Rebirth: Starting from Lighting Up the Tech Tree

Chapter 173 Breakthrough

There were more than twenty people sitting in the large conference room on the third floor of the research institute.

Li Guodong stood in front of the projection screen, which displayed a task breakdown table divided into three main modules from top to bottom. At the top, three headings were marked in red: Attitude Control, Thrust Adjustment, and Landing Buffer.

"One month," Li Guodong said. "That's the timeframe set by President Zuo. If we can produce simulation verification results within a month, we can use the data to discuss cooperation. If we can't, we'll have no leverage at the negotiating table."

The conference room was quiet. Qiu Pei sat in the first row, Wei Jia was a few rows behind, Ma Hao and two others sat by the window, and Shen Yiming was in a corner.

Zuo Cheng stood at the doorway, not going in. After listening to Li Guodong's opening remarks, he turned and walked towards the stairwell.

He's not here for a meeting. He's here to work.

He was confident in the attitude control module. The algorithm framework provided by the blade on the system panel was essentially a hierarchical control strategy. The bottom layer used traditional PID control for fast response, while the upper layer used a lightweight neural network for disturbance prediction and adaptive parameter adjustment.

He compiled this idea into a technical memo and handed it to Li Guodong the night before last. After reviewing it, Li Guodong said, "This framework is at least a generation ahead of our previous solutions."

Zuo Cheng did not explain the source. He said, "Let's verify it first."

The verification work was carried out in the simulation laboratory. Qiu Pei was responsible for translating Zuo Cheng's algorithm framework into code, and Wei Jia was responsible for building the dynamic model of the rocket's reentry stage.

On the first day, the two of them sat in front of a monitor for eight hours, adjusting the parameters.

"Attitude angle tracking error, steady-state value 0.3 degrees." Qiu Pei looked at the curve on the screen, "It's an order of magnitude smaller than traditional PID."

"But that's only under ideal conditions," Wei Jia said. "The real question is whether we can maintain this level of accuracy when we factor in wind disturbance and uncertainties in aerodynamic parameters."

Zuo Cheng stood to the side for a while and said, "There's a time constant in the perturbation prediction layer; change it from 0.5 seconds to 0.3 seconds."

Qiu Pei was taken aback: "How did you know you needed to change this parameter?"

"We figured it out through trial and error," Zuo Cheng said.

It wasn't actually determined through trial and error. It was determined by the parameter tuning rules built into the algorithm framework within that blade. The rules explicitly state that, under atmospheric disturbances during reentry, the time constant should be approximately 0.3 seconds.

Qiu Pei modified the parameters as instructed and ran the simulation again. The results showed that even after adding 30% aerodynamic parameter uncertainty and a gust of wind of 15 meters per second, the attitude angle tracking error remained within 0.5 degrees.

"This level of precision is sufficient," Wei Jia said.

The breakthrough in the attitude control module took five days. The subsequent thrust adjustment module was even more complex.

The core challenge in adjusting engine thrust lies in the fact that multiple ignitions are required during rocket recovery, and the magnitude and duration of thrust for each ignition must be precisely controlled. Too much thrust will cause the rocket to decelerate excessively. Too little thrust will cause the rocket to crash to the ground.

Li Guodong, along with two engineers from the propulsion system background, worked three consecutive all-nighters to develop a thrust planning algorithm based on convex optimization.

"But there's a problem," Li Guodong said in the conference room at 3 a.m., "The computation time is too long. For real-time control, the solution time must be within 50 milliseconds, but our algorithm currently takes 200 milliseconds."

Zuo Cheng flipped through the algorithm documentation and pointed to one of the constraints: "This second-order cone constraint can be removed."

"Removing it will change the feasible region."

"It won't change much. The physical constraints of thrust planning are mainly the thrust magnitude and the rate of change of thrust, which are linear and can be retained. The second-order cone constraint can be replaced by iterative correction."

Li Guodong had the engineers modify the algorithm according to Zuo Cheng's suggestion. The solution time was reduced from 200 milliseconds to 35 milliseconds.

The last module is the landing buffer.

Two engineers brought by Ma Hao from the Unmanned Systems Division played a crucial role. They drew inspiration from the folding buffer mechanism of drone landing legs and designed a four-legged folding landing system.

"The landing overload can be controlled within 3G," said Ma Hao's engineer. "3G is a design specification for the Falcon 9."

After all three modules were completed, Li Guodong organized a full-system joint simulation. The attitude control, thrust adjustment, and landing buffer modules were linked together to simulate the complete process of the rocket from first-stage separation to vertical landing.

The simulation started at 9:00 AM.

The results came out at 11:30.

"Landing accuracy: horizontal deviation 2.7 meters. Vertical velocity: 0.8 meters per second. Attitude angle deviation: 0.4 degrees. All meet design requirements."

Li Guodong stood in front of the screen and looked at the three sets of data three times.

"Let's try again," he said, "with a different set of perturbation parameters."

The second simulation incorporated stronger wind disturbances and larger aerodynamic parameter deviations. The results still met the requirements: landing accuracy of 3.1 meters, vertical velocity of 1.1 meters per second, and attitude angle deviation of 0.6 degrees.

Li Guodong tried three different perturbation conditions, and all the results were within acceptable limits.

The sixth simulation ended at 7 p.m.

Li Guodong turned around from in front of the screen and looked at the dozen or so people in the lab. Some were asleep on the table, while others were staring blankly at their cold coffee cups.

"The entire system has been tested and approved," said Li Guodong.

There was a two-second silence in the lab, then someone started clapping.

Zuo Cheng stood at the very back and did not applaud. He looked at the system panel: 517 points, six branches, and a 1.3x increase in technical skill.

"Patent," Zuo Cheng said. "Compile the algorithm into a patent application and submit it to the legal department tomorrow."

Li Guodong nodded: "Someone is already organizing it."

That night, twice as many people were working overtime at the research institute as usual.

It was already 9 p.m. when Yu Ying came up from the second floor.

"The data verification for the second-generation satellite is complete," she said. "All three sets of abnormal data have been resolved; they were caused by calibration errors in the testing equipment."

"Okay," Zuo Cheng said.

Yu Ying glanced at the data on the screen.

"The entire system has passed?"

"It passed," Zuo Cheng said. "All six simulations, with different disturbance conditions, met the standards."

Yu Ying sat down next to him. The others in the lab were talking in hushed tones.

"What's next?" Yu Ying asked.

"Talk to the China Aerospace Science and Technology Corporation," Zuo Cheng said. "Bringing simulation data along will carry more weight than just empty words."

Yu Ying nodded.

What was the landing accuracy when SpaceX successfully recovered the object for the first time?

"The publicly available data indicated a deviation of about four meters from the target point," Zuo Cheng said. "Later, we optimized it to within one meter."

"Then we made it 2.7 meters tall for our first simulation."

"Simulation is simulation, and reality is reality," Zuo Cheng said. "There are hundreds of test launches in between."

"But at least the direction is right."

"The direction is right," Zuo Cheng repeated.

He turned off the screen in front of him, stood up, and stretched. Outside the lab window lay the night sky of Hangzhou, with distant streetlights forming a ribbon of light.

Next, the China Aerospace Science and Technology Corporation.

He wants to take this data with him to discuss a real collaboration.