It’s more than making products. It’s the forging of our spirit.
Let’s turn the clock back to the winter of 2020.
After the kickoff meeting for the XPeng P5 LiDAR project in November, the teams from XPeng and Livox hit the ground running and left immediately for the HAP “production site” to conduct a series of studies and tests. On the dark and empty site, the teams gathered and discussed details of the production plant drawings. Little could they imagine that in only nine months, they would be able to achieve a milestone in the automotive industry: China’s first automated mass-production line for automotive-grade LiDARs.
The XPeng and Livox teams were on-site discussing the production plant drawings.
Photographed in November 2020
Site renovation was completed and pending the setup of production facilities.
Photographed in January 2021
The automated HAP production line commenced operations
Photographed in September 2021
In the following September, XPeng P5 with Livox HAP was officially launched. The challenges, however, did not end there. On-time delivery of the initial batch was only the first step, and lying ahead were enormous supply chain hurdles arising from rapidly growing orders for XPeng P5.
1. Being a pioneer means you have no prior experience to fall back on.
2. With multiple delays caused by chip shortages, we had to invest significantly more resources to test alternative solutions.
3. There was no turning back for us after we had received nearly tenfold the amount of expected customer orders.
“It was an incredible challenge, much like crossing a river. But instead of feeling our way through the water with the rocks, we were stumbling across only to realize, while in the middle of the river, that a mountain torrent was heading our way. Whether we pressed on or fell back, the chances of survival were 50%. So, all we could do was pick up our pace and forged ahead,” project manufacturing manager Zhang Wenda said. Despite the gargantuan task, the goal could not be clearer for the teams:
To deliver a tenfold increase in production capacity within three to four months from September.
Fast forward to this day, HAP production has achieved its target capacity 12 months ago, and XPeng P5 with Livox HAP is now a common sight on the roads. To document the entire process, we interviewed team members who were deeply involved in the project and share the highlights and their insights in this article.
“The machine that makes the machine
is vastly harder than the machine itself.”
On New Year’s Day of 2021, XPeng and Livox officially announced their partnership on XPeng P5. Before that, Li Shuai from the project manufacturing team had started laying out an overall plan for the production line. With his extensive experience in assembly processes, he was able to identify the details of various challenges in the project.
“I kept asking myself and my team how we were going to build the first automated production line for automotive-grade LiDARs without any industry standards; and how we would improve our process designs to meet the three production requirements of ‘high performance, stability, and efficiency’. We were also pondering how the supplier’s equipment could be moved to the site within four months; how we would ensure materials with a long lead time of six to eight months––which was common industry practice–-were delivered to us as soon as possible; and how our product and assembly line designs could be done hand-in-hand.”Each challenge on its own posed a risk to the project. But when combined together, they made the project tougher exponentially.
The team brainstormed on project risks and discussed solutions at a project manufacturing meeting
Photographed in January 2021
As the HAP product manager, Shao Peiyuan shared the same sentiments as Li Shuai: “Fundamentally, these issues that he mentioned stem from common challenges facing the LiDAR industry today.”
Even as the industry moves from mechanical to hybrid solid-state technology, LiDARs remain the type of automotive sensors with the highest precision requirements. Due to the complexity of these products, they pose a greater challenge to a manufacturer’s production and hardware capabilities.
Over the past decade, most manufacturers shipped only hundreds or thousands of mechanical LiDARs per year on average, with their production methods being essentially manual. However, as we advance to pre-installing LiDARS for mass production, the target capacity increases to thousands or even tens of thousands of units a month, which drives the need to transform the mass production of LiDARs urgently.
“While Elon Musk may not like the application of LiDARs in cars, I do agree with him on the topic of production at scale, that “the machine that makes the machine is vastly harder than the machine itself” said Peiyuan. Before 2021, the focus of many players in the industry was on the explicit parameters of products, such as their detection range, FOV, resolution, and precision, as well as price-performance ratio. But since ushering in pre-installed LiDARs, we have had to explore the iceberg below the water, meaning to look closer at underlying aspects such as our engineering, manufacturing and organizational capacity.
The industry has entered a new era where players will be competing on multiple levels.
Successful product delivery hinges upon the combination of manifest and underlying abilities
Facing the Unknown
The first challenge faced by the manufacturing team at the outset was: How do they build a production line that is compatible with product design and development upfront, while being able to ensure timely deliveries in the late stages. With how production lines are built traditionally, the risk involved is lower as the facilities are only set up after the design has been confirmed and approved. In our case, the project development cycle was less than nine months, so we had to build the production lines concurrently. Yet with the HAP product design still to be finalized, we also could not afford to alter the assembly line later should there be any change in the product design. To secure delivery to the customer in September, the project team decided to invest 50% more of the budget upfront to build a more flexible production framework.
