Cloud coverings can prevent satellite optical instruments from clearly capturing the surface of the Earth. Still in testing, JPL's dynamic targeting avoids cloud imaging using AI, increasing the proportion of available data, and focused on this 2015 volcanic eruption-like phenomenon at Landsat 8 Indonesia. Credit: NASA/USGS
Editor's Note: For over half a century, we have learned how to use a perspective from space to study the world of our hometown. The instruments we develop often find the path to the spaceship we send to observe the rest of the world of our solar system. The smarter and more capable, the more likely it is to strengthen the vehicles that transmit across the solar system if they can build a spacecraft that operates in spaces close to Earth. Repairing or upgrading space vehicles after the space vehicles are launched is a problem, but it is possible. However, self-monitoring and implementing operations becomes more effective the more you reach on missions, in order to enhance the capabilities of spacecraft farther away from Earth. Artificial intelligence is at the heart of the development of these features. Dynamic targeting is an example of how these advanced features are applied to study our home planets.
In recent testing, NASA showed how artificial intelligence-based technology can help spacecraft provide valuable, more targeted scientific data. This technology allowed the first time Earth-obsessed satellite to look ahead along its orbital path, quickly processing and analyzing images with onboard AI to determine where to point to the instrument. The entire process took less than 90 seconds without human involvement.
This graphic shows how JPL's dynamic targeting uses Lookahead sensors to see what's in future paths of the satellite. Onboard algorithms process sensor data and identify clouds and targets to avoid for closer observation as satellites pass overhead. Credit: NASA/JPL-Caltech
The concept, known as dynamic targeting, has been developed for over a decade at NASA's Jet Propulsion Laboratory in Southern California. The beginning of a series of flight tests occurred in mid-July on a commercial satellite. Goal: To enable orbiters to avoid clouds and improve ground imaging, demonstrating the possibilities of dynamic targeting to autonomously hunt certain short-lived phenomena such as wildfires, volcanic eruptions, and rare storms.
“The idea is to make the spacecraft look like humans. We're not only looking at the data, but we're thinking about how it will be displayed and how it will respond,” said Steve Chien, a technical fellow at JPL's AI and a lead researcher at the Dynamic Targeting Project. “When humans see a picture of burning wood, they understand that it may show a forest fire as well as a collection of red and orange pixels. We have the ability to say 'It's fire' on our spaceships, and we are trying to concentrate that sensor on fire. ”
Avoid the clouds for better science
This first flight test for dynamic targeting will come later rather than hunting certain phenomena like fire. Instead, the point was to avoid the ubiquitous phenomenon: clouds.
Most scientific instruments on orbital spacecraft look down on what's underneath. However, for Earth-focused satellites with optical sensors, the clouds get in the way by as little as two-thirds of the time, blocking the view of the surface. To overcome this, dynamic targeting looks 300 miles (500 km) and has the ability to distinguish between clouds and clear sky. If the scene is clear, the spacecraft image the surface as it passes overhead. If it's cloudy, the spacecraft cancels image activity and stores data storage for another target.
“If you can be smart about what you're taking photos, you'll just image the ground and skip the clouds. That way, this image researcher hasn't saved, processed or downloaded all the images that are not available.” “The technique helps scientists get a much higher percentage of available data.”
How dynamic targeting works
The test will be conducted on the Cognisat-6, a briefcase-sized cubesat released in March 2024. Designed, built and operated by Open Cosmos, the satellites host payloads developed by Ubotica, featuring commercially available AI processors. While working with Ubotica in 2022, Chien's team conducted tests at the International Space Station, which runs dynamic targeting algorithms and similar algorithms on the same type of processor. The results showed that combinations could work for space-based remote sensing.
The Cognisat-6 lacks a forward-looking imager, so the spacecraft tilts 40-50 degrees forward, showing both visible and near-infrared light, a camera pointing to a visible sensor. Once a good looking image is acquired, advanced algorithms of dynamic targeting trained to identify clouds are analyzed. Based on that analysis, the dynamic targeting planning software determines where to direct the cloud-free view sensors. Meanwhile, the satellite tilts towards Nadir (looks directly below the spacecraft), snaps the planned image, capturing only the ground.
All this is done in 60-90 seconds, depending on the angle of the original appearance. This is because low Earth orbit spacecraft speeds at around 17,000 mph (7.5 km/sec).
What's next?
With cloud evasion capabilities proven, the next test is to hunt storms and harsh weather. This is essentially targeting clouds rather than avoiding them. Another test is to search for thermal anomalies such as wildfires or volcanic eruptions. The JPL team has developed a unique algorithm for each application.
“This first deployment of dynamic targeting is a very important step,” says Chien. “The ultimate goal is to make operational use in science missions and to create extremely agile instruments that take new measurements.”
There are multiple visions as to how that will happen – perhaps even if the spacecraft is exploring the solar system. In fact, Chien and his JPL colleagues used data from the ESA (European Space Agency) Rosetta Orbiter to inspire dynamic targeting work from another project they also worked on to demonstrate the feasibility of autonomous detection and imaging diagnostics released by Comet 67p/Churyumov-Gerasimenko.
On Earth, adapting dynamic targeting for use in radar allows scientists to study dangerous extreme winter weather phenomena known as rare and short-lived deep convective ice storms to closely observe existing technologies. Specialized algorithms identify the formation of these dense storms with satellite-looking instruments. Then powerful, focused radar pivots to keep ice clouds in sight, and the spacecraft speeds up overhead and collects data in six to eight minutes, so they are “glaring” at them.
Some ideas include using dynamic targeting on multiple spacecrafts. The results of onboard image analysis from major satellites could be quickly transmitted to trailing satellites that could impose targeting a particular phenomenon. Data can even be fed into the constellations of dozens of orbital spacecraft. Chien is leading the test of that concept called Federated Autonomous Measurement, which begins later this year.
Astrobiology
