July 21, 2019 - 83 views|
To mark the 50-year anniversary of the lunar landing, we used Evolutionary AI to virtually land on the moon, with optimal accuracy and intricate control.
It's been 50 years since the first person stepped on the moon, proving there are no limits to how far we can go when we apply our passion, bravery and innovation to go beyond what seems possible. Landing a person on the moon was only theoretical until Neil Armstrong, Buzz Aldrin and Michael Collins eased the lunar lander onto the surface of the moon, executing a complex series of maneuvers to land at the right location, with just enough fuel and at the right speed. They were supported by rooms filled with some of the greatest minds in physics, engineering and computing, and they were watched by the entire world. So many things could have gone wrong. This amazing accomplishment had to succeed – for the lives of the astronauts and for the future of the U.S. space program. And this success relied heavily on software that at the time was state of the art. The code to control the lunar lander, developed by Don Eyles, was simulated and tested as much as possible, but in the end, it had to work in the real world, on the moon, nearly 240,000 miles from any help.
From Five Decades Ago to Today That moment was not just an achievement in science; it was an inspiration for a generation of future engineers and programmers who grew up in a new world where anything was possible. These children, with posters of Neil Armstrong on their walls and toy boxes filled with rockets and spacemen, studied science and computing while dreaming of a future in space. They went on to change the world as they invented the personal computer and the internet, created video games and self-driving cars, and brought artificial intelligence from science fiction to practical solutions. Today, 50 years later, advanced software systems have been adopted ubiquitously, and recent AI breakthroughs have triggered an unprecedented wave of AI enablement, with far-reaching business and consumer impacts. We’re at the point with AI that new techniques, like evolutionary computation, are even able to automatically create AI models, learning from scratch to optimize decisions with multiple goals. This is applicable to almost every important business decision we make, allowing us to optimize business areas like supply chain, sales and marketing, and patient care, to name a few.
To demonstrate how this works, Cognizant's AI team has shown how evolutionary AI algorithms can be used to virtually land on the moon, with optimal accuracy and fuel levels, and intricate control of speed, deceleration and location using the various engines. The programmers at NASA would have been fascinated to see this in action.
Based on prior landing data, our Evolutionary AI solution builds a model for how various controls interact with the lunar environment, with the resulting outcomes in fuel consumption and landing accuracy. This model is similar to how human pilots think about the world they see. Just as pilots can strategize and virtually experiment with various maneuvers in their head, Evolutionary AI builds optimum decision strategies against the model, which is a surrogate for the real world, and can explore through simulation every possible outcome, learning as it explores.
This enables creativity in a cost-effective, low-risk manner because the various strategies are being tested virtually. Then, as a new, often unexpected, situation presents itself, the AI system is able to utilize the best decision strategy to deal with it, learning to land the aircraft accurately, efficiently and safely.
The progress in AI over the last decades has been driven by the passion of data scientists and data engineers, designing neural networks and deep-learning systems to understand data and recognize patterns. With new technologies that can use AI to create AI and to solve complex, multi-objective problems, just imagine what the next generation of engineers and scientists will be able to do. Humanity’s next steps to explore planets and stars will certainly be aided by AI.
Click here to learn more about Evolutionary AI.