AI Trends

Niwa Robotics Founder Yang Ruigang: The Data Pitfalls Autonomous Driving Fell Into, Robots Will Step Into Again

Sina TechUpdated Nov 18 19:15

AI summary

Niwa Technology founder Yang Ruigang has chosen community last-mile delivery as the implementation scenario for embodied intelligence, proposing the concept of "embodied action intelligence" and building a system around the World Travel Model (WTM) that encompasses data generation, physical simulation, and model training. He assesses that generalizing robots for homes and factories is difficult or demands strict cost requirements, whereas community delivery can leverage autonomous driving technology and gradually expand its scope of application. To address Corner Cases in real-world environments, Niwa emphasizes advancing routes by combining lightweight topological maps with real-time perception, prioritizing the resolution of interaction issues such as opening doors and riding elevators to control ROI.

Body · AI translation

New Crowdsourced Testing Promotion Copy New Crowdsourced Testing Promotion Copy end Add App-exclusive traffic diversion

After more than a decade in autonomous driving, Yang Ruigang has decided to shift his next entrepreneurial venture from highways to residential communities.

Before this, he had studied 3D vision at U.S. universities, joined Baidu Research to lead the Robotics and Autonomous Driving Laboratory, and served as CTO of Inceptio Technology, promoting the pre-installation mass production of autonomous heavy trucks. From researching algorithms to getting dozens-of-ton trucks truly running on highways, he witnessed the journey of autonomous driving from technical exploration to commercialization.

In 2024, Yang Ruigang began preparing for his second startup. At that time, humanoid robots had become the most watched direction in the embodied intelligence industry. He had also explored related entrepreneurial opportunities but ultimately chose a scenario that seemed less "sexy": community last-mile delivery.

He judged that while home robots have a large market space, every household environment is different, making generalization extremely difficult; factory robots must meet strict production cycle times and cost requirements. In contrast, community delivery can reuse positioning, navigation, and obstacle avoidance technologies accumulated from autonomous driving, and allow robots to begin interacting with the real physical environment through tasks such as opening doors, riding elevators, and delivering packages.

This was the starting point for Yang Ruigang's founding of Niwa Technology. He hopes to first let robots solve mobility and passage issues, then gradually move into more complex tasks. To this end, Niwa proposed "Embodied Mobility Intelligence," building a system for data generation, physical simulation, and model training around the World Travel Model (WTM).

From witnessing autonomous trucks driving on open roads to robots handling various complex tasks between buildings, Yang Ruigang has always been challenging technically unsolved problems. Yet he says the real opportunity lies precisely in those technical problems that remain unsolved.

"We are not afraid of difficulty."

From Lab to Mass Production: Re-understanding the Value of Technology

Yang Ruigang: Actually, I didn't start out doing computer vision; I did computer graphics.

I was in the CAD and CG lab when I was at Tsinghua University. Later, I went to the University of North Carolina at Chapel Hill for my Ph.D. because it is very famous in the field of graphics.

When I first arrived, I mainly worked on graphics rendering and display, and later participated in a project called Telepresence. It aimed to solve the problem of whether real-time 3D information of a person could be collected through multiple cameras and rendered from different perspectives, allowing people far apart to hold meetings face-to-face.

As we worked on it, we found that rendering and display were gradually no longer the hardest problems; the real challenge was how to reconstruct 3D information in real time.

So my doctoral research shifted to 3D vision, studying how to recover 3D structures from images, point clouds, or depth information. Later, I began researching semantic information, not just knowing the 3D shape of objects but also what they are.

I entered this field in 1998, much earlier than the rise of deep learning. UNC was very strong in graphics at the time, and I was among the early batch of people submitting papers to CVPR.

But I didn't feel like I was sitting on a cold bench, because there were still a large number of unsolved problems in 3D vision itself.

Yang Ruigang: The earliest applications of deep learning in computer vision were mainly concentrated in tasks such as image classification, detection, and segmentation. The widespread adoption of deep learning in 3D vision actually came later.

My earliest attempt was to introduce constraints from traditional 3D geometry into deep learning frameworks. At that time, I also collaborated with Su Hao and others to study how to combine 3D geometry with deep learning.

Later, I worked on directions such as generating new views, diffusion models, and video generation control.

