simplified examples but between six and 12 months' work remains to obtain an industrial version. The success of this process will depend on the emergence of any competing technologies during this period. Big mechanical software publishers have also analysed the needs of industrial firms and can produce high-performance solutions in a short period of time. Nevertheless, developing a solution with researchers, although necessarily slower, allows ideas that I consider to be innovative and interesting to emerge. Furthermore, so far, we have not seen anything similar to Mosela at the conferences where we have presented the prototype.
Do you plan to work with Inria again in the future?
In the long term, this process could be used to produce codes that allow models to be adjusted more easily on the basis of flight tests or to optimise equations. Another avenue that has been considered would involve making use of interval mathematics, an area in which Inria is an expert. The aim would be to assess the impact of uncertainties on modelling data more easily, such as a 10% variation in a characteristic of a shock absorber or a tyre.
'Airbus was not satisfied with existing solutions and wanted a solution that was closer to its concerns regarding applications.'
Yves Papegay - © Inria - Photo Kaksonen
Yves Papegay, Inria researcher within the Coprin team
'Airbus wanted to accelerate and at least partly automate the production line of computational paradigms which are encoded in flight simulators. We started to work together for the first time in 2001 on the development of an environment which would computerise the writing of models. We put forward a high-level symbolic program based on the calculation program, Mathematica, which was very close to the program used by the engineers for their paper-based work, easy to use and capable of describing a model accurately. Our work resulted in the creation of an industrial editor of physical models which has been used since the summer of 2004 by Airbus to design its aeroplanes. In the second part of the project, undertaken with Laurent Farenc, we fine-tuned this program and created a tool that is capable of interpreting this language and producing the corresponding simulation code as well as a digital assessment engine, which facilitates the development of the model and the business sector documents. The Mosela environment, which is the outcome of this work which involved standardising, understanding and defining the activities of a computer scientist, is a comprehensive working tool, of which a prototype was presented in 2012 and which has since been transferred to Airbus. Inria will undoubtedly be asked to take part in bringing Mosela to market by providing advice on the implementation of a maintenance structure. '