19 – 21 November 2024// Bremen, Germany

SPEAKER INTERVIEW

spacer

Dr Aubrey Dunne, Co-Founder & CTO, Ubotica Technologies 

 

Please could you tell us more about your role as CTO at Ubotica?


We founded Ubotica to develop solutions to real world problems in the low-power compute vision domain. Visual data is extremely rich in content, making it a valuable source of information for extracting insights and for automatically controlling processes. We saw an opportunity to develop solutions around power-efficient state-of-the-art embedded Vision Processors, and to address the vision challenge using artificial intelligence (AI) directly in the embedded solution. Very quickly the European Space Agency and others saw value in using our solutions to deliver Space AI.

My current role as Chief Technology Officer involves researching, developing and evaluating solutions to deliver on the opportunities that on-satellite AI can realise. This involves exploring new technologies for integration into our product line, evaluating new approaches to on-board Earth observation (EO) data processing, selecting EO processing bottlenecks and challenges that can be addressed with on-board processing, and leveraging our technology to solve the latest challenges in the EO and Space AI domains. I guide our wider team in identifying enhancements and additional features for existing products and concepts for new products and manage the R&D operations within the company. We focus on real-world problems and solutions, and this inherently requires Interfacing with partners, customers, and competitors to keep abreast of advancements in the general EO and satellite domains, where Ubotica’s technologies can help to realise additional return on investment in these Space assets.


The Earth observation applications are numerous for terrestrial end-users – how have you seen the application of satellite imagery grow over the last couple of years?


Our first satellite payload went into orbit in 2020, and in doing so we were the first to enable the demonstration of hardware acceleration AI technology on an EO CubeSat. We have many partners in the Space industry, working with the European Space Agency on a number of projects, with NASA JPL, and with other private companies that develop applications for space and that build satellites. Over the next 18 months our solutions will fly on a further five satellites, and we have a joint mission with Open Cosmos to launch our own satellite - CogniSat-6. The CogniSat-6 mission will demonstrate a number of unique satellite imagery applications that are empowered by on-board AI processing. These include using our latest Space AI solution to control autonomous image scheduling, and to generate real-time insights from EO data that will be delivered direct to users with minimal latencies.

Ubotica’s technology enables customers to deploy intelligent data processing solutions easily and reliably on-orbit. We provide technology that accelerates the application of computer vision and AI to satellite imagery, enabling the extraction of valuable information from raw data directly on-board the spacecraft, in a power efficient, robust, and easy to deploy fashion.

We develop applications in-house and empower our customers and partners to develop and deploy their own applications efficiently and effectively on satellites. Already applications from cloud detection to ship detection to road detection are using our AI technology for satellites. The availability of compact multispectral and hyperspectral sensors, with ever increasing ground sample distances, can be coupled with on-board processing to enable a new class of applications – those delivering time-sensitive insights for end-users and autonomous control alike.

 

One of the main challenges with satellite data gathered, is the amount of it. How is this challenge being tackled?  

Ubotica’s AI technology helps to deliver insights to customers faster, using less bandwidth, and providing them with exactly the insights that they need (rather than just raw data). By processing raw data on the satellite to extract information and downlinking the generated insights reduces the latency and bandwidth required by orders of magnitude compared to sending the entire raw image database to Earth for processing. It is important to get the information to customers as fast as possible, especially when monitoring emergencies or natural disasters that require rapid responses. In many EO applications, the end-user requires not raw data (or even data per se) but rather insights that are of the most value for their particular use case. By extracting these insights directly on-board using AI techniques and communicating them rapidly to the end user via always-available downlink paths, there is no longer a need to transmit all the raw data to ground. Leveraging our on-board AI solution enables the dynamic swapping of Neural Networks so that multiple applications can be addressed on a single platform.


Another challenge is the quality of satellite data – which technologies can help customers understand the data better and how can end-users ensure they don’t lose return on investment on money spent on data gathered?

 

In current Earth observation satellite systems, images are captured on satellite and queued for downlink to Earth at the next available ground station pass. Some process or person on the ground then filters through these images, for example discarding all optical EO images that are cloudy. This can be a slow and expensive process.

With Ubotica’s on-board AI technology we can deliver cleaner images with improved information density to ground by filtering out the cloudy images directly on the satellite, so that only good quality cloud-free image data are sent to ground. Our on-board cloud removal solution is coupled with hardware-accelerated lossless image compression to enable satellites to be more efficient at capturing and transmitting valuable – cloud-free – data, thereby increasing the value realised by the satellite asset.

 

If you could travel anywhere into space, without the restrictions of time and resources, where would you go, how will you travel there, and why?


I would travel to the International Space Station via Dragon capsule, so that I could experience life in LEO, and observe the beauty and majesty of the Earth in a single awe-inspiring view. While there I would also check in on the Ubotica AI experimental testbed that is operational on the ISS – work conducted with JPL and for which we recently received a team achievement award from NASA.

spacer
spacer