Smart Navigation with Light Boats: Pressing the Accelerator for Mass Production of Autonomous Driving Technology | In-depth

Author: Zheng Wen

Editor: Zhou Changxian

In the race towards mass production and commercialization of autonomous driving, the “land-grabbing” movement is in full swing, with cost reduction being the evident expression at this stage.

In December of last year, Element AI released an L4 level autonomous driving solution that costs less than $10,000 (approximately RMB 67,000), a price that fell again to $3,000 (approximately RMB 20,000) in April of this year.

On May 18th, autonomous driving company, Lightboat Intelligent Navigation, announced during the first QCRAFT DAY that their fourth-generation autonomous driving solution DBQ V4 production costs have dropped to RMB 10,000.

At the same time, Lightboat Intelligent Navigation announced strategic partnerships with Volcano Engine, T3 Travel, and Horizon Robotics, the leading Chinese intelligent driving chip manufacturer.

Among them, the most notable partnership is with Horizon Robotics. The two parties will jointly develop and adapt high-cost-effective, high-level autonomous driving solutions based on Lightboat’s independently developed “Parking and Anchor” integrated solution and Horizon’s Journey® series of car-grade AI chips.

The continuous decline of the autonomous driving industry’s “bottom line,” as well as the collaboration among upstream and downstream companies in the ecosystem, seem to indicate that the commercialization of autonomous driving is indeed coming.

“Build Rockets, Not Ladders”

Like many high-tech companies, Lightboat Intelligent Navigation’s founding story began with a garage in Silicon Valley. The founding team has extensive experience in the autonomous driving industry, with core members from top companies such as Waymo, Tesla, Nvidia, Facebook, and more.

From the beginning, Lightboat Intelligent Navigation has planned a clear development path, adhering to the principle of “building rockets, not ladders,” to achieve extensive deployment of autonomous driving.

“Our mission has always been to bring autonomous driving into reality, not just starting with ADAS and L4 story, but from scratch,” said Lightboat Intelligent Navigation’s founder and CEO, Yu Qian.Based on this philosophy, QZJH (Qingzhou Intelligence Navigation Co., Ltd.) has created an efficient methodology called the “Autonomous Driving Super Factory”. This is a systematic and automated autonomous driving infrastructure that is data-driven and committed to creating a data automation loop, with the aim of continuously improving efficiency and providing strong support for the research and rapid iteration of QZJH’s technology, products, and various autonomous driving solutions.

In addition, QZJH has proposed a “dual-engine” strategy to build a “power engine” by leveraging its public road L4-level autonomous driving capabilities to drive the efficiency of urban transportation, and to build an “innovation engine” with a high-performance, cost-effective autonomous driving pre-fitted mass production solution, which accelerates the entry of self-driving vehicles into reality.

On May 18th, based on the “Dual Engine” strategy, QZJH released two heavyweight solutions to drive the SPACE mobile travel space solution, which drives the “power engine”, as well as the DBQ V4, the fourth-generation mass-produced L4-level autonomous driving solution, which drives the “innovation engine”.

DBQ V4 supports 1-5 LiDARs, 0-4 blind spot radars, 6 millimeter wave radars, and 12 perception cameras to achieve 360-degree perception with no blind spots and no dead angles, as well as mutual redundancy in both the left and right fields. With advanced software algorithms, DBQ V4 has stable perception performance in the face of different lighting conditions and vehicle motion states.

Yuan Qian further explained, “We can provide customers with different configuration solutions. On the one hand, we can provide full L4-level autonomous driving capability on the flagship version of DBQ V4. On the other hand, we can also achieve 99% of L4-level capability at 10% cost on the standard version of DBQ V4, with mass production costs as low as RMB 10,000.”

“The last 1% is the most difficult autonomous driving scenario, requiring a large redundancy of central software and hardware, various safety backup strategies, and a large number of tests to achieve. But other capabilities can already be well commercialized and provide our mass production car users with a very good driving experience,” said Qian.

It is reported that DBQ V4’s L4-level support includes autonomous lane changing, overtaking, three-point turning, entering and exiting the ramp, and lateral avoidance, including unprotected left turns, narrow roads, and detours around car lanes with no protection. Yuan Qian added, “Many of these capabilities are not defined in L2’s capabilities and are not included. These capabilities are part of a holistic driving experience.”DBQ V4 has high scalability, and it can provide different versions of configuration schemes according to the production needs of different levels of autonomous driving required by the automaker, while maintaining the same technology stack to adapt to different models and different scenarios. It can efficiently adapt to various vehicle types such as sedans, SUVs, MPVs, and buses.

It is worth mentioning that in the field of motion planning, QZJH relies on the “Autonomous Driving Super Factory” to independently create “Spacetime Unified Planning”, becoming the first autonomous driving company in China to adopt spacetime unified planning and realizing automatic driving technology that is most suitable for Chinese roads.

Currently, the industry widely uses the autonomous driving scheme of “spacetime separation planning”, which has limited road response ability and can easily cause poor experiences such as sudden braking. The “Spacetime Unified Planning” system of QZJH endows the vehicles with more sensitive timing control ability, which can smoothly solve various complicated road problems and bring passengers a comfortable riding experience without jitter.

