Rebirth: Starting from Lighting Up the Tech Tree

Chapter 50 Closing the Net

At the end of February, the LoRa project entered the on-site testing phase.

Zuo Cheng led Liu Wei and the two newcomers on a 40-minute drive to the experimental field base of the Blue Star Provincial Academy of Agricultural Sciences located on the outskirts of the city. The three thousand acres of flat farmland stretched as far as the eye could see. Every twenty meters between the ridges, a half-meter-high white pole was planted, with 412 soil sensor nodes from Ruilian Technology at the top of each pole.

Zhao Kai waited for them at the edge of the field, his face flushed red from the wind.

"All nodes were deployed yesterday, and all hardware self-tests passed. Now we're just waiting for your communication solution to go live."

Liu Wei opened his laptop and connected it to the LoRa gateway on site. A status list of more than 400 nodes popped up on the screen, all of which showed "offline".

"Start flashing the firmware," Zuo Cheng said.

For the next two hours, the four men split up and walked through the fields, using handheld terminals to flash the smart access protocol developed by 402 onto each sensor node. The February suburbs were bitterly cold; the soil in the fields hadn't completely thawed and felt hard underfoot. Zuo Cheng's shoes were covered in mud, and his fingers were numb with cold, but he didn't slow down his firmware flashing.

At 2 PM, all 412 nodes were refreshed.

"Start a network-wide synchronous test." Zuo Cheng stood next to the gateway, watching Liu Wei type the command.

The node status on the screen began to change from "offline" to "online" one after another. Within ten seconds, 387 nodes successfully came online.

But there are still 25 that have not responded.

"The packet loss rate is six percent." Liu Wei frowned. "That's significantly higher than the one percent in the lab simulation."

Zuo Cheng squatted next to the gateway and read the log for three minutes.

The problem lies in the terrain—the experimental field isn't perfectly flat; there's a gentle slope in the northwest corner, and there's obstruction between the sensor nodes and the gateway on the slope, causing signal attenuation to be more severe than expected. The laboratory simulation used an ideal planar model and didn't account for the impact of this micro-topography.

"Two options," he said, standing up. "First, add a relay node at the top of the slope to cover the blind spot in the northwest corner. Second, adjust the transmission power and backoff parameters of the nodes in that area. We'll work on both options simultaneously, with results within half an hour."

Liu Wei took someone to the top of the slope to set up a relay, while Zuo Cheng modified the retreat parameters himself.

Forty minutes later, the test was repeated—all 412 nodes were online, and the packet loss rate dropped to 0.2 percent.

Zhao Kai looked at the neat rows of green status lights on the screen and let out a long sigh: "Awesome."

"We can't celebrate yet." Zuo Cheng didn't relax. "We need to conduct a 72-hour continuous stability test. If the packet loss rate remains below one percent after three days, it will be considered a pass."

He left two new employees on duty at the scene, while he and Liu Wei returned to the company first.

In the car, Liu Wei asked, "Brother Cheng, how did you know it was a terrain issue just by looking at the logs?"

"The signal attenuation pattern is wrong," Zuo Cheng said, leaning back in the passenger seat with his eyes closed. "If the packet loss is caused by channel collisions, the lost nodes should be randomly distributed. But the twenty-five nodes with packet loss in the logs are all concentrated in the northwest, which means it's not a protocol problem, but a physical layer coverage problem."

Liu Wei was silent for a few seconds, then said, "I've learned something."

Three days later, the results of the 72-hour stability test came out—the average packet loss rate was 0.18 percent, and the peak rate was 0.43 percent, far below the 2 percent target required by the Academy of Agricultural Sciences.

The project passed acceptance. The final payment of 120,000 yuan from Ruilian Payment arrived in the account within a week.

That evening, the light screen lit up in Zuo Cheng's consciousness.

