Moonshots How Engineers and Scientists Are Achieving the Impossible

[MUSIC PLAYING]

Welcome to MATLAB Expo 2023. 20 July, 2019, marked the 50th anniversary of the first ever lunar landing. It was the definitive scientific and engineering achievement of that era, and it is still one of the greatest human achievements of all time.

The Apollo mission was the first literal moonshot. It captured the imagination of people across the world because of the sheer audacity of the idea, and the perception that it is dangerous, unprecedented, and virtually impossible. Many technologies that enabled Apollo probably wouldn't have been invented or perfected without the bold vision of landing humans on the moon.

Many things that we now take for granted, such as metals and alloys that can withstand intense heat, fireproof fabrics, freeze-dried food, and photovoltaic cells all trace back their origins to the Apollo program. Integrated circuits, computers that we use every day, and software tools, such as MATLAB, that we all love, owe a debt of gratitude to that original moonshot that brought these technologies into the mainstream. So a big thank you to the Apollo program and to NASA.

Moonshots have come to mean projects with lofty and seemingly impossible goals. Moonshots increase our knowledge and understanding of the world around us, and they often lead to innovations that fundamentally improve our standard of living. In this presentation, I will show you inspiring examples of the vision, courage, creativity, and determination of engineers and scientists who are pursuing lofty and seemingly impossible goals in the unlimited clean energy space, improving quality of life through health care, and of course, space exploration.

Through these examples, I will also show you how MATLAB, Simulink, and model-based design enable each of us to dream, define, pursue, and achieve our own moonshots. Let's start with the critical moonshot of our generation, unlimited clean energy for everyone. The rapid growth in population and economic activity, particularly in Asia, Africa, and the Middle East, is increasing our energy consumption.

Global energy demand is expected to grow by 50% or more by 2050. Meeting this increasing energy demand is essential for securing our planet's future. But we must do this while protecting the environment and tackling the existential threat of climate change. Photovoltaic arrays have come a long way from their exotic and expensive beginnings in the Apollo program. Solar and wind energy are now mainstream, renewable energy sources.

We are also experimenting with harnessing ocean waves and other green sources. But we need more, a lot more to satisfy the projected demand. By exploring the universe around us, we learn that the sun and the other stars produce enormous amounts of energy through fusion. In many ways, fusion is the ultimate clean energy source.

Fusion reactors don't create carbon dioxide or other harmful emissions. They are safer than traditional fission reactors because any disturbance in the fusion will cause them to stop instead of melting down. The fuels being tested, such as hydrogen and deuterium, are abundant. So can we actually mimic the sun and achieve net energy positive fusion? This is the grand challenge that several teams, from large national labs, to new startups, are trying to answer.

Tokamak Energy is the first private company to achieve 100 million degrees Celsius in a fusion device. This is the temperature at which fusion is expected to be commercially viable. The Lawrence Livermore team recently announced that they achieved the first break-even controlled fusion. Lawrence Livermore's achievement has already been called one of the most impressive scientific feats of the 21st century.

So how does it work? They start by creating a perfect 2 millimeter diamond shell and coating the inside with a microscopic layer of hydrogen isotopes frozen at a temperature near absolute zero. This target is zapped by a bank of 192 lasers that, together, are 50 times more powerful than any other in the world.

The lasers zap the shell for 100 trillions of a second. When the hydrogen shell implodes, it releases a flood of neutrons, carrying 1.5 times the energy put in by the lasers. This net gain of energy has never been achieved before.

Lawrence Livermore combined the power of MATLAB, Image Processing Toolbox, and a layering toolbox that they built themselves to grow and regulate the microscopic layer of hydrogen ice on the inside of the tiny shell. You can hear more details of this major breakthrough in Lawrence Livermore's presentacion in the algorithm development and data analysis track.

