UrbanMatrix Technologies Showcases at WAIC: From Navigation to Data Collection, Completing the Critical Link for Embodied AI Implementation

In July 2026, at the WAIC in Shanghai.

Stepping into the Embodied Intelligence Exhibition Zone, a subtle atmosphere mixed with anxiety and fanaticism permeated the entire pavilion. At the WAIC two years ago, only 18 enterprises ventured into this field; last year, the number had soared to 80; and this year, the number of exhibitors has quietly exceeded 200.

Behind this growth rate faster than Moore's Law, not only are various end players shining brilliantly; as the architectural dividends of model algorithms hit a phased bottleneck, data Infra supply chain players that were once hidden deep underwater, represented by Aegis Technology, have also begun to step into the spotlight.

I. How Data Became the Achilles' Heel of Embodied Intelligence

To understand the importance of data to embodied intelligence, autonomous driving is the best mirror.

If we break down embodied intelligence into different segments, autonomous driving is actually just a specific sub-task within the embodied intelligence landscape. Yet even giants like Tesla and a host of Chinese new-energy vehicle startups, despite millions of user data points continuously uploaded to the cloud and nearly a decade of relentless focus, still tread carefully when facing all kinds of bizarre corner cases.

The scenarios faced by embodied intelligence are more than an order of magnitude more complex than those of autonomous driving:

The first layer is the geometric dimensional upgrade from 2D to 3D: cars only need to handle movement on the X-Y plane, while the hand-eye coordination of robots operates in a fully open three-dimensional space.

The second layer is the exponential explosion of degrees of freedom: cars only have a steering wheel, accelerator and brake, while humanoid robots often have dozens of degrees of freedom (DoF), which means the range of possible movements increases exponentially.

The third layer is severe data asymmetry: compared with autonomous driving, the number of robots that can provide high-quality data feedback is pitifully small.

More critically, embodied intelligence does not require ordinary video data, but embodied experiences with spatial structure and task logic.

"There are massive amounts of videos on the internet, but for embodied intelligence, videos without reliable pose data are just useless junk materials," a relevant expert from Aegis Technology stated bluntly. "2D maps only need to describe a plane, but robots need to know their absolute position, posture, motion trajectory and task status in three-dimensional space. Only data with stable trajectories, time synchronization, spatial alignment and quality assessment can be regarded as trainable silicon-based experiences. And these require data such as SLAM/VIO, robust 6DoF trajectories, real-time online output and trajectory scoring."

Over the past year or more, Aegis Technology has precisely identified this data desert, and with underlying components such as the Insight spatial intelligent camera and the TinyNav navigation algorithm library, it has secured partnerships with a number of leading large enterprises.

But this year, Aegis Technology found that the industry's pain points are no longer limited to a single piece of hardware or navigation algorithm. A scalable, verifiable and reusable data closed-loop infrastructure is the foundation that enables all kinds of embodied intelligent terminals to evolve rapidly.

That is why, at this year's WAIC booth, this company positioned as a physical AI infrastructure provider launched two representative products: Looper Embodied Intelligence Data Acquisition System and NavCore Embodied Intelligence Autonomous Navigation Brain.

They are precisely targeting the two major problems encountered in the data dimension of the embodied intelligence industry: how to achieve large-scale mass production of high-quality data in the real world, and how to realize highly stable environmental perception and autonomous navigation on the robot's own end.

II. Looper Data Acquisition System: Equipping Robots with First-Person Learning Capabilities

On the battlefield of embodied intelligence data acquisition, there are currently three mainstream technical routes: teleoperation + VR, wearable multi-view, and fixed multi-camera studio.

The biggest advantage of teleoperation is that it can directly obtain robot motion data, which is suitable for training robot execution strategies. However, it usually relies on special equipment, with high acquisition costs and limited efficiency. At the same time, the data distribution is easily restricted by the robot's own form and operation mode.

Fixed multi-camera acquisition has strong controllability and can obtain high-quality, multi-view data, but it relies on specific sites and equipment, making it difficult to cover a large number of real tasks and long-tail scenarios in homes, factories and commercial environments.

Considering the above background, Looper chose the Ego-centric multi-view solution. Its core concept is: let humans complete tasks from a robot's perspective, and convert the physical experience accumulated by humans over tens of thousands of years into digital assets that robots can understand, in an imperceptible and low-cost manner.

In the industry's view, this perception method that is closest to how humans complete tasks is also the most promising route to achieve large-scale real-scene data at a relatively low cost.

In terms of configuration, Looper adopts the golden combination multi-view solution of head + both hands: the head-mounted Insight camera is responsible for recording first-person visual information, spatial environment and its own motion trajectory; the hand cameras are responsible for capturing close-up operation details, including hand movements, object interactions and local spatial changes; the backpack end provides edge computing, synchronous recording and data storage capabilities.

But this is by no means simply strapping several cameras to the body to record videos. "It's easy to collect data, but collecting valid data is the real barrier. What Looper does is collect valid data and high-quality data," Yan Qinrui, Co-Founder and Chief Operating Officer (COO) of Aegis Technology, gave us an example: during the data acquisition process, spatial and temporal alignment between multiple cameras is a very critical industry difficulty. If the images, depth information, inertial data (IMU) and pose trajectories from different perspectives cannot maintain high-frequency synchronization in the same spatiotemporal coordinate system, the trained robots will suffer from movement incoordination.

To solve this problem, Aegis Technology chose a technical solution that directly completes the 6-degree-of-freedom trajectory calculation of VSLAM and outputs it on the end side.

In the past, the common industry practice for embodied intelligence data acquisition was to collect data for a whole day, take it back and run 3D reconstruction offline on a server for several days, only to find that the data alignment failed and the entire batch was scrapped. With Looper's direct end-side output of VSLAM, the spatial position of each frame can be known in real time on site, and unqualified data can be re-tested immediately.