“At the same time, considering the trajectory in time and space can make it more flexible and adaptable to dynamic obstacles on the road, which is very suitable for the complex road situations in China.” Yu Qian revealed that in relatively narrow roads, with a large number of opposing vehicles and dynamic obstacles that are very complex, QZJH vehicles shuttle with ease, “which is highly correlated with our algorithm framework.”

In addition to the characteristics of trajectory planning, relying on real road testing and simulated scenarios constructed from generated data, QZJH Matrix can help customers reduce the testing costs of autonomous driving to 1% of pure road testing and generate millions of long-tail scenarios. Every day, it completes millions of algorithm training, testing, verification, and iterative optimization, making the evolution of autonomous driving software capabilities no longer restricted by vehicles and personnel, and achieving continuous rapid improvement of vehicle intelligence.

“We use automated testing tools extensively, which greatly ensures the quality of our engineering, from unit testing to integration testing, simulated testing to scenario library testing, including our board testing.”

Interestingly, the testing ideas of QZJH team coincide with those of Waabi, an autonomous driving company in Canada.

Raquel Urtasun, founder of Waabi, once said that since the first DARPA challenge in 2004, it cannot be denied that the autonomous driving industry has made significant progress. However, from a commercial deployment perspective, unmanned driving is still limited to very simple and limited scenarios.
This is because traditional development methods for autonomous vehicles cannot fully utilize AI and instead rely on complex and time-consuming manual adjustments, which result in high business expansion costs. Yu Qian confidently stated, “From a design and testing perspective, we have a very strict, data-driven and efficiency-boosting tool chain to ensure the engineering implementation rather than a magical algorithm, a single point of thing, some simple concepts and ideas.”

At the conference, QZJH announced that their autonomous driving development toolkit, QZJH Matrix, which is based on simulation, will officially provide services to customers, assisting them in building their own “autonomous driving super factory.”

It is understood that the toolkit has linked the entire process from data processing, labeling, training, large-scale simulation, and technical output, realizing the efficient use of autonomous driving data and driving the efficient iteration of technology development, becoming a vital support for the QZJH autonomous driving super factory.

Thus, through innovative solutions and an entire toolkit, QZJH has created a more efficient and practical “QZJH Solution” to achieve the front-loaded production of autonomous driving.

Compared to other peers in the competition with abundant resources in the field, what are the advantages of QZJH’s construction? According to Yu Qian, “We must achieve it more efficiently, find a balance between commercialization and technology, and find more suitable application scenarios. Through the implementation of a commercial cycle and a data cycle, we will promote a larger commercial cycle, and gradually achieve autonomous driving.”

In December 2021, QZJH and Dongfeng Yueda Kia jointly released the Sharing Bus based on the Longboat Space scheme. The Sharing Bus has now completed landing operations in Wuhan, Dali, and other cities, and has been widely used in intelligent connection, scenic sightseeing, and other scenarios.

Robobus is the most impressive and recognizable product of QZJH, but it is only a small opening in the autonomous driving market and the first step of its long journey.

As Yu Qian stated, “As a lightweight, efficient autonomous driving company, QZJH has quickly led the commercialization landing of Robobus in just three years, but this is only the first milestone accomplished.”At the end of last year, QCraft unveiled a video, once again attracting widespread attention from the industry. The video showcased the proficiency of QCraft’s autonomous car driving through busy city traffic with ease, handling various complex road conditions.

From the success of Robobus to the growth of Robotaxi, QCraft has chosen a path of “driving greater business opportunities through one complete commercial ecosystem, step by step realizing autonomous driving” in the autonomous driving market, with a near-paranoid pursuit of efficiency in both the underlying technology and commercialization processes. This not only secures its position in a cut-throat business competition, but also demonstrates CEO Yu Qian’s profound consideration of the company’s future competitiveness.

So, what is a more efficient approach? Yu Qian’s answer is “to use war to fuel war”, which is also the most practical and intelligent growth strategy for QCraft.

At the first QCraft Day, the company condensed its thoughts and practices in exploring autonomous driving into the phrase “choose a high place and walk towards a wide future”. This summarizes QCraft’s strategic thinking on the commercial deployment of autonomous driving, indicating a new phase of development.

“Choosing a high place” refers to the company’s mission to bring autonomous driving into reality, as emphasized earlier. “Walking towards a wide future” means efficiently improving the scale of deployment by building an autonomous driving super factory and realizing data loop closure through innovation in the front-loading mass production business.

From a product perspective, the two solutions, Longzhou SPACE and DBQ V4, respectively achieve power engine and innovative engine in depth and width. Driven by these two engines, QCraft ultimately fulfills its mission of bringing autonomous driving into reality.

To AutocarMax, “choosing a high place and walking towards a wide future”, proposed by QCraft, is a wise and practical development path compared to Baidu Apollo’s “climbing Mount Everest and laying eggs along the way”. In other words, it seeks both lofty ideals of the future and current commercial viability, thereby enabling feedback in data and funding.