[Side Quest: Internet of Everything Completed]

[Rating: Excellent]

Completed projects: IoT communication module (delivered), sensor protocol optimization (delivered), LoRa smart access (delivered)

[Reusable Deliverable: IoT Smart Access Protocol Stack (Four-Module Architecture)]

[Reward: 15 points, unlock the blade's "IoT communication protocol"]

Current points: 128

Current number of blades: 13 (including 1 fusion-class)

Thirteen leaves. One hundred and twenty-eight points.

The newly unlocked "IoT Communication Protocol" blade and the previous "Embedded System Development" blade are adjacent on the technology tree, and they both belong to an unactivated branch node - "Internet of Everything".

Zuo Cheng stared at the gray branch node for a long time.

Activating the second branch requires three leaves from the same field. Currently, the Internet of Things (IoT) field only has two—"Embedded System Development" and "IoT Communication Protocols." One more is needed.

But that's not urgent. The most urgent matter is the heavens.

In early March, Fang Ze's simulation platform was finally debugged, and the parallel simulation environment for 120 satellite signals was officially ready. Zuo Cheng had been waiting for this environment for almost a month—without it, the core algorithm of the multi-satellite parallel architecture could not be verified.

Tang Xu's beam coordination algorithm has been theoretically derived and is now ready to be fed with simulation data.

"Run." Zuo Cheng only said one word on the day the simulation platform was launched.

Fang Ze pressed the Enter key. One hundred and twenty simulated signals were simultaneously fed into the multi-satellite pipeline scheduler, and the data curves on the screen jumped densely like an electrocardiogram.

The first round of simulations took six hours. After the results came out, Tang Xu stared at the data for twenty minutes, his expression changing from expectant to solemn.

"The beam coordination switching latency exceeded the standard," he said, pointing to a set of data. "When the number of satellites being tracked simultaneously exceeded eight, the average beam switching latency jumped from three milliseconds to eleven milliseconds. Our target is within five milliseconds."

Eleven milliseconds. More than double the limit.

Zuo Cheng walked over to Tang Xu to look at the data. The simulation log showed that the root cause of the spike in latency was the computational load of the beamforming matrix—when more than eight satellites were in the field of view at the same time, the matrix dimension increased dramatically, and the computing power of the embedded platform could not keep up with the demands of real-time computing.

"Matrix dimensionality reduction," Zuo Cheng said.

Tang Xu looked up at him: "What do you mean?"

"It's unnecessary to perform full-precision beamforming on all satellites within the field of view. Instead, we prioritize signals—full-precision calculations are used for primary serving satellites, simplified models for candidate switching satellites, and only coarse tracking for the rest. These three levels of precision correspond to three different computational costs, reducing the total computing power requirement to one-third of the original."

Tang Xu froze for a few seconds, then slammed his hand on the table.

"Yes! Layered precision! I've been thinking about how to optimize matrix operations themselves, I never imagined that we could reduce dimensionality from the demand side!"

He immediately began modifying the algorithm. Zuo Cheng helped him derive the boundary conditions for the layered precision—which satellites should use full precision, which should use simplified models, what the criteria were, and how to set the switching threshold.

The two of them postponed the meeting from 3 p.m. to 11 p.m.

The new algorithm was immediately put into the simulation platform and rerun after it was written. This time, the beam switching delay for eight satellites tracking simultaneously was 3.7 milliseconds.

Twelve stars – 4.2 milliseconds.

Fifteen stars – 4.8 milliseconds.

All within five milliseconds.

Tang Xu looked at the numbers on the screen and took a deep breath.

"Chengzi," he called Zuo Cheng for the second time, "your brain really works."

Zuo Cheng smiled and said, "It's not that I'm particularly clever; it's that your beamforming theory has a solid foundation. I'm just looking at the problem from a different angle."

"Stop being modest." Tang Xu stood up and stretched. "Come on, let me treat you to a late-night snack. The barbecue stalls near the school gate should still be open."

"No need." Zuo Cheng picked up his phone and glanced at it. Yu Ying had sent him a message half an hour ago—"The insulated bag is on your desk, it contains braised pork ribs rice. Remember to return it when you're done."

He opened the insulated bag on his desk. The food was still warm.