1. Framework for AI Development and Certification
Development of AI function for aerospace has some specifics and the right framework for AI development may help to accelerate deployment of certifiable AI functionality for your cockpit. In general, following main steps need to be done:
- Define well your functionality and operational space.
- Select the algorithm type which fit the best for target functionality.
- Establish data management and start to collect data. Ensure you have enough data even for corner cases.
- Test your algorithm at various levels iteratively during the development.
- Do not omit human factors aspect – including subjective evaluation of AI functionality.
As the whole process depends on data availability, the generation of data at simulator, collection of data in field and auto-generation of data – all these options are interesting from cost saving point of view. The investment into the infrastructure (i.e. cloud solution for in-field data collection) may pay back soon and it may also support future growth of the application. Next, once the data collection and supporting infrastructure is well working, there are options to automate future function improvement – the functionality may improve automatically and once significantly improved performance is achieved, you may deploy new version of the product (after formal verification of the change). Described development framework which may be applied to various aerospace AI applications is depicted in the figure below.
Development framework for AI based applications.
The application of the framework and iterative improvements of the AI functionality can be demonstrated at following example. Let’s consider the training organization which manage fleet of airplanes and train pilots. The very first AI application may be on-ground instructor assistant – based on recorded flight data, on-ground assistant prepares feedback for a student. The instructor reviews the feedback and store it with correction into the organization data cloud. This function safe instructor time – helps to prepare post-flight briefing with students, it also helps to the organization to collect training records – and last but not least – data for future AI improvements are collected. Next – as side benefit – data owned by the organization may help to improve training efficiency, safe cost, etc.
Now – in the example above, there are following actors:
- Instructors – they benefit from pre-prepared post-briefings.
- Training organizations – they benefit from detailed training data and related data analysis and services.
- Supplier of the solution – the multiple training organizations data (anonymized, secured,…) can be used for further growth of the functionality – the large number of records from various training organizations can support deployment of improved assistant which assists on-board to a pilot during his solo flights.
- Pilot – benefits from more effective training (at first), then have assistant on-board after his training is done who helps him recognize piloting mistakes and correct them prior they may evolve to real problem.
The example is illustrated in the figure below.
Are you interested how your use case of AI Application can be solved? Or are you interested how our framework may help you get your AI function quickly to the market? Then contact us, we can help you to get certified AI Functionality into the cockpit!
1. Framework for AI Development and Certification
Development of AI function for aerospace has some specifics and the right framework for AI development may help to accelerate deployment of certifiable AI functionality for your cockpit. In general, following main steps need to be done:
As the whole process depends on data availability, the generation of data at simulator, collection of data in field and auto-generation of data – all these options are interesting from cost saving point of view. The investment into the infrastructure (i.e. cloud solution for in-field data collection) may pay back soon and it may also support future growth of the application. Next, once the data collection and supporting infrastructure is well working, there are options to automate future function improvement – the functionality may improve automatically and once significantly improved performance is achieved, you may deploy new version of the product (after formal verification of the change). Described development framework which may be applied to various aerospace AI applications is depicted in the figure below.
Development framework for AI based applications.
The application of the framework and iterative improvements of the AI functionality can be demonstrated at following example. Let’s consider the training organization which manage fleet of airplanes and train pilots. The very first AI application may be on-ground instructor assistant – based on recorded flight data, on-ground assistant prepares feedback for a student. The instructor reviews the feedback and store it with correction into the organization data cloud. This function safe instructor time – helps to prepare post-flight briefing with students, it also helps to the organization to collect training records – and last but not least – data for future AI improvements are collected. Next – as side benefit – data owned by the organization may help to improve training efficiency, safe cost, etc.
Now – in the example above, there are following actors:
The example is illustrated in the figure below.
Are you interested how your use case of AI Application can be solved? Or are you interested how our framework may help you get your AI function quickly to the market? Then contact us, we can help you to get certified AI Functionality into the cockpit!