Another major advantage of this method is that the images, IMU and 6DoF (six degrees of freedom) trajectories are naturally aligned on the end side, and the subsequent costs of 3D reconstruction and task segmentation drop sharply.

During the final retrieval, what customers get is not a bunch of disordered video files, but high-quality experience packs that can be accurately retrieved according to spatial regions, motion trajectories and task stages, truly transforming the original videos into learnable and verifiable data.

It is reported that Aegis Technology is also planning to launch a new generation of four-camera large-field-of-view data acquisition headband, with the core optimization direction of lightweight wearing and unrestricted natural collection. The new solution will restore human hierarchical perception logic: using peripheral vision to take into account the global environment, and the line of sight to lock the interactive objects in front of you, completing the upgrade of the robot data acquisition solution from the underlying perception level.

III. NavCore Navigation Brain: How to Enable Robots to Move from "Being Able to Walk" to "Knowing How to Walk"

If Looper solves the problem of data production, then what NavCore needs to solve is the autonomous navigation problem of robots.

In the past, embodied intelligence companies lived like high-end system integrators when it came to navigation:

Buy cameras from Manufacturer A, select computing boards from Manufacturer B, purchase positioning algorithms from Team C, and then the R&D team stays up all night writing drivers, calibrating sensors, and aligning timestamps.

As a result, trivial and tedious tasks such as frequent disconnections of wireless networks during on-site deployment, uneven power supply of multi-board cards, and interface adaptation often consume a lot of the R&D team's time.

In addition, throughout the entire link, a tiny delay in the camera will amplify the positioning error, the positioning error will directly contaminate the global map, and a slow map update will lead to path planning that hits walls. Any jitter in any link may directly cause the task to fail.

Not to mention, compared with the past 2D navigation solutions, the data volume of 3D navigation at the same resolution will increase by more than an order of magnitude, and it will also increase the computing pressure of point cloud processing, collision detection and path planning.

The difficulty of software algorithms has escalated, but the conditions faced by hardware have become more stringent: in the past, 2D-based autonomous driving could be deployed on a spacious vehicle, but embodied intelligence requires deploying 3D navigation capabilities on low-cost embedded platforms.

Facing the above problems, as the autonomous navigation brain for embodied intelligence, NavCore's approach is: large-scale integration and standardization. It can encapsulate complex robot navigation engineering into deployable products, so that customers do not need to build perception, computing, mapping, positioning and navigation links from scratch.

In terms of hardware, it uses the NVIDIA Jetson series as the computing power base, deeply integrating into the world's mainstream developer ecosystem; at the same time, it modularly and integrally designs its self-developed Insight spatial intelligent camera, computing, communication, and power management to reduce integration complexity.

What is particularly worth mentioning here is the Insight spatial intelligent camera self-developed by Aegis Technology. Traditional cameras can solve 90% of the problem as long as they can capture clear images from a fixed position. But in embodied intelligence scenarios, it not only requires high image quality, but also a sufficiently large field of view, extremely low perception delay, reliable sensor synchronization, and the ability to continuously output stable position and posture under conditions such as vibration, sharp turns, low light, strong light and dynamic occlusion.

This requirement determines that the lens, sensors, inertial measurement units, chips and algorithms cannot be designed independently. To this end, Insight adopts a 188-degree ultra-wide field of view, an industrial-grade inertial measurement unit and end-side computing power, completing visual positioning, depth calculation and spatial perception locally on the device, so that embodied intelligence can not only see, but also accurately understand complex spatial relationships.

In the final software algorithm, Aegis Technology also uses algorithm pruning, computing optimization, hardware acceleration and model quantization compression to embed spatial intelligence, which originally required server-level computing power, into the low-cost robot end side, enabling it to run stably for a long time.

In addition, considering that in the past many robot projects needed to re-implement a complete set of system integration for different bodies and different scenarios, including sensor selection, computing power adaptation, driver development, coordinate system calibration, algorithm deployment and on-site debugging, etc. This has led to long robot development cycles and difficult-to-replicate solutions, which is also an important bottleneck for the large-scale implementation of the industry.

NavCore not only has Aegis's hardcore TinyNav visual navigation algorithm built-in, but also can provide standardized interface output based on ROS2. Whether it is humanoid robots, quadruped robots, AMRs or drones, or scenarios such as full-domain spatial perception and multi-robot scheduling in smart parks, centimeter-level autonomous inspection and dynamic obstacle avoidance for industrial inspection, real-time navigation and task execution in complex port logistics environments, and intelligent path guidance and immersive explanation for cultural tourism tours, it can be quickly migrated and used out of the box.

By reconstructing the entire navigation link at the underlying level, NavCore outputs the most complex and error-prone integration work of the entire industry through standardized interfaces. The R&D teams of embodied intelligence can naturally release their limited energy to the upper-level task intelligence and application scenarios, accelerating innovation.

IV. Epilogue

As scenarios move from simple demos with fixed trajectories to real scenarios full of uncertainties, embodied intelligence is ushering in its coming-of-age ceremony.

In the next three to five years, whether the invisible links such as data closed loop, perception navigation and body control can be solidly implemented will become the new industry survival standard.

The two answers Aegis Technology presented at WAIC: Looper aims to encode the labor experience that humans have accumulated in the physical world for tens of thousands of years into the digital world in a low-cost and large-scale manner; NavCore draws a definite travel route for robots in the continuity and irreversibility of the physical world, which may be the key.

Only when infinite human experience begins to continuously feed models through standardized pipelines, and the engineering threshold of autonomous navigation is reduced to out-of-the-box usability, can embodied intelligence truly take the key step from expensive demos to general labor.