sbdcsierra Chapter 146 Joint Research and Development
Chapter 146 Joint Research and Development
Sunlight streamed into the laboratory through the floor-to-ceiling windows. Zuo Cheng stood at the door, watching the busy figures inside.
The joint laboratory has been officially operational for three days. The entire 30-person technical team in room 402 is in place, separated from the car company's engineers by a glass partition, each debugging their equipment. On the other side, several brand-new test vehicles are quietly parked beside their workstations, waiting to be activated.
"President Zuo," Shen Yiming approached, lowering his voice, "the coordination process is more complicated than expected."
"How so?"
"There are more than a dozen mismatches in the interface protocol. The car manufacturers use traditional automotive electronic architecture, while our algorithm is based on software-defined design; the underlying logic is completely different."
Zuo Cheng frowned. This was a problem he had anticipated, but he still felt a bit of pressure when he actually faced it.
The first plenary session was held in the afternoon.
Zuo Cheng stood in front of the projection screen, with a clear technical architecture diagram behind him.
"The perception layer, decision-making layer, and execution layer are separated into three layers." He pointed to the screen with a laser pointer. "Our 402 team is responsible for perception and decision-making, while the car manufacturers are responsible for interface adaptation for the execution layer. Our common goal is to produce a mass-producible demo version in three months."
Some people nodded in agreement, while others whispered among themselves in the meeting room.
A young engineer raised his hand: "Mr. Zuo, your perception algorithm is trained on drone scenarios and then directly transferred to the vehicle platform. Can you guarantee the accuracy?"
That's a very good question.
Zuo Cheng smiled slightly: "Good question. We conducted virtual tests covering tens of thousands of kilometers using simulators, and also performed transfer learning on the environmental perception data accumulated by the drones. But what's truly convincing is real-vehicle testing."
Shen Yiming continued, "The simulation accuracy is 97.2%, but a simulator is not a real road condition. Therefore, a rigorous three-stage verification plan was designed, with the closed test track being the second stage."
A murmur rippled through the conference room. "97.2%? On the simulator?" The young engineer's eyes widened.
---
The first technical meeting began immediately after the conference.
Problems soon arose.
"Your lane-keeping assist command output frequency is 20 Hz," the car company engineer frowned. "Our ESC system requires 50 Hz; this is a fixed protocol for Bosch products and cannot be changed."
Shen Yiming flipped through the document: "20 Hz is the optimal frequency for simulation verification; forcibly increasing it will sacrifice accuracy."
"If you can't do it, you can't do it."
The air froze.
Zuo Cheng stepped forward, glanced at the documents on both sides, and drew a simple diagram on the whiteboard: "Add an edge computing unit between the decision-making layer and the execution layer to handle frequency conversion. Our algorithm doesn't need to be modified, and the protocol will suffice."
"Adding a hardware node?" the engineer asked hesitantly. "How much will the latency increase?"
"The edge computing unit has a latency of 0.5 milliseconds, and the additional latency from frequency conversion does not exceed 1 millisecond, for a total latency of 1.5 milliseconds, which is within the acceptable range for vehicle control."
Vice President Bao walked over, looked at the plan, and nodded: "Feasible. Adding a hardware node in exchange for maximum flexibility. We can afford the cost."
The problem was resolved on the spot.
---
Over the next week, similar interface issues arose one after another. Communication protocol mismatch, inconsistent data formats, timestamp synchronization errors... Each one was minor, but when accumulated, they created a huge workload.
Shen Yiming slept in the laboratory for three consecutive days.
At 11 p.m. on the fourth night, Zuo Cheng pushed open the laboratory door and saw Shen Yiming still staring at the screen, with a cup of cold coffee on the table.
"Not back yet?"
Shen Yiming looked up, his eyes slightly dark: "The interface documentation is complete. Three of the four major differences have been resolved, and the last one is in the approval process."
"Thanks for your hard work."
Shen Yiming shook his head: "This break-in period is normal. 402 has a strong technical foundation, and the car company has strong engineering capabilities; they complement each other well. After this break-in period, things will go much smoother."
Zuo Cheng sat down next to him and opened his phone to look at the tech tree panel.
Points: 297. The technology radar is still cooling down.
He calculated in his mind that, at the current pace, producing a mass-producible solution within three months shouldn't be a problem. However, automotive-grade certification would be the real challenge, as the ASIL-B standard is more than ten times stricter than that for consumer electronics.
"There's been a small breakthrough," Shen Yiming suddenly said.
"What?"
"Tests revealed that the perception algorithm could identify obstacles at a distance 15% greater than expected. Originally designed for 250 meters, actual tests showed it could reliably reach around 290 meters."
"reason?"
"The background sky data accumulated by the drones came in handy," Shen Yiming's eyes lit up. "High-altitude flight inherently provides a stronger ability to identify small targets at long distances, and this advantage is amplified when flying on the ground."
290米。左城在心里转了一圈这个数字。特斯拉Autopilot有效感知距离是200米，如果402能做到290米，意味着高速行驶时能早将近0.5秒做出反应。
"We won't release this data yet," Zuo Cheng said. "We'll confirm it after the actual vehicle test results are available."
Shen Yiming nodded. He knew Zuo Cheng's style: cautious and never releasing data that hadn't been fully verified beforehand.
---
Two weeks later, the first integration test took place.
A white test vehicle slowly drove into the closed testing ground. Shen Yiming sat in the driver's seat, his hands hovering on either side of the steering wheel, ready to take over at any moment. Zuo Cheng sat in the passenger seat.
"Test begins," Vice President Bao's voice came through the walkie-talkie.
Shen Yiming pressed the automatic driving mode button, and the steering wheel turned on its own. The vehicle smoothly drove onto the test track, increasing its speed to 60 kilometers per hour. A simulated obstacle appeared ahead, the system issued a soft warning, and then the vehicle automatically decelerated and smoothly bypassed it—the whole process was seamless.
Curving tests, lane change tests, emergency braking tests... each test was successfully completed without a single incident requiring intervention.
Test complete.
The car company engineers had mixed expressions. There was surprise, admiration, and a little bit of resentment.
Wei Dongsheng stood at the edge of the test track, watching the test vehicle slowly come to a stop, a glint of light flashing in his eyes.
"President Zuo." He walked to Zuo Cheng's side, his voice tinged with barely concealed excitement, "This is even better than I expected."
Zuo Cheng smiled slightly: "This is just the beginning. The closed testing ground is only the first hurdle; the real test lies on the open road."
Wei Dongsheng nodded, but Zuo Cheng could see that this fifty-something-year-old automotive veteran had a fire burning in his heart.
He believed that 402 could produce results that would shock everyone within three months.
And this is only the first step.