Early on, the team tested multiple options such as semi-automatic human-machine cooperation, multiple turntables, non-calibrated multi-processing, and fully-automated standard machines. After countless tests and analyses, the team eventually found a solution: a modular design based on standardized workstations. The core concept lies in the standardized planning of equipment: Each equipment is a separate and standard process, such as material placement, dispensing, and screwing, which can be set up in various combinations as needed.
The innovative modular design allows the production line to be modified at a minimal scale within specific modules to address any design changes. A flexible production line greatly mitigates late deliveries caused by external issues during short lead times, and limits the impact of design changes on the overall production framework.
This was the first critical decision that the Livox team had to make early in the project. “Decision-making” is different from “standard process management”. Standard production-line management principles concerning “people, machines, materials, methods, and environments” are applicable in most cases––as long as the team is focused and skilled enough, they will find the one best solution, just as they would in answering a multiple-choice question. “Decision-making”, on the other hand, is subjective and it usually takes a longer time to check the outcome of a decision. Zhang Wenda said: “I had to answer a classic question in a written test for this job interview. ‘A high-speed train is about to hit five people. If you divert the train to another track, only one person will get hit. What would you do?’ At the time, I felt that the question was irrelevant to working in the real world. But now I understand its purpose. The company was trying to find out about the values that supported our decision-making, through behavioral models.”
But whether a partnership is truly win-win has to be tested over a long term, while the making of short-term decisions is easily motivated by self-interest. At the time, the Livox team had a single belief: “Since we’re in the B2B line, our success depends on that of our customers and whether we create value for them. We must be aligned with and share the same values and goals as our customers.” From another perspective, the fact that XPeng, an industry leader, chose to work with Livox for their first LiDAR project meant that they had faith in us and we must never let them down.
“I feel like the team has grown over the course of the project. It was more than accomplishing a mission, as we also had an opportunity to mature and develop ourselves mentally. We will still face many unknowns like this in the future. After this experience, our team members began looking at things from a broader perspective, and our marketing team has revamped their strategies to focus on serving existing customers well and ensuring their needs are met,” said Li Shuai.
Achieving 97% yield with 0.01mm precision
In February 2021, the procurement team began onboarding equipment suppliers as required by the project strategy. Project procurement representative Jin Hongye recalled: “Throughout January, we were discussing and evaluating the production line strategy, because everyone understood that any strategic error would disrupt the entire plan and incur staggering amount of resources to make changes in the middle and later stages. So in February, we started discussing our production line design with potential suppliers.”
At this juncture, the team faced another race against time. They had to decide which equipment, among several top-quality models, best fit the HAP production line.
“For hybrid-solid state LiDARs, the scanning module is a core component, whose quality is determined by motor stability. In turn, the dynamic balancing machine is one of the most important workstations for ensuring highly consistent output from the motor,” Li Shuai stated. On a certain level, assembly precision is fundamental to securing optimal dynamic balance.
The project team indicated to the manufacturing and procurement teams that a 0.05mm precision was required for the dynamic balancing process, while for equipment assembly the standard would be even higher at ≤ 0.01 mm. Despite the wealth of experience of the manufacturing team, whose members had been involved in multiple Tier-1 projects worldwide, they were stumped by the request.
Recalling the scene at the time, manufacturing head Jiang Mingyuan said: “After leaving the meeting room, I spoke to my colleagues in the procurement team about the project. I said we had to work independently first then converge at the top.” The procurement team began sourcing top brands and high-precision robotic arms that meet the efficiency and precision requirements, while explaining in detail the background and matrices of the project to the suppliers so that they fully understand our needs and can provide us with optimal support. Meanwhile, the process team drilled down into every minor detail of the production process. The design of the concentricity jig itself was altered as many as five times over the existing standards.
“We believe that redundancy is essential, especially for new sensors in cars. Everyone in the team is aware that, even when the quality criteria have been met, market launch is only the beginning of the true test of a product,” said Wenda on a more serious note.
Through close cooperation between and repeated tests by the R&D and process colleagues, the dynamic balance yield of the production increased from an initial 50% to over 97% within three months after the equipment was in place.
The Livox and supplier teams commissioning workstations at the production line
“In addition to dynamic balance, the stability of electrical performance is also a key determining factor for product life and reliability,” Li added. The quality of the plugins and circuit connections is critical. To ensure reliable circuit connections, the commissioning of workstations has become one of the most crucial processes in the assembly line.
From the principles of fusion welding to incorporating them into the product design, the Livox team conducted detailed research into the differences between various welding techniques, including the choice of welding wire, temperature resistance of components, thermal conductivity of circuit boards, selection of tooling materials, and even the cleaning frequency for welding nozzles and their preheating time. Finally, after nearly 100 DOE tests and more than 20 slice analyses, the production process achieved a stable permeability of over 75% and a yield of more than 99.7%.