For researchers, new technologies require relearning. It's just that the learning frequency is getting faster now; before you fully understand the previous things, new methods may already have emerged to replace them.

I believe scientists must possess the ability to learn from scratch.

Yang Ruigang: We have done large-scale street scene 3D reconstruction for simulation training. At that time, the U.S. placed great emphasis on using simulation systems, hoping to understand streets and terrain through 3D reconstruction before entering real environments. We were doing this even earlier than Google Street View.

We also researched how to convert very dense point clouds into simplified polygon models while retaining high-quality textures. Later, Apple Maps' 3D maps adopted similar technical approaches.

Commercialization attempts were made too. I received SBIR project funding twice in the U.S. These projects provide hundreds of thousands of dollars, with the purpose of pushing paper results toward product conversion.

But at that time, it was basically informal. If interested, we continued working; if not, we might look for jobs.

At that time, we lacked strong business awareness and didn't have anyone particularly skilled in commercialization to drive it, so nothing really took off.

New Smart Drive Physical AI: You had already conducted years of research at U.S. universities and attempted technology commercialization. Why did you decide to join Baidu in 2016? Who invited you?

Yang Ruigang: Lin Yuanqing invited me, Andrew Ng was responsible for the interview, and finally, I had an online exchange with Robin Li.

Lin Yuanqing and Andrew Ng left successively later on. For about two and a half years, I mainly worked under the leadership of Wang Haifeng.

In 2016, Baidu Research had about five or six scientists like me. Each was responsible for a relatively independent laboratory, having significant autonomy over personnel and technical directions. The operational style was actually quite similar to university labs. Later, I was responsible for the Robotics and Autonomous Driving Laboratory.

"New Intelligent Driving Physical AI": You promoted the ApolloScape dataset during your time at Baidu. Why did you start this project? How much resources were invested?

Yang Ruigang: I owe this project to Robin Li.

After Robin Li joined Baidu, he placed great emphasis on building a technical brand. He felt that since there was already a well-known dataset like KITTI in the autonomous driving field, Baidu should also have its own data brand. This task was assigned to me.

The core team for ApolloScape consisted of about 10 people and worked intensively for around half a year. Participants included colleagues such as Huang Xinyu, Geng Qichuan, and Zhou Dingfu, who were responsible for tasks like data collection, cleaning, and tool development.

We also received support from other teams within Baidu. For instance, we used high-precision data collection vehicles equipped with high-precision LiDARs, navigation-grade IMUs, and GPS, and added some extra sensors.

One hundred percent of the ApolloScape data came from real-road collection, not simulation-generated. However, later on, the company's business focus shifted, and the dataset did not receive the same level of continuous construction and maintenance investment as originally planned.

Yang Ruigang: What impressed me most was the business model proposed by Inceptio, called 'Dual-Driver to Single-Driver'.

Many long-haul freight operations in China previously required two drivers to take turns driving. If a sufficiently reliable assisted driving system could be developed, allowing one driver to transition into a safety officer role, there would be an opportunity to reduce the crew from two drivers to one. This would not only improve transportation costs but also allow for continuous data accumulation during actual operations, gradually evolving towards L4 autonomy.

I believe this is a progressive route. If we aimed for L4 right from the start, mass production and large-scale deployment would be extremely difficult. Especially for heavy trucks operating on highways, issues regarding road rights are even more complex.

In comparison, starting with commercially viable assisted driving and then gradually increasing the level of automation is a more realistic choice.

Yang Ruigang: When I joined, my main responsibility was to establish a complete R&D system for autonomous driving technology.

At that time, the company had R&D centers in both China and the US. The initial approach was somewhat similar to 'R&D in the US, implementation in China'. But after I joined, I explicitly stated that this model would be hard to sustain in the long term. Autonomous driving requires extensive real-world testing. Since the product would ultimately be implemented in China, the R&D team needed to be as close to real scenarios as possible.

Later, the Chinese R&D team grew from about 10 people to around 200, while the US team shrank from approximately 50 to about 10, and further down to single digits.

I have always believed that research can adopt a relatively loose cross-border collaboration model, but for true mass-produced products, R&D, testing, and delivery must be tightly integrated. Beyond on-site testing, cross-border collaboration involves practical constraints such as chip supply and legal regulations.

