Artificial intelligence the key to autonomy – and safety - in the skies
SkyGrid is creating a new way to fly.
June 25, 2020 in Innovation, Technology
In 2018, Amir Husain became the founding CEO of SkyGrid, a joint venture between Boeing and artificial intelligence company SparkCognition. SkyGrid is building an aerial operating system to support the next generation of autonomous aviation. In the conversation below, Husain peers into the future.
IQ: Fast forward decades, even centuries. When we look up, what do we see?
Husain: Envision a future with millions of concurrent flights — both piloted and unpiloted. To achieve safe route planning and high-traffic density, we must be able to predict a variety of factors that can change quickly, such as weather conditions, vehicle health and maintenance needs. We can’t just default to the old way of doing things by defining large airspace “bubbles” around every aircraft. This manual approach will drastically impact traffic density and system efficiency by reducing the number of aircraft that can share airspace.
As more unpiloted aerial vehicles (UAVs) become operational, artificial intelligence (AI) models will be necessary to help define the optimal route based on the drone’s health and predicted environmental conditions. AI technology will also be critical to safely sense and avoid new obstacles inflight or completely reroute the drone if the new conditions are extreme.
IQ: How are you collaborating with Boeing to make urban air mobility a reality? Husain: SkyGrid is a joint venture between Boeing and SparkCognition with a mission to accelerate urban air mobility and enable a wide variety of commercial drone services, including package delivery, industrial inspections and emergency assistance.
As we move from a world with thousands of piloted aircraft flying concurrently to a world where millions of unpiloted aircraft are operating at the same time, many elements of the underlying infrastructure will need to evolve.
We’re clearly not there yet. How does that kind of complexity in the skies happen safely and efficiently?
First, today’s air traffic control system was designed to cater to the needs of thousands of concurrent flights. As we move to millions of unpiloted flights, the inherent design and heavy human involvement in this system will prevent it from keeping up with flight volume. As such, air traffic control can be augmented by AI to enable large scale UAV traffic.
Second, the system will require the use of artificial intelligence to generate optimal routes that take into account dozens of underlying criteria, from airspace classes to weather and microweather, traffic patterns of terrestrial vehicles on the ground below, environmental considerations, noise and safety considerations as well as the maintenance condition of the UAV in question. These complex decision factors require machine “intuition,” implemented perhaps by techniques such as Deep Hashing, because they cannot always be computed from scratch using deterministic algorithms that may turn out to be computationally expensive.
Third, the infrastructure must also leverage AI to predict maintenance needs. To safely monitor drone health and performance as UAV use grows, we will need to complement manual inspection with AI.
Last, the system must enable advanced cyber protection. Seen a certain way, each autonomous aircraft is essentially a flying computer with a network connection that can be hacked if not secured properly. AI-powered cybersecurity will be the key to detecting malicious activity on the edge and preventing it from executing on a drone. This type of security must function even when the drone cannot contact a host network and must be particularly able to deal with sophisticated, “zero day” or never-before-seen attacks.
By combining Boeing’s aviation expertise with SparkCognition’s advancements in AI, SkyGrid is bringing this system to life to safely integrate drones in the global airspace.
IQ: What might an engineer need to concentrate on to dive into the AI world?
Husain: If I were to emphasize just one thing, it is to focus time and attention on truly understanding the core underlying concepts of computing: programmability, algorithmic efficiency, search, graph algorithms and recursion. Thinking from first principles is essential in developing mastery in a subject. I am not a big fan of neglecting first principles and developing familiarity only with the highest, most abstract layers of the technology stack. Your towers of understanding must be built on a firm and deep foundation.
There’s almost no area of engineering or science that isn’t dependent on or significantly enhanced by the tools of computer science and AI. And to the extent that AI allows you to almost magically build solutions based on observation — on data — it is an incredibly powerful science. When you realize AI will amplify everything and that it does not operate in a silo, then you will continue to find ways it adds value to an enterprise. It’s a field with a large number of known rewards but also an uncountable number of unexpected ones!
For example, SparkCognition recently hosted a competition in the U.K. where Southampton University students were asked to use our automated model building platform, Darwin, to serve a real-world problem of their choice. To our surprise, students used Darwin to invent water purification technology. This is just one example of an unexpected reward for the field and also illustrates that technology will, in fact, truly change everything.
IQ: What drew you to the field of AI?
Husain: I’ve been fascinated with computing since a very young age. I developed a lifelong love affair with programming when I first saw a Commodore 64 in action at the age of 4. I started formal research in AI while I was a teenager and published my first Institute of Electrical and Electronics Engineers (IEEE) research papers while I was still 17. If you think about it, AI is the ultimate use of computers. If you look at early pioneers like Alan
Turing, a lot of their attention was focused on how computers could “think.” Even Charles Babbage’s mechanical computer design from the 1800s was called the Analytical Engine. Once you get deep into computer science, it's hard to not be fascinated by the potential of computers as thinking machines!
IQ: But the idea of a “thinking machine” may sound ominous to some.
Husain: On my journey to understand artificial intelligence, I’ve reflected deeply on how we humans think. In my book, “The Sentient Machine: The Coming Age of Artificial Intelligence,” I discuss many ways human intelligence differs from artificial intelligence. Despite the fears that surround this topic, I ultimately believe AI will complement human thought and elevate the human condition.