Technology invades the modern world

Chapter 288 Isn't your progress a little too fast?

Cheers erupted at the Wenchang control center. Everyone there was a professional and well aware of the difficulty of this soft landing.

In short, the US has conducted 11 manned lunar landings, and has successfully sent astronauts to the lunar surface 6 times. However, landing on the edge of the lunar south pole is a first.

In the conference room of Apollo Technology in Kunshan, it was already past midnight Beijing time. The experts from Russia had coffee on their tables; they simply couldn't stay awake without it.

At this time, they were still in Kunshan and had not yet moved to Shenhai, because the office building in Shenhai had not been vacated yet, and the security work had not been fully resolved.

As the commotion caused by Apollo Technology grows, the overall security level is constantly being upgraded. At this point in time, the focus is on ensuring absolute safety within a 20-kilometer radius of the office area.

Therefore, Apollo Technology is not expected to move there until the end of this year.

Kunshan and Suzhou are incredibly envious. Our unicorn, which we finally managed to escape, has been snatched away by Shenhai.

Back at Apollo Technology, the Russian experts were dumbfounded when they saw the data and images transmitted from Wenchang.

I knew Apollo technology was amazing, but I never imagined it would be this outrageous.

To put it simply, in general understanding, your soft landing at the lunar south pole is a groundbreaking event, so the priority is definitely to be safe and cautious.

In short, the approach will definitely be automatic navigation, with remote intervention only in case of unexpected events. As the number of launches accumulates, experience will be gradually gained, and the automatic navigation scheme will be optimized.

This is normal logic, but Apollo Technology's solution is abnormal.

Although they also had remote intervention, this remote intervention did not mean that I wanted to ensure that Wu Gang 0001 could successfully complete a soft landing, but rather that I could ensure that it could land at the lunar south pole. Whether it was a soft landing or not was irrelevant.

Either achieve a soft landing through fully automated navigation, or I will have to remotely intervene to land.

As long as the landing site is on the edge of Shackleton Crater at the south pole of the moon, it will be fine.

As it turned out, they watched as Wu Gang 0001 made a soft landing, completing one difficult maneuver after another, and successfully landed on the edge of Shackleton.

"No, of course it is very difficult. The whole process requires precision in navigation and guidance. From launch to lunar injection to mid-course correction and lunar orbit insertion, it requires extreme precision."

Although China's Long March series rockets achieved precision launches, they still required intervention from the Earth control center mid-flight.
Regarding mid-course correction capabilities, China's Chang'e series demonstrated such capabilities; however, fully automated systems would require even greater robustness.

Alyosha and Alexander were both experts who came along this time.

The former is responsible for orbit calculations during rocket launches, while the latter is involved in communication technology.

After exchanging glances, the two began to discuss in hushed tones.

After Alyosha finished speaking, Alexander added, "I'm not saying these two steps aren't difficult, but rather that what follows is even more difficult."

Whether it's Earth orbit insertion, lunar injection, mid-course orbit correction, or lunar orbit insertion, precise calculations are required.

In the past, China's aerospace precision calculation capabilities reached 80 points. Now, Professor Lin and his team have simply improved those 80 points to 90 or 95 points.

There is already a foundation, and from a technical point of view, there are many solutions, which are nothing more than theoretical solutions, depending on their applicability in reality.

As a master mathematician and aerospace expert, Professor Lin's judgment is beyond question. He can choose the best solution to optimize past technologies.

I think the biggest challenge lies in descent and landing.

There was no precedent to follow in this process; neither China's aerospace industry nor NASA had any.

The Antarctic region is full of mountains and craters, and the permanently shadowed areas are dark all year round. If you use a visual navigation solution, the low-angle sunlight at the edges will have a serious impact.

Think about it, the temperature in the shadowed areas of the moon is -203 degrees Celsius, while the sunlit areas are 54 degrees Celsius. You would hardly find a similar scenario on Earth to test this.

This is the hardest part.

As for NASA's Lunar Node-1 solution, it remains only at the theoretical level. In reality, it would be completely unusable in such a complex scenario! Alexander shook his head, his face full of surprise and admiration.

In the past, everyone was about the same, scoring around 70 points. At most, in recent years, with the money, resources, and investment from China, the score jumped from 70 to 75. This refers to the aerospace industry as a whole. But then, out of nowhere, a monster emerged that could score 95 points, far surpassing NASA, which had previously scored 80 points.

The Russian experts were understandably shocked.

The Lunar Node-1 scheme mentioned by Alexander was proposed by NASA. It is a scheme that uses radio signals to support the lander, ground infrastructure and astronauts to jointly build a precise geolocation, provide navigation and observation services, and digitally ensure that they can quickly determine their position on the moon relative to other spacecraft, ground stations or moving rovers.

This scheme is primarily used to assist lunar spacecraft in orbital maneuvers and guide landers to a successful landing on the lunar surface.