Yang Ruigang: I often say: We are 'dancing in shackles'.

Because for mass production, many hardware conditions are determined in advance.

For example, chip computing power and sensor configurations are already set. You cannot just swap in a better card if you feel the computing power is insufficient, as you might when doing research.

In mass production, a few serious failures can prevent the product from leaving the factory. Making a product is like a wooden barrel; the shortest plank must be long enough.

Heavy trucks are a good example. The difference between empty load and full load is significant, and the dynamic relationship between the tractor and trailer is complex. We also spent a lot of time on fuel efficiency. Commercial vehicles are production tools; customers calculate costs and efficiency, so these issues require technological answers.

Yang Ruigang: Yes. I have accompanied vehicles many times, communicated with drivers, and met with vehicle manufacturers and heads of express logistics companies.

I really enjoyed this process because it allowed me to understand customer needs from completely different perspectives.

As a CTO, one cannot only care about technology; one must also know who the customers are, why they buy this product, and what problems they truly want to solve. It is the biggest taboo for technical leaders to work behind closed doors.

Embodied Intelligence Implementation: Find Fit First, Then Discuss Generality

Yang Ruigang: When I seriously considered starting a new venture again in 2024, our earliest name idea was Numa, which corresponds to 'Mu Niu' (Wooden Ox) referencing the Wooden Ox and Flowing Horse. I thought the English pronunciation was simple, easy for foreigners to read, and sounded somewhat like Puma. I had even registered the domain name.

Later, younger colleagues, especially HR, reminded me that this name absolutely could not be used because in Chinese it sounds like 'Niu Ma' (beasts of burden/slaves). While I, as the founder, could joke about it myself, it seemed inappropriate for employees to constantly say they were working at a 'Niu Ma company'.

So we continued adjusting the original name, eventually arriving at 'Niu Wa'. 'Niu' means connection or bond, and 'Wa' comes from Nuwa, the creator of life in mythology. We hoped to create a new form of robot, or even a new species.

Yang Ruigang: I always ask where the Product-Market Fit lies. There is still a long way to go from technological breakthroughs to large-scale applications.

Both homes and factories have their own difficulties.

Homes seem relatively enclosed, but every home environment is different. Item placement, furniture structure, and user needs vary greatly, requiring very high generalization capabilities.

The main issues in factories are cycle time and efficiency. A robot might be able to perform a certain action, but if its speed is far slower than human labor or its cost is several times higher, customers will find it hard to accept.

In comparison, last-mile delivery does not have such high requirements for fine manipulation and action cycle times.

In community delivery, part of the problem is familiar to us, such as positioning, navigation, and obstacle avoidance. Moving forward, the robot needs to swipe access cards, open doors, take elevators, and hand packages to users, interacting with the physical environment. This is an interesting intersection between autonomous driving and embodied intelligence.

Another benefit of this scenario is the ability to choose which residential communities to start with. There is no need to solve all problems from the beginning; instead, you can first find communities with suitable conditions, get the system running, and then gradually expand its scope of application.

Yang Ruigang: We initially thought that old residential communities without elevators and with the most difficult delivery conditions would be the ones that needed robots the most. However, after talking more with customers, the market taught us a lesson.

They suggested starting with high-end communities. On one hand, these communities may have a stronger willingness to pay; on the other hand, existing technology is easier to meet their environmental requirements. This feedback directly affects the product form. Originally, we prioritized stair-climbing capabilities, but now we look more seriously at which scenarios can be met with wheeled chassis.

We now have a clearer understanding of which functions customers truly need and which are temporarily unimportant, leading to an adjustment in the mass production schedule.

New Smart Driving Physical AI: Have you ever fully followed the workflow of a courier or food delivery rider? What is the hardest 100 meters?

Yang Ruigang: We have done this many times and talked with many couriers and food delivery riders. The thing they are most unwilling to do is go upstairs.

For couriers, it is best to deliver packages directly to the station, where sorting, shelving, and subsequent distribution are handled by others.

Food delivery riders face similar situations. For example, in Huaqiangbei, Shenzhen, some commercial buildings have people specifically helping riders take food upstairs, some of whom are retirees. During lunch and dinner peaks, they wait downstairs, charging about 1 yuan to deliver the food upstairs.

