AI in the sky: automatic map updates and autonomous flight control with ML


This evening we have two speakers, Sven Briels from Readar and Ruben Polak from Daedalean.

Agenda estimate:
18:00 - doors open
18:30 - pizzas arrive
19:15 - talk 1: updating maps from aerial imagery by Sven Briels
19:45 - break
20:00 - talk 2: autonomous flight control with ML by Ruben Polak
20:30 - post talk networking and drinks
21:30 - end

Sven's talk description:
Basically every map we use is largely handcrafted. Whether it be the map you use while navigating using your smartphone, or the maps the government uses to determine the tax you should pay on your house. These maps are updated frequently by manually identifying changed regions from aerial imagery. This process is error-prone, often subjective and costly. The question is, can we automate this?

At Readar we specialize in datamining on aerial imagery. We have a track record in object detection (for instance solar panels) and object classification (for instance classifying which roofs are made of asbestos-cement plating). Can we use this experience in determining which areas of a map need to be updated? How do we prevent shadow, perspective change and seasonal influences to trigger false positives?

During this talk I will give some insight in how we learn the computer to update a map.

About Sven:
Sven Briels is co-founder of Readar, a company that specializes in data mining on aerial imagery. He has a background in Aerospace Engineering and has over 10 years experience in Remote Sensing.

Ruben's talk description:
After more than a century of aviation, navigation, guidance and control of aircraft is still a rather manual affair. With the advances in battery
technology, computing capacity and powerful machine learning algorithms, the stage is set to work towards introducing autonomous systems into the aviation industry and, in time, change urban transport forever.

At Daedalean we develop certifiable safety-critical software systems for the aviation industry that adopts technology from robotics, computer vision and machine learning. The challenge lies in the certification of these safety-critical systems. Safety is tightly regulated in aviation and rightly so, because it has made flying one of the safest means of transportation. Yet these regulations also present a barrier for the adoption of new technologies.

In this talk, we explore these challenges, specifically those for machine learning, and how we seek to overcome these barriers.

About Ruben:
Ruben Polak works as a Machine Learning engineer at Daedalean AI, where he focuses on the specification and certification of visual perception and localization systems for autonomous air navigation, guidance and control. He has a background in Aviation Engineering and Artificial Intelligence.