Multimodal AI
Multimodal AI that combines spatial, visual, language and voice understanding and turns it into robot action.
Multimodal AI That Unifies Space, Vision, Language and Voice
Multimodal AI understands both the operator's command and the work environment
and translates them into real robot actions.
By processing spatial, visual, language and voice information together,
it lets the robot grasp both the intent of a task and the conditions around it.
Key Features of Multimodal AI
01Semantic Grasping
Semantic grasping lets the robot identify and pick objects by understanding not only their shape but also the context of the task. By combining spatial perception with VLM (Vision-Language Model) image recognition, it can distinguish items down to a specific brand logo or individual product, then generate grasp coordinates and a motion path for the target object.
02Voice and Natural-Language Interface
The system recognizes the operator's voice, interprets natural language and generates task commands for the robot. With an LLM-based agent control architecture, robots can be operated in plain language, without complex system control.
03SKU-Level Recognition and Operation
Even in logistics environments with a wide mix of products, automation can run on SKU (Stock Keeping Unit)-level recognition, and on-site workers get an operating environment in which they can control the robot themselves.
Key Specifications
Detailed specifications for technical review.
| Perception modalities | Integrated processing of spatial, vision, language and voice inputs |
|---|---|
| Visual recognition | VLM (Vision-Language Model)-based object recognition (distinguishes brand logos and individual products) |
| Identification level | SKU (Stock Keeping Unit)-level recognition |
| Language input | Natural language via voice or text |
| Agent architecture | LLM-based agent control (generates robot actions suited to the task goal) |
| Task output | Grasp coordinates and motion path for the target object |
| Operation | Natural-language robot operation without programming |
| Application environment | Logistics and industrial sites with a high mix of products |
Ongoing Collaboration
Our multimodal AI robot perception technology is being advanced with government R&D support.
Selected for 2025 Deep Tech TIPS
Selected for Korea's Ministry of SMEs and Startups' Deep Tech TIPS program, we are advancing our multimodal AI technology with approximately KRW 2 billion in R&D funding.
Looking for the right product or solution?
Request a Demo & Site Assessment