Riders calculate the costs carefully. If a single order takes 20 minutes, and the last 10 minutes are spent waiting for the elevator and going upstairs, the rider might earn only 6-7 yuan for that order. They would rather pay 1 yuan to someone else to complete the final leg, allowing them to use those 10 minutes to pick up another order. If they can earn an additional 3 yuan during that time, minus the 1 yuan paid out, they still make an extra 2 yuan. They certainly calculate this ledger more clearly than we do.

SF Express even has a vivid internal term for this: the "Upstairs Machine Project."

Yang Ruigang: I think the business models are quite different.

The demand for delivery services in hotels is limited, and the number of tasks a robot can perform daily is also finite. If the utilization rate is not high enough, the economic equation is hard to justify. Moreover, the technical threshold for hotel delivery is relatively low, mainly involving SLAM and path planning. As more participants enter, it easily leads to price competition.

However, the number of express and food delivery orders generated in a residential community every day far exceeds the delivery needs of a single hotel. Last-mile delivery involves larger task volumes and potentially higher usage frequencies.

Of course, this premise assumes that the technology can truly solve the problems.

New Smart Driving Physical AI: When robots enter residential communities, what other Corner Cases arise that were not anticipated in the lab?

Yang Ruigang: There are too many such cases. We encountered an interesting scenario.

A robot was on a very narrow road when it met a person. That person was polite and kept waving, signaling the robot to go first. But the robot did not understand the gesture and kept stopping there. The person also didn't move, continuing to wave. As a result, the person and the robot were stuck in an impasse.

Another overseas customer asked us if the robot could stand back up and continue executing tasks if kicked over. Especially for quadruped robots carrying large bags, if they are flipped over like a turtle with four legs in the air, it is actually difficult for them to restore their normal posture on their own.

Previously, colleagues working on shopping mall cleaning and security robots encountered a situation where four or five people lifted a robot weighing over 100 kilograms to another corner. Because that robot relied on SLAM navigation, losing its initialized position after being suddenly moved, it ultimately just stopped there.

We are also more concerned about safety issues. For example, if the mechanical structure loses power suddenly, the robot might automatically sink. If a child places their hand near the joints at this moment, there is a risk of pinching or crushing injuries.

Many problems are only discovered on-site once robots truly enter human living environments.

Yang Ruigang: ROI must still be considered.

For instance, opening fire doors is technically feasible, but some fire doors are heavy and require significant torque, meaning heavier and stronger robotic arms are needed. Once the robotic arm becomes heavier, the main body must also be reinforced. However, some community elevators are very small; if the robot is made too large, it will struggle to turn around once inside.

So the question is how much cost is added to open this door, and what new limitations it brings.

Sometimes, simply installing a door opener for 500 to 1,000 yuan can solve the problem. In comparison, modifying elevators involves more laws, regulations, and approval processes, making it much more complex.

Currently, the clear requirements from customers are: first, do not modify the elevator's control system; second, the robot must be able to pass through access controls. I feel that at this stage, there is no need to require robots to solve all Corner Cases before commercialization.

There is another major difference between last-mile delivery and Robotaxi: Robotaxi requires operation across a large continuous area for the commercial service to be viable; delivery robots can first select communities with suitable conditions, truly run successfully in one community to generate revenue, and then gradually scale up.

Yang Ruigang: Light mapping still uses maps. We use lightweight topological maps to provide basic spatial relationships, do not rely on high-definition maps, and combine this with real-time perception to advance the route.

This direction aims to reduce the burden of creating and maintaining high-definition maps. However, what initialization, facility configuration, and testing are required for new sites depend on the specific environment. Ultimately, judging its value comes down to how long it takes to prepare a new site and how many adjustments are needed after changes occur on-site.

The True Moat of Mo Robot Lies in System Closed-Loop

"New Smart Driving Physical AI": Many embodied intelligence companies are now working on end-to-end models, world models, or VLA. What is special about Nuwa's model architecture?

Yang Ruigang: Frankly speaking, I feel that at this stage, the differences in model architecture among various companies may not be as significant as imagined. The core capabilities that Nuwa has already verified mainly lie in simulators, data generation, and data augmentation.

