This Four-Legged Robot Inspects Factories Using AI, No Cloud Connection Needed
Avnet and Weston Robot have built a dog-like inspection robot that thinks for itself on-site, spots problems in real time, and works even when the internet is not available.
Key points
- Avnet and Singapore-based Weston Robot announced a joint autonomous inspection platform in 2025 built around a four-legged, dog-shaped robot.
- The system delivers up to 50 TOPS (trillion operations per second) of AI processing power directly on the device, without needing a cloud connection.
- It uses AMD Ryzen AI Embedded processors to run thermal imaging, visual sensors, and 3D lidar mapping at the same time.
- Weston Robot sells access on a subscription basis, meaning companies pay a recurring fee rather than buying the hardware outright.
- Edge AI processing, where the AI thinks on the device itself rather than sending data to a remote server, lets the robot flag problems faster than traditional manual inspections.
Imagine a dog-sized robot trotting around a factory at 2 a.m., scanning equipment for overheating parts, flagging a leaking pipe before the morning shift arrives. That's the pitch from tech distributor Avnet and Singapore-based robotics company Weston Robot.
The two companies have built an autonomous inspection platform using a quadruped robot, which is simply a robot that walks on four legs. Avnet fitted it with a custom backpack of its own computing hardware, giving the machine enough brainpower to handle serious AI tasks on the spot.
What does it actually do?
The robot patrols a pre-set route, stops at inspection points, and uses onboard cameras and sensors to check for problems. It handles thermal imaging (detecting heat), visual analysis, and 3D lidar mapping, a technique that uses laser pulses to build a detailed three-dimensional picture of the surroundings.
All of that processing happens on the robot itself. Engineers call this edge AI: the artificial intelligence runs on the device rather than uploading data to a remote server and waiting for a reply. Decisions arrive faster, and the robot stays operational in GPS-denied or low-connectivity areas like underground facilities or heavily shielded buildings.
The platform runs on AMD Ryzen AI Embedded processors and delivers up to 50 TOPS of AI performance. TOPS stands for trillion operations per second, a measure of how much number-crunching a chip can handle. That's enough to run multiple AI vision tasks at once without slowing down.
New inspection and recognition tasks can also be trained and pushed to the robot after deployment, so it isn't locked to a fixed set of checks from day one. Our 3 August story on Reimagine Robotics showed a similar drive to make on-the-job robot retraining faster and more accessible.
Who is this for, and what does it cost?
Weston Robot sells access through a robots-as-a-service model, where customers pay a recurring fee instead of buying the hardware outright. Think of it like leasing a clever guard dog. The company, first reported on by The Robot Report, works with industrial clients across inspection and facility management.
No public per-month price is listed yet. The subscription model does lower the barrier compared to buying industrial robotics hardware outright, which can run into tens of thousands of dollars.
| Feature | Detail |
|---|---|
| Robot type | Quadruped (four-legged) |
| AI performance | Up to 50 TOPS |
| Processor | AMD Ryzen AI Embedded |
| Sensors | Thermal, visual, 3D lidar |
| Connectivity needed | None (edge AI, works offline) |
| Pricing model | Robots-as-a-service (subscription) |
Should workers be worried?
Not immediately. The robot handles dull, repetitive patrol work that currently ties up human staff or gets skipped on night shifts. Weston Robot frames it as freeing workers from dangerous or tedious inspections, not replacing skilled roles.
One thing worth your attention: the robot captures continuous thermal and visual data inside facilities. Any company adopting this needs a clear policy on where that footage is stored, who can access it, and how long it's kept. That conversation is easy to skip when the technology feels new and impressive, and it shouldn't be.
Common questions
Does the robot need Wi-Fi to work?
No. The edge AI setup means it processes everything on its own hardware. It operates in GPS-denied or low-connectivity environments without sending data to the cloud.
Can it learn new inspection tasks after it is deployed?
Yes. The Avnet computing architecture lets operators train and push new recognition tasks to the robot over time.
How is this different from a standard security camera system?
A fixed camera watches one spot. This robot moves through a space, uses multiple sensor types at once, and can actively investigate something it flags as unusual, which a static camera can't do.