Currently, with more advanced sensors being assembled and high computing power platforms being gradually applied, advanced-level autonomous driving is rapidly entering the “mass production era.” If computing power is compared to oxygen, then the concentration of oxygen in the environment has greatly increased, resulting in a lot of data from these sensors.

Finding the best partner and seizing the “Golden Turning Point”

With the advent of mass-produced autonomous driving, it is crucial to find the best partner and seize the “golden turning point”.”What kind of company can adapt to such environmental changes? Obviously, autonomous driving companies with useful L4 capabilities like us are better suited for this. ” Yu Qian is very confident.

At the same time, more and more OEM customers have begun to pay attention to intelligent manufacturing. He believes that “industry development is very advantageous for self-driving companies like us. It’s like increasing the oxygen concentration and food abundance by ten times.”

Yu Qian asserts that “under the joint drive of multiple factors such as policies, technology, and market, the autonomous driving industry is at the golden turning point of development. Therefore, cooperation between upstream and downstream companies in the industry needs to be closely coupled and coordinated to create greater value for customers.”

“Intelligent automobiles will become the next mother ecosystem in the field of human technology and require extensive and in-depth industrial collaboration to achieve true scale landing.” Dr. Yu Kai, founder and chief scientist of Horizon Robotics, made the same judgment.

He believes that “in the era of high-level autonomous driving, the best model is to fully open, support our partners to develop their own perception algorithms, map positioning algorithms, prediction algorithms, and regulation algorithms in all aspects. Otherwise, it will be difficult to achieve large-scale production in the end.”

In fact, as autonomous driving scales up, the consensus of the industry has become a safer, more cost-effective, and more comfortable driving experience, which is the direction that everyone is racing towards.

However, Yu Qian said, “Our development philosophy has never been based on single algorithms as scene abilities, but more on engineering implementation capabilities, efficient data loop closure, and efficiency enhancement, which we believe to be the long-term competitiveness for us since day one. We have always emphasized the ability of landing engineering on this basis, and we have also accumulated very strong engineering implementation capabilities.”

“We are prepared to go at least 20 years. This is a very long-term thing. Therefore, technology and commercialization must be well combined on this road.” Yu Qian believes that Horizon Robotics and QZVS Technologies, through technology-driven implementation of the combination of technology and business, are moving towards the ADAS market in an era when the pace of the times is being followed.

Due to the very consistent development philosophy between the two sides, they quickly reached a consensus. With the strategic investment of Horizon Robotics and QZVS Technologies, the two companies are committed to promoting the scale application of high-level autonomous driving technology with higher cost-effectiveness.

Expected in the third quarter of 2022, the two parties will open road testing of autonomous driving sample cars based on the Journey 5 chip. By 2023, the joint development of a high-level autonomous driving solution based on Journey 5 chip by both parties will reach mass production level.

As the only domestic company to achieve mass production of AI chips for automobile regulation, Horizon Robotics is currently the only provider of a full-scenario intelligent chip solution for vehicles from L2 to L4 level, and an industry promoter to build an open-source and innovative ecology. Objectively speaking, it is indeed the best partner to translate advanced autonomous driving into reality.

After the press conference, the affection and admiration between the two parties overflowed during the media communication session. Yu Kai said that he deeply admired the strong strength of the Horizon Robotics team. Yu Qian pointed out explicitly that he valued the high-efficiency gene and ecological position of Horizon Robotics in the industry.

“Horizon Robotics has a deep understanding of the domestic autonomous driving ecology, and has years of cultivation and very deep experience in the entire Chinese autonomous driving industry, which gives us very good support.” Yu Qian sincerely exclaimed, “Our cooperation with Horizon Robotics is carried out in a very efficient and fast pace. This feeling is very strong.”

Yu Kai added, “We are close to local consumers, local car factory customers, and close partners. For example, if Horizon Robotics needs any support in the tool chain for their chips, we can respond quickly, iterate quickly, and truly create long-term value for consumers and manufacturers.”

Their dialogue was sincere and warm, but not polite. Previously, Li Xiang, the founder of Ideal Automobile, expressed his admiration for the Horizon Robotics team without reservation in an interview with the media.

“Their large professional teams work directly with us, which is completely different from the cooperation with chip manufacturers before. Our autonomous driving team has encountered various suppliers before, and Horizon Robotics is the team with the highest cooperation degree and professionality we have ever seen, which greatly exceeds our expectations.”

In Yu Qian’s view, cooperating with Horizon Robotics can further strengthen the customer service of Horizon Robotics, “In the future, for mass-produced passenger cars, it is difficult to distinguish between L2, L3, or L4-level vehicles. What we care more about is whether we can provide customers with the best value experience, comfortable driving experience, and reasonable cost. This is the question I need to answer.”

“We cooperate with Horizon Robotics with an ideal in mind.” At the end of the interview, technology expert Yu Kai emotionally said, “In the era of smart cars, we not only have Chinese brands like NIO, XPeng, and Ideal, but we must also have Chinese solutions. Thank you for the trust of Horizon Robotics at this time point. We must make Chinese solutions.”This is the opportunity bestowed by the era.

These are the people who are making the greatest efforts for the “core chips + powerful software” autonomous driving solution in China at such a golden turning point.

This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.