What truly affects performance is how much data you have, what situations these data cover, and whether you know which data should be used to train the desired capabilities. There is a lot of experience involved; obtaining a model architecture does not mean you can accomplish the task.

When I was working on ApolloScape at Baidu, I realized that real-world collection could hardly cover all long-tail scenarios.

Yang Ruigang: For example, to collect fire trucks, we would station our collection vehicles near fire stations and follow them once they came out. But you cannot use this method to cover all situations. As time goes on, the proportion of truly valuable information in the collected data decreases.

So we considered synthetic data early on. Real data tells you what happens on-site, and then you need to consider whether you can increase the complexity and diversity of the data. This understanding has persisted until now.

"New Smart Driving Physical AI": What are the respective relationships between WorldWeaver, SimWeaver, and WTM in Nuwa Technology's core technical architecture?

Yang Ruigang: WorldWeaver is responsible for organizing real-scene data, 3D assets, and generated data, while SimWeaver provides physical simulation capabilities. This data and simulation system serves the training and evaluation of WTM.

We hope that problems discovered on-site can enter a repeatable testing process. For instance, if an interaction fails, and if the corresponding scenario and conditions can be reproduced in simulation, different handling methods can be repeatedly compared before returning to the real machine for verification. In this way, data becomes connected to product issues.

Yang Ruigang: We have conducted migration between quadruped robots of different brands. Since the two bodies are relatively similar, it was easier. Now we are also migrating the system to wheeled chassis.

Quadrupeds and the wheeled chassis we selected have different modes of movement. Some lateral movements possible with quadrupeds need to be reprocessed when switched to this type of wheeled chassis, requiring adjustments in trajectory tracking, control, and smoothing. The complexity of underlying control also differs.

Currently, the entire system basically runs on wheeled chassis.

Yang Ruigang: There is no clear commercial demand yet. At this stage, both quadrupeds and wheeled platforms have customer demands, so we still hope to prioritize doing things that can generate actual value as soon as possible.

We look at which tasks in human living environments require robots to move and execute. Capabilities extracted from community delivery can be applied to other tasks in the future, but they need to be verified according to specific scenarios.

Yang Ruigang: Definitely needed; we are currently customizing our own wheeled robots.

Because community delivery imposes very specific requirements on the robot body, you cannot simply take any off-the-shelf wheeled robot to complete the task.

For example, some customers explicitly stated that the robot must be able to climb slopes of 8 to 10 degrees and cross curbs of 5 centimeters. Five centimeters is actually not low.

There are also size requirements. Although some residential building lobbies have double doors, usually only one side is opened, so the actual passage width might be only 50 to 60 centimeters. If the robot is too wide, it won't fit.

But from the perspective of delivery efficiency, there is a desire for it to carry more items per trip. The larger the load capacity, the taller and heavier the robot becomes. However, if made too tall, it becomes prone to instability or even tipping over when going uphill. Therefore, body design is essentially a continuous trade-off process.

Yang Ruigang: I think there is no contradiction.

First, the market is large enough; robot companies like Deep Robotics cannot rely solely on us as their only customer.

Second, if partners are willing to use our "brain," we are also very willing to work with them to design robots better suited for specific application scenarios.

Many robot manufacturers themselves accept OEM or joint development cooperation models. We can be responsible for the brain and application capabilities, while they handle body manufacturing. Both parties can jointly develop products, forming a complementary relationship.

English text is being prepared.

We do not plan to hold a large number of robots ourselves for large-scale operations. Operations require strong professionalism and can cooperate with enterprises skilled in this area. I hope Nuwa's core capabilities remain focused on robotic intelligent systems and hardware-software development.

Yang Ruigang: At the end of September this year, we were invited by K2 to participate in LIVEX in Abu Dhabi, where we jointly showcased products with their subsidiary AutoGo. Currently, we are also advancing PoC cooperation with K2 in the UAE.

Next, we hope to clarify specific scenarios and tasks together with local partners. What roads will the robots travel on? Which facilities will they use? Who handles on-site problems? All these matters need to be addressed item by item. Attending exhibitions allows customers to see the products, but actual usage requires entering specific scenarios for verification.