(The image shows a lunar lander equipped with the Lunar Node-1 signal sensor.)
However, the prerequisite is that you need enough signal transmitting and receiving units on the moon to assist each other in the construction of this complete system.

This is also part of a series of lunar navigation infrastructure projects that America plans to build on the moon.

“Imagine getting confirmation from a lighthouse you’re approaching on shore, instead of waiting for news from your home port you left days ago,” said Evan Anzalon, principal investigator for the technology and a navigation systems engineer at NASA’s Marshall Space Flight Center in Huntsville, Alabama, in an interview. “What we’re looking to provide is a lunar network of lighthouses that provides sustainable, localized navigation capabilities, enabling lunar spacecraft and ground crew to quickly and accurately determine their position, rather than relying on control centers on Earth.”

Of course, it's still on Earth; it hasn't gone to the moon yet.

If Lin Ran were still working at NASA, he could simply use a door, build a small sensor, and drop the sensor directly onto it to complete the initial system setup. There would be no need for such a complicated process.

NASA's system is, firstly, only on Earth, and secondly, they need to be able to launch things to the edge of Antarctica. They haven't even taken the first step, so they are far from successful.

That's why Russian experts think your idea is just theoretical.

What they are seeing now is that Apollo Technology's automatic navigation has directly achieved the most difficult soft landing on the edge of Antarctica.

Everyone wants to know how you did it.

Valentin was no exception. He sensed the whispers and eagerness among the experts he had brought. He asked, "Professor, this is truly a remarkable achievement. Apollo technology has created another miracle. Please allow me to offer you my sincere congratulations."

Valentin's compliments were sincere, both because he was genuinely convinced after witnessing the entire process and because the achievements of Apollo technology were undeniable.

If you get into a top-two university in the college entrance exam and people praise you for having a bright future, it's a completely different story. Even if both praises are sincere, the latter might sound sarcastic or insincere.

"But Professor, could you explain to us how you did it?" Valentin asked. "We are all very curious."

Lin Ran thought for a moment, then said, "We've used a lot of technological innovations in this area."

I'll just pick a few points that I think will be of interest to everyone to talk about.

I will mainly talk about the innovations we have made in the field of algorithms to improve the accuracy of overall navigation.

We used convolutional neural networks for lunar terrain relative navigation to perform visual crater detection.

Terrain-relative navigation can improve the accuracy of spacecraft position estimation by detecting global features that act as supplementary measurements to correct drift in inertial navigation systems.

We primarily used convolutional neural networks and image processing methods to build an algorithm that tracks the position of a simulated spacecraft using an extended Kalman filter.

This allows for the visual detection of craters in simulated camera frames during the process, and the matching of these detections with known lunar craters in the currently estimated spacecraft location region.

These matched craters were considered as features tracked using a convolutional neural network.

This system is able to perform more reliable position tracking of image brightness changes and more repeatable crater detection frame by frame throughout the trajectory.

When we tested on trajectories using standard brightness images, the new method reduced the average final position estimation error by 90% and the average final velocity estimation error by 50% compared to Kalman filters using image processing-based crater detection methods.

Oh, by the way, you can see this method in a paper accepted at the 2020 America Control Conference. We made some minor optimizations to that paper. With this algorithm, we ensured that we could detect craters and rocks, and find flat terrain.

The quick-thinking Russian experts had already started researching on their laptops.

"At the sensor detection level, we have collaborated with technology companies in our country that have extensive experience. We combine LiDAR, camera and IMU data and use particle filtering and Kalman filtering algorithms to fuse multi-source data and reduce single sensor errors."

Okay, let me briefly explain. This mainly focuses on the lunar lander navigation solution based on the Terrain Relative Navigation method.

An algorithm was developed on a scaled-down simulated lunar scenario, against which a three-axis motion frame was constructed to reproduce the landing trajectory.

At the tip of the three-axis moving frame, long-range and short-range infrared ranging sensors are installed to measure height.

We all know that distance sensor calibration is crucial for obtaining good measurement results.

Therefore, the sensor is calibrated by optimizing the nonlinear transfer function and bias function using the least squares method.

Therefore, the covariance of the sensor is approximated by a second-order function of distance.

These two sensors have two different operating ranges that overlap within a small area.

To achieve optimal performance within the overlap range, a switch strategy was developed.

After evaluating the switching strategy, a single error model function for distance is found.

Due to different environmental factors, the temperature deviation at the edge of the meteorite is large. Therefore, the bias drift of the two sensors will be evaluated and appropriately considered in the algorithm.

In order to reflect information about the lunar surface in navigation algorithms, a digital elevation model simulating the lunar surface has been considered.

The navigation algorithm is designed as an extended Kalman filter, which uses altitude measurements, a digital elevation model, and acceleration measurements from the moving coordinate system.

The goal of the navigation algorithm is to estimate the position of the simulated spacecraft during its journey from a landing altitude of 3 kilometers to a landing point near the edge of the crater.

