Innovation Endorsed: Landmark Patent for Physical AI Granted by U.S. Patent and Trademark Office (USPTO)
- Ting-Yuan Wang
- 2 days ago
- 2 min read
Updated: 4 hours ago
We are excited to share that the patent for our underpinning technology, which bridges machine vision to Physical AI, enabling robotic agents to identify and understand real-world objects via digital twins, has been granted by USPTO.
The Invention
This patent, "Machine-learning Method on Vectorized Three-Dimensional Model and Learning System Thereof", describes a novel method for utilizing vectorized 3D models to generate synthetic data of real-world objects for AI training. It can be applied for various applications, including object identification and anomaly detection, where training data are difficult or economically infeasible to acquire, and the robotic agents need these capabilities to function and swiftly interact with the physical world.
Specifically, this patent recognizes our novel approach of:
Scanning physical objects using stereoscopic imaging
Constructing vectorized 3D representations
Generating diversified synthetic images for AI training and validation
Note that this patent was filed before Nvidia made the concept of Digital Twins popular with embodied applications not in theory, but in practice.
Enabling the Upcoming Physical AI
Imagine when a robot sees a cup and knows exactly its purpose (holding liquid), weight, gripping area (handle), and fragility (maximal tactile force), instead of cautiously examining, feeling, and interacting with a cylindrical, hallow object to figure out what to do. This sounds intuitive but in fact nontrivial because we have not had a World Model available.
Why This Matters Now?
In current industries like automotive remanufacturing, PCB manufacturing, and solar farm operations, automating object inspection and identification with AI is avidly sought after. However, data scarcity and inconsistency have been the biggest bottlenecks to economical AI adoptions.
It has been proven that using our technology to turn physical inventory into standardized, reusable digital assets can enable a wide spectrum of applications through an Identification-as-a-Service (IDaaS) platform, even across their fragmented vertical ecosystems where data chains for uniform product IDs are often broken or nonexistent, and automated visual identification become the only reliable method.
Let Us Work Together
Having applications that may utilize this invention? We are seeking to collaborate with partners across industries to solve real-world challenges. If you’re exploring AI-powered identification, digital twins, or synthetic data pipelines, let’s connect.