UK and Canada AI Innovation Challenge: How can machine learning improve aerodynamic testing for airplanes?

Do you have innovative ideas about the power of AI and machine learning that could be applied to real-world challenges?

Together with global aerospace and transport company, Bombardier, the Foreign Commonwealth Office Science and Innovation Network and CARIC, Digital Catapult is looking for established startups, scaleups and researchers to develop and pitch innovative AI solutions that could help aircraft reach new levels of performance in extreme weather conditions.

Six finalists (three from UK and three from Canada) will be selected to pitch in Montreal to a panel of industry experts. The finalists will have opportunities to network and develop business relationships that could lead to potential future collaborations. The winner will engage in follow up conversations with interested parties and will be awarded a trip to either Canada or the UK to explore the country’s AI ecosystem and create new relationships that will add business benefit.

Why has the the UK and Canada AI Innovation Challenge been created?

Ice formation on aerodynamic surfaces is one that affects the whole aerospace industry. The challenge of developing ice protection systems for aircraft involves detailed physical modelling and extensive testing in experimental facilities or in flight. During the testing process, video cameras provide information on the surface water behaviour as it runs over and around protected surfaces, sometimes creating ice accumulations. Detailed surface water dynamics are currently difficult to observe. Improving the characterisation of such icing test data has the potential to enhance physical understanding for engineers and reduce development costs.

The challenge

Main Challenge (Part 1): Develop solutions that process videos and segment images to help engineers analyse and categorise surface conditions

During icing tests, icing specialists look at areas where water droplets may collect and freeze as an airplane flies through clouds. Unfortunately, water and ice are often hard to see. Poor lighting, glare, or simply the inevitable presence of the icing cloud can make it difficult to clearly see and categorise what’s happening on the surface.

We’re looking for startups to propose solutions for video processing, to remove the inevitable imperfections of the experimental footage, whilst clearly revealing the surface conditions. Solutions to this challenge should look to use image segmentation to identify different types of surface water: (1) impingement or water film areas, (2) beads or rivulets, (3) glaze ice, (4) rime ice.

In support of this specific challenge multiple fixed angle video sources will be provided by Bombardier, along with labeled images provided by its icing specialists. The solutions to this challenge will help to improve the understanding of icing dynamics and contribute to the evolution of ice protection simulation models.

Beyond the main challenge (Part 1), we are also challenging you to put your expertise to the test and encourage you to stretch AI to its limit to see if you can find a solution to Part 2.

Bonus (Part 2): Create forecast imagery that extends the duration of a given icing test and predict imagery for an icing test for which no experimental data is available.

The bonus challenge is to create a video prediction for a test condition not part of the initial dataset. Solutions to this challenge will enable Bombardier to: (1) create forecast imagery to extend the duration of a given icing test, and (2) create predictive imagery for icing conditions that are not included in the available dataset.

Access to the sample data will be granted if your application is selected and an NDA has been signed.

Register interest below for more information on both challenges, including real-time needs for the bonus challenge.

Focus

Please note, the main focus for applicants should be on creating ideas and developing an approach for how you would solve the challenge. Focus should be directed towards the main challenge. The judging panel will see it as a bonus if innovators present processed results and/or pitch for the bonus challenge.

Who we’re looking for

Startups, scaleups, research institutions or academic bodies. Specifically applicants that have expertise in fields such as computer vision, generative models, predictive modelling, and physical simulations.

Applicants should ensure at least one person from their company is available to fly to Montreal in December and for the key dates outlined in the knowledge pack.

Why apply?

The UK and Canada AI Innovation Challenge is a high profile competition for pioneering AI innovators. Reasons to apply include:

• Raise market awareness of your organisation on a global scale.
• One return flight to Montreal with accommodation for one representative per company.
• The opportunity to pitch to key stakeholders within Bombardier and a judging panel of industry experts including the AI council and Real Ventures.
• Meet with key internal Bombardier stakeholders to discuss the proposed solution in more detail.
• The UK winner of the Challenge will be awarded a trip to Canada to explore the country’s AI ecosystem and create new relationships that will add business benefit. The trip to Canada is expected to take place before the 31 March 2019.
• How to apply and what’s the selection process?

Apply below to register your interest. You will be sent an email with a link to the Open Call application form with a Knowledge Pack outlining more information on the challenges outlined above and the selection process. You must apply by 5th October 2018 to be considered.