Furthermore, the algorithm is continuously updated during landing. To this end, we specifically devised a crater peak detector to reset the navigation filter using a new state vector and a new state covariance.

Everyone listened very carefully.

Alyosha had already found the paper from the America Control Association that Lin Ran had mentioned earlier. Alexander glanced at the abstract for a moment and then muttered, "Pervert!"

Alyosha did not ask why he was a pervert.

Because America's paper stated in its abstract that the average final position estimation error was reduced by 60% and the average final velocity estimation error was reduced by 25%, while Lin Ran's paper, which claimed to have made a small optimization to the scheme, actually reduced the error by 90%.

The two Russian experts racked their brains but couldn't figure out how China had managed to achieve such a small optimization.

"Regarding landing accuracy, as you all know, our launch ultimately aims to ensure that the distance between the fuel tank and the lunar lander is no more than 200 meters."

Including this landing, as you have all seen, the error between our target location and the actual location should not exceed 20 meters.

Our limit is that we can go even lower than 20 meters.

Each landing is in an adjacent location, ensuring that the construction of the lunar base can make the most of existing resources, and that every spacecraft launched to the moon can be put to use.

This, too, is built upon the shoulders of those who came before.

This approach can be traced back to Capuano's work in 2015, where they studied code-based Earth navigation system signal receivers for ensuring accuracy during landings throughout lunar orbit. In that study, they achieved an accuracy of 700 meters.

This means using Earth navigation system signals to support lunar missions. You've probably heard of this, since the European Space Agency researched GNSS receivers in 2021, which they wanted to use on the ESA-SSTL Lunar Pathfinder spacecraft to reduce the accuracy to 100 meters.

Back then, you hadn't fallen out with Europe yet, and they should have been informing you about many of their projects.

GNSS: Global Navigation Satellite System, which includes GPS, Russia's GLONASS, Europe's Galileo, and China's BeiDou.

The Russian experts present were somewhat embarrassed.

What do you mean? Have we fallen out? We've only temporarily suspended our cooperation; it will resume soon.

These Russian experts, deep down, still hope to integrate into Europe. On the one hand, it's their Russian nature, and on the other hand, they have had a lot of cooperation with their European counterparts in their past work. Who doesn't have a few European expert friends?

Lin Ran ignored their expressions and continued, "Later, there were also needs for real-time and accurate position and velocity information to guide lunar probes to the moon, especially during the approach and braking phases. Navigation information is provided by ground-based tracking stations, including S-band ranging, Doppler systems, and very long baseline interferometry. Our Chinese experts proposed an intelligent heterogeneous sensor data fusion method for lunar probe descent navigation in the Chang'e 5 series, and achieved a positioning accuracy of several kilometers."

Building upon the wisdom of our predecessors, Apollo Technologies has developed a system that uses lunar gravity gradient measurements during the approach phase to autonomously navigate the landing of lunar spacecraft.

As the spacecraft approaches the moon, the intensity of the gravity gradient signal will increase.

The spacecraft's onboard gravity gradiometer can accurately measure local gravity gradients and uses the latest lunar gravity model to provide reference values.

To account for the decrease in spacecraft altitude, the cutoff degree and order of the gravity model are gradually increased in order to achieve a trade-off between computational cost and model accuracy.

We developed an iterative Kalman filter for orbit and attitude coupling estimation using gravity gradient measurements and attitude quaternions obtained from star sensors.

The gradient meter noise level was also taken into account.

We conducted simulations before this launch, and the simulation results showed that the spacecraft's position converged rapidly, achieving an accuracy of less than 10 meters in the final period, that is, during descent.

There was an uproar.

The precision that Lin Ran mentioned, the 100-meter precision he mentioned earlier, refers to track precision.

Regarding the final landing accuracy, Lin Ran also mentioned that in China Aerospace's plan, the orbital accuracy is several kilometers.

No, the track accuracy is several kilometers, how come it becomes 10 meters in your case?

From several kilometers to ten meters, isn't that a bit too big a leap?

No one could understand it.

Even more outrageous is that your simulation result is 10 meters.

After all, simulation is simulation, and actual landing is actual landing.

As a result, your actual landing effect has now reached within 100 meters.

How did you manage to do that?
Moreover, Lin Ran explained things in great detail, which is more than enough for an outsider.

But it wasn't that detailed, and even if they understood that much, they felt they couldn't replicate it.

Valentin knew he shouldn't ask, and that he shouldn't pry into other people's privacy from any angle, but he couldn't help himself: "Professor, could you tell us how the specific algorithm is designed?"

Lin Ran's expression changed, and he said, "Of course not!"
I have already told you about our technological evolution roadmap, including the thought process we went through, the previous papers we used, and the core idea of ​​our algorithm design: gravity gradient.

Asking again would be impolite!

Valentin immediately apologized: "I'm sorry, Professor, I was being presumptuous. Please forgive my rudeness, because what you have accomplished is so, so incredible."

We have never seen such a technological leap.

(End of this chapter)