The pursuit of unlimited clean energy is not limited just to big, government-funded programs. Modern engineering tools and workflows have made it possible even for startups and small teams to pursue and achieve moonshots. Tokamak Energy, a global leader in commercial fusion based near Oxford, UK, built a spherical Tokamak device, the ST40.

On 10 March, 2022, they announced that the ST40 reached 100 million degrees Celsius, the temperature threshold required for commercial fusion. This was the first time a private company had achieved this critical milestone. The Tokamak team used Simulink and Simulink Coder to develop and deploy the plasma control algorithms, and MATLAB for real-time and post-pulse data analysis. Tokamak expects their next generation device to demonstrate the full potential of high-temperature superconducting magnets and start operating the world's first commercial fusion pilot plant in the early 2030s.

From unlimited clean energy, let's turn our attention to revolutionizing health care to improve our quality of life. Treating and curing diseases, helping people recover from devastating injuries, enabling everyone to live healthier, fuller lives, these are overarching and enduring goals that require many moonshots. Here are two recent moonshots in health care, an autonomous pediatric exoskeleton and brain surgery assisted by AI.

For thousands of years, we looked at the moon, but we couldn't go there. But we dreamed that one day we would. For thousands of years, many children with severe neurological disorders couldn't walk, but we dreamed that one day, they would. The vision, innovation, and determination that made our first steps onto the lunar surface possible also empower a physically disabled child's first steps on Earth, and their impact is similarly profound and life-altering.

More than a dozen commercial, nonprofit, and academic partners have joined hands with the robotics and mechatronics group at the Junior HEI Graduate School of Science and Engineering in France. They built an exoskeleton for 8 to 12-year-old children with severe neurological disorders, including lower limb paraplegia. The exoskeleton is designed to accommodate changing heights and weights of the children.

Here, we see an adult member of the research team testing it. The research team tested the sensors in each foot that detects successful weight transfer, which is important for maintaining balance while walking. Initially, the researchers tried to design the exoskeleton in C++, but the manual coding process dragged along. They needed to write a lot of code to implement each functionality.

The team then turned to Model-Based Design with Simulink, MATLAB, and Simscape Multibody for designing motor dynamics and controllers, including the Speedgoat real-time environment for prototyping and testing. The junior exoskeleton is currently being evaluated with adult patients in the Netherlands. After that, the team will start clinical trials with children.

The modeling, simulation, and implementation track at Expo includes several talks that show how many organizations from a variety of industries are using Model-Based Design to eliminate error prone manual steps, accelerate prototyping, perform predictive maintenance, and much more. Now, let's look at the second example of health care moonshots that showcases the courage and determination of scientists going after Parkinson's disease, a disease with no known cure, using AI, one of the newest technologies available.

The earliest signs of Parkinson's include tremors, difficulty walking, rigidity, and loss of balance. The severity of symptoms grows as the disease progresses, and it severely degrades the patient's quality of life. According to the Parkinson's Foundation, more than 10 million people in the world are living with the disease, and it is the second fastest growing neurodegenerative condition in the world, second only to Alzheimer's.

While there is no cure yet, one of the most effective tools for treating symptoms of Parkinson's is deep brain stimulation. This procedure implants a stimulus electrode in the brain that delivers electrical pulses to disrupt the irregular and hyperactive electrical signals in the brain. For this to work, the neurosurgeon must place the stimulus electrode precisely in the subthalamic nucleus, the STN, a structure smaller than an almond located deep within the brain.

This procedure is quite complex. A misplaced electrode could adversely impact other parts of the brain. According to Dr. Ciecierski of NASK, the National Research Institute in Poland, it can alter the patient's emotions and put the patient at the risk of requiring additional surgery. This procedure is also long, taking three to four hours in the operating room.

The challenge, then, is to reduce the surgery time while improving the accuracy of the electrode replacement. This is Dr. Ciecierski in action in the operating room. He developed an AI software program to assist neurosurgeons as they perform this complex procedure.