We also want to see, in a different environment, which capabilities can be directly applied and which need adjustment. How much of the navigation, facility interaction, and task execution capabilities accumulated from community delivery in China can be reused overseas? After architectural and service processes change, what new work needs to be added? These questions will help us determine which are already product capabilities and which still rely on development for individual projects.

We do not plan to conduct large-scale operations ourselves; instead, we hope to focus our energy on intelligent systems and hardware-software products. Therefore, it is necessary to clarify the division of labor in overseas cooperation: how far Nuwa is responsible, what support local partners provide, and how to ensure that on-site issues are promptly fed back to R&D. First, solidify specific projects, then look at how to expand.

Yang Ruigang: We already have revenue, but it is mainly project-based and POC-stage revenue; scaled revenue has not yet been formed.

Yang Ruigang: Our team consists mainly of two parts. Currently, including full-time employees and interns, there are more than thirty people in total.

One part consists of young research talents, mostly students and research teams from Shanghai Jiao Tong University. The other part comprises individuals with mature engineering, product, and commercialization experience, many of whom are former colleagues of mine from Baidu and Inceptio.

University resources have been very helpful to us. Now, SJTU can provide excellent students and research talent, while Nuwa can provide more computing power, data, and real-world scenarios.

Yang Ruigang: The first institution clearly willing to invest was PNP.

We started contacting the financing market around November 2025, and received PNP's first TS in December.

The communication at that time was quite fast. After chatting with their partners about twice, things became quite clear. They have three partners in total, one of whom is primarily based in the US. I specifically went to meet him in January 2026, after which the process mainly involved due diligence and other procedures.

What they valued most was actually the completeness of the team. As for which specific direction to pursue, they felt that more exploration could be done, and it wasn't necessary to be completely certain from the start. Because at that time, we were indeed still a very early-stage team, with basically everything still on PowerPoint slides.

After the seed round, we quickly completed another angel investment round of 50 million yuan.

The second round progressed even faster. A person in charge at an investment firm was a former colleague of mine from Baidu who knew me well and understood our team's capabilities, so many things did not require repeated explanation. After one meeting, the other party even said that verbally, they could give the TS right now.

I think this also shows that early-stage financing is largely about investing in people and teams.

New Smart Driving Physical AI: Originally planned for mass production in 2028, why is it being significantly accelerated now?

Yang Ruigang: There are mainly two reasons.

On one hand, it is competition. When we first started doing last-mile delivery, it felt like few people were working in this direction, but now several companies have begun to layout.

On the other hand, it is market demand. As we contacted more customers, we found that actual demand was more urgent than originally anticipated.

From receiving the TS in the previous round to deciding to accelerate the next round of financing, there was only about three months. But during these three months, we achieved some very clear technical milestones, and the feedback obtained commercially was much clearer than before.

Which features customers truly need and which are temporarily unimportant, we now have a clearer judgment. So the mass production plan has been advanced by at least half a year compared to the original schedule, and the financing rhythm has accordingly accelerated.

Yang Ruigang: I believe cumulative financing must reach at least the hundred-million-yuan level to have sufficient funds to bring the product to a mass-production state that is truly meaningful and capable of sustainable, stable operation.

Doing POCs is relatively easy, but going from 10 units to 100, and then to 1,000, model training, data, and manufacturing will begin to consume massive amounts of capital. More importantly, truly enabling robots to operate stably in 100 communities is a completely different matter from making a few POC prototypes.

We are currently only operating in a few communities. There is at least one order of magnitude of growth needed to reach 100 communities. The real challenge is ensuring long-term stable operation across a large number of different real-world environments.

I often tell the team: if this thing were not difficult, it wouldn't fall to a team like ours to do it. There are many people outside who understand the industry, operations, and sales better than we do.

Our real opportunity lies precisely in the fact that some technical problems have not yet been solved. Because it is difficult, that is why people like us who do technology are needed.

We are not afraid of difficulty. Leifeng.com Leifeng.com Leifeng.com (WeChat Official Account: Leifeng.com) article_adlist [

Leifeng.com original articles, unauthorized reproduction is prohibited. See reprint notice for details. ]article_adlist Statement: Sina News exclusive manuscript, unauthorized reproduction is prohibited. tech_web_article_v2015_block_bottom Non-targeted 300*250 button 17/09 wenjing begin

Non-targeted 300*250 button end

Source

Sina Tech