Besides making breakthroughs in typical processes such as dynamic balancing and fusion welding, the team has investigated and resolved difficulties in numerous other production aspects. Each challenge not only tested the technology, cost control, and delivery timeline on our end, but also mounted pressure on our suppliers. “Before this, most of the suppliers were used to catering for traditional vehicle-mounted devices and sophisticated production lines. But now, our partners were willing to explore building LiDAR production lines with Livox. Not only have we overcome new challenges throughout the entire process, from equipment design, process innovation, to operation, we have also gained a deeper understanding of where the industry is heading in terms of technology and how best we can work together as partners,” shared Mingyuan. They have been an invaluable partner to Livox in its journey of driving innovation and progress in the automation of production lines for automotive-grade LiDARs.
IF NOT ME, WHO? IF NOT NOW, WHEN?
In five years, we have successfully incubated six industrial-grade platforms, and the experience forms a solid foundation for Livox’s production testing technology. However, what was constantly on the mind of product manager Peiyuan was how the production line under test could meet the automotive-grade management standards and achieve zero defects in product delivery.
Hardly any user of XPeng P5 would know that the “square box” under their headlights has undergone more than 850 rigorous tests for roadworthiness to provide them with a safer and more comfortable driving experience.
In just six months, the production test team successfully built the first complete and fully automated production line for automotive-grade LiDARs in the industry, having accomplished various project stages from conception, design, commissioning, to operation. The team also improved the project’s UPH productivity by up to six times and first-pass yield to 95% from an initial 10%.
Fully automated focusing and testing of transmitter and receiver modules
Gu Yue, head of production testing, was by nature a no-nonsense man. However, when he started talking about how once they had to work around the clock over two weeks to resolve production test issues, he perked up and turned surprisingly talkative.
“For a period of two weeks last year, I basically slept in the factory. At the time, the situation was very tense with the building and commissioning of the production line. It also didn’t help that our colleagues from sales had told us that the customer had added a lot more delivery criteria for their orders, which meant we needed to scale up our production capacity within a shorter time. So the testing team had to be responsible for not just the test results, but also the overall efficiency of the production line. We decided to review the UPH and first-pass yield on a daily basis, and form a daily action plan right after identifying the critical issues at hand. Then we would go off and fight our battles, then reconvene the next morning to evaluate the new results.”
Even as a “veteran” who has been involved in testing and process onboarding since the team’s founding in 2016, Gu admitted that the challenges of this project were unprecedented. Nonetheless, during those dark two weeks, his only thought was: “The fate of the industry’s first automotive-grade LiDAR production line lies in my hands”. Driven by such strong passion and determination, he overcame all hurdles and delivered outcomes that went beyond expectations.
It’s more than making products
it’s the forging of our spirit
In early 2022, the production capacity for HAP reached nearly 8,000 units a month, with their monthly delivered volumes breaking industry records. The process was not smooth sailing, but the teams achieved their goal of increasing their production capacity by tenfold in 3 months.
For everyone involved in the project, the word “tenfold” is more than a goal or number. It also represents the grit, tenacity, and practical spirit that is built into Livox’s DNA.
Towards the end of the interview, Li Shuai said: “I am so thankful to my colleagues and also happy that they have grown through this rewarding journey. Qiang who’s in charge of the machines used to be a person of few words, but as the project progressed, he grew more cheerful and even chatty. There is also Yue, a true ‘ironman’ at the production line who never backs down from any difficulty. He’s our key liaison on the frontlines that connects our R&D and the suppliers. Mingyuan also added: “Besides doing our best in our own duties, the most precious thing is that we always had each other’s back at the most critical points of the project. Whenever a team was struggling at one phase, the other teams would show up and support them fully. Usually, we may identify ourselves with different departments such as the factory, R&D, procurement, and sales. But while in a project together, we become ‘One Team’––sharing the same purpose and responsibilities, and working closely towards the same goal.”
Peiyuan also weighed in from another perspective: “Back to my point on ‘machine that makes machines’ earlier, what I’ve witnessed in the building of this ‘machine’ of ours was more than just everyone’s passion, but also the ability of the teams and the systems to operate sustainably into the future. We know that passion is important when we’re attempting something new, but the key to ensuring an organization does something consistently well in the long term is ‘habit’, which is reflected in the development of teams and systems.” After delivery to the customer, product supervisor Peiyuan led the team in reviewing the project for another three months and identifying weaknesses in areas from decision-making, management to implementation. They compiled their findings into over 300 items spanning five categories, which serve as a valuable reference for other ongoing projects.
At a time when industry trends and environmental conditions are unpredictable, Livox remains grateful to each of its customers for their faith and trust, and will continue to embrace challenges without fear and deliver its promises to customers, with honesty and integrity.
As shown in the HAP production process, we believe that what Livox is doing is more than making products––it’s the forging of our spirit; and what fuels our uphill climb is not just production capacity, but the power of our conviction.