OmniHead is a head module developed by VMR specifically for embodied intelligence, equipped with a high-precision RGB-D vision system and a large AI model. By integrating OmniHead, developers can easily enhance the robot’s visual perception capability and environmental robustness, thus eliminating the need for remote control.
On August 18th, VMR officially launched OmniHead (“Aoshi Zhinao”), the world’s first head module for humanoid robots.
Currently, humanoid robots generally face the pain points of weak environmental perception ability and decision-making relying on preset logic in practical applications, which makes it difficult to get rid of manual remote control and restricts the expansion of practical application scenarios. The launch of OmniHead will effectively solve these technical bottlenecks.
OmniHead is a head module developed by VMR specifically for embodied intelligence, equipped with a high-precision RGB-D vision system and a large AI model. By integrating OmniHead, developers can easily enhance the robot’s visual perception capability and environmental robustness, thus eliminating the need for remote control.
01 Visual Panoramic Perception
The RGB-D depth vision system of OmniHead consists of three groups of ten RGB-D array cameras, distributed at the front, left rear, and right rear of the head module, forming a 360°x90° panoramic field of view, enabling the robot to locate, perceive, and navigate autonomously in complex scenes.
Exclusive Visual Positioning Technology
OmniHead adopts exclusive visual positioning technology, which can obtain real-time 3D information of the surrounding environment through RGB-D array cameras and quickly build maps, providing reliable data for robot positioning and navigation.
Multimodal Perception and Semantic Recognition
The RGB-D camera of OmniHead integrates a semantic perception algorithm to achieve accurate recognition of environmental object categories. After training, it can recognize various objects such as tables, chairs, humans, robots, and dogs, providing an accurate basis for robot behavior decision-making.
Multi-machine Collaborative Data Synchronization
The three groups of RGB-D array cameras can realize synchronous collection and transmission of visual data, and finally achieve panoramic 3D reconstruction and real-time perception.
02 Large AI Model + Microphone Array
It is deeply integrated with DeepSeek or Tongyi Qianwen large models and a circular microphone array, supporting 360° sound source localization and natural semantic understanding, and realizing multi-turn dialogue and emotional response.
The circular microphone array with 6 microphones works together for noise reduction, supports 5-meter far-field sound pickup, and is equipped with a voiceprint recognition algorithm to achieve sound source separation and voice enhancement in multi-person dialogue scenarios, and supports Chinese-English bilingual communication.
03 Open Interface for Multimodal Data Fusion
OmniHead has rich open interfaces, supports joint reasoning of visual and voice multimodal data, is compatible with various humanoid robot platforms, and can expand the application of embodied intelligence in multiple scenarios, including commercial office, intelligent manufacturing, smart home, medical care, and other fields.