He used MATLAB for the entire process, data preparation and cleanup, signal processing, training, classification, everything. We always hear of people working with AI and using AI to significantly improve what they do. This is a great example of AI helping professionals do what they do better. It shows what is really happening within domain-specific AI.

Continuing with our brain surgery story, once they are near the STN, the surgeon moves the microelectrodes 1 millimeter at a time, taking a 10-second recording from each electrode, and repeats this action until the electrodes pass through the STN. One of the key signals being monitored and recorded is the referential voltage data that measures the cortical activity. This is how the raw voltage data looks like.

It clearly shows the fundamental challenges in developing and using data-specific and domain-specific AI. You need a lot of data, and that data will be messy. That messy data will need to be cleaned and filtered, both when training the model, and while using it in production. You need to train your AI model, our classifier, using that cleaned data and deploy the model on the cloud as an embedded system or as an app.

Dr. Ciecierski did all of this in MATLAB because MATLAB is a great platform for developing data-specific and domain-specific AI. The algorithm cleans the voltage data by removing high frequencies, spurious spikes, and other artifacts, and processes the data by applying a variety of techniques, including wavelet transformations and power spectral analysis. This clean data is fed to a pre-trained machine learning classifier which estimates the likelihood that the voltage spikes originated from the STN. Amazingly, the entire signal processing and classification process takes only about two minutes and has 97% accuracy. The four hours long surgery is reduced to a 30-minute procedure.

In case you are wondering about the remaining 3% of the cases, the surgeon can easily see the discrepancies and correct the position of the electrode. Think about that. Neurosurgeons can now do brain surgery in 30 minutes using domain-specific AI developed in MATLAB. Several other organizations are also using MATLAB and Simulink as the platform for developing and deploying data-specific and domain-specific AI. A few of these are featured in the AI tracks at this Expo, including Moffitt Cancer Center's use of MATLAB and machine learning for cancer research and discovery.

Finally, let's get back to where we started, to space exploration. The technology, computers, and software that the Apollo program brought to life are now returning the favor and contributing immensely to the next generation of space programs, to new and loftier goals and to distant planets. Here are two recent moonshots in space exploration, Artemis 1's visit to the lunar orbit and MOXIE making oxygen on Mars.

Artemis 1 launched on November 16, 2022, and released the Orion spacecraft into orbit. As we take a quick tour of Artemis 1 and Orion, I will show you how MATLAB and Simulink served as a common engineering platform for various teams across different organizations and enabled the development of many critical components and subsystems. Artemis 1 was an uncrewed mission to test the safety and functionality of key hardware and software systems, and to show that it can get to lunar orbit and return back safely before putting a human crew on board the spacecraft.

During the launch, the rocket is supported by a new launch tower that includes automated swing arms. NASA engineers used Simscape and Simscape fluids to model the hydraulics of these swing arms to make sure they retract safely and swing away correctly during launch, and make a clear path for the rocket once the engines start. After launch, the rocket's ascent is controlled by a highly automated and precise mission management logic.

The flight software controls this logic and autonomously triggers events like booster separation. NASA used Stateflow to model and validate the mission management algorithms that orchestrate this sequence. It is a very complex state flow model with over 12,000 states and 27,000 possible state transitions.

Once the rocket delivers Orion to orbit, the capsule needs electrical power for its computers, life support, communications, basically everything, and depends on solar panels and batteries for its power. Lockheed Martin and NASA model the Orion power system in Simscape Electrical to validate the design. Once power is established, the journey to the moon can begin.

The Guidance, Navigation, and Control, the GNC system, tells the spacecraft where to go and how to get there, and it has to work perfectly for Orion to reach the moon and return back safely. These algorithms are entirely designed in MATLAB and Simulink using Model-Based Design, including generating the embedded flight control code. MathWorks has worked with Nasa and Lockheed Martin engineers for the entire 15-year development cycle of the GNC system.

Orion's power system design is a great example of the innovation and progress in the electrification megatrend. You can find several more examples in the electrification track. Here is a summary of some of the critical components and subsystems in Artemis and Orion we saw in our quick tour. MATLAB, Simulink, and Model-Based Design enabled their development and served as a common engineering platform for various teams across different organizations. These tools and workflows will continue to be used on future Artemis missions as well, including a planned human mission to Mars.

This is really exciting, right? Sometime in the future, maybe we'll be talking about Marsshots rather than moonshots to describe visionary projects with Simulink impossible goals. So are we really going to Mars? Can we live on Mars? Well, many scientists and engineers are already hard at work to answer some of these questions.

One thing we know for sure is that a journey to Mars would require a lot of oxygen, some of it for the astronauts to breathe, but the majority to help burn fuel for the return trip. A crew of four would need about 25 metric tons of oxygen. Sending that much mass from Earth would cost billions of dollars and involve complicated logistics.

So NASA plans on using what they find on site and extract it by separating oxygen atoms from carbon dioxide, which makes up 95% of the Mars atmosphere. Sitting in the belly of the Perseverance Rover is a gold-coated box about the size of a car battery, the Mars Oxygen In-situ Resource Utilization Experiment, or MOXIE. The device has two main parts, a gas compressor and the solid oxide electrolyzer, SOXIE, which splits carbon dioxide using heat and electricity.

Surrounding these components are pipes, valves, filters, sensors, a power supply, and a small computer. We have all heard of digital twins. It turns out that MOXIE has two twins. So they are triplets, actually.

One of the siblings is a hardware twin at the Jet Propulsion Lab, JPL, in California. It is used for testing commands in hardware before mission control sends them to MOXIE on Mars. A traditional digital twin implemented in software also exists and it is at MIT. The team uses this to test the commands in simulation before trying them on the JPL hardware twin. In the case of MOXIE, it makes sense to also have a hardware twin to test and verify commands because it is impossible to do any repairs to MOXIE on Mars in case something goes wrong. Of course, for most earthly applications, just a traditional digital twin is sufficient

Simulink is a very powerful platform for building digital twins, and that is what this team used to build their digital twin. Three MOXIE control loops were modeled in Simulink. The first loop controls internal pressure by adjusting the speed of the compressor blades. The second maintains a set temperature of 800 degree Celsius in SOXIE, which is needed for electrolysis.

The third loop controls the electrical voltage to keep the current constant. The loops also need to model fault detection and shut down a run when things go awry. On 20 April, 2021, MOXIE made its first run and successfully generated oxygen on Mars.

As we saw earlier, Simulink is a powerful platform for building digital twins, and also, designing control systems using these models. Simulink was used for modeling MOXIE's critical control loops for internal pressure, temperature, and electrical voltage. MIT scientists are already working on MOXIE's successor, using MATLAB to optimize the hardware layout, minimize the mass, and simulate a suite of operating conditions. That new device will spend 14 months on Mars, generating tons of oxygen in preparation for the arrival of the first human visitors to the red planet.

I showed you three sets of moonshots in this presentation, ensuring unlimited clean energy, improving quality of life through advances in health care, and going to Mars. As we step back and look at the big picture, a common thread in these achievements is the vision, determination, and courage of engineers and scientists to pursue lofty and Simulink impossible goals. Another common thread is the tools and workflows that they are using, MATLAB, Simulink, and Model-Based Design, that provide a platform to empower their creativity and accelerate transformative innovations in many disciplines, including AI, electrification, digital twins, robotics, and autonomous systems.

In MATLAB Expo, you will see many more examples of amazing innovation and inspiring discovery that engineers, scientists, researchers, educators, and students are pursuing. It is a great opportunity to learn from your peers, see what's new in the tools and workflows, and take away actionable insights that help you dream, define, pursue, and achieve your own moonshots. Thank you for listening. Enjoy MATLAB Expo 2023.

[MUSIC PLAYING]