Synergy between visual perception and cognitive perception
Co-driving AI to accelerate development
When robots actually enter homes, factories, warehouses and public spaces, the productivity of physical intelligence will be further released. It will not only identify objects, understand the environment, plan paths, but also touch the world in the moment, perceive forces, judge friction, control strength, and correct movements in a timely manner.
Vision enables robots to see the world; cognition enables robots to feel the world. Only when the two work together can an intelligent body form a perceptual decision-making mechanism that is closer to humans.
02
I'm not sure what I'm going to do.
Adding a key touch to robots
In 2019,Huawei has pioneered six-dimensional sensing technology based on opticsIt is the world's first multi-dimensional force/torque sensing technology based on the optical principle, and it is also a fundamental reason for the simultaneous breakthrough of microscopy, high performance, high reliability and automated mass production.
Based on pioneering optical six-dimensional sensing technology, Huawei has pioneered the Photon-Finger series of product matrices tailored to smart fingertips.The world's smallest optical multi-dimensional power sensor with a diameter of only 8 mm and a thickness of only 4.5 mmFor the first time, the fingertips of the robot are also equipped with a stable, reliable, resistant to interference, optical multi-dimensional sensors that can be delivered in bulk.
Fingerprinted multi-dimensional sensing software developed by Huawei Creative Sciences, covering from 0.01N/20N (resolution/measurement, below)It's a great way to get to know each other.It's about 0.5N/100N.Common household dishesIt's going to be 1.5N/300N.Heavy-loaded aluminum from the aluminium industrythe use of fingerprints; andFor the first time, the integrated sensing-calculating-dissolving process was implemented inside the fingers.The program has passed rigorous verification of scenarios such as hands-on, multi-finger electrical claws and achieved mass delivery with the help of the automated production/labelling/testing production line built by Huawei Creative Sciences.
The continuous investment in original sensing principles, program feasibility, engineering reliability and large-scale production is also a technical consideration of the EGO-UMI program for intelligent visual, auditory, tactile and multi-module data collection.
03
From a mother to a mother to a mother to a mother
Joint imaging to build visual infrastructure
With 12 years of visual engineering accumulation, Joint Pictures has built a multi-modal data collection infrastructure from front-end to embedded intelligent visual perception software to build full-chain vision capabilities for embedded intelligence. Around robots and embedded intelligence, we provide visual hardware, multi-sensor integration, synchronization, edge processing, embedded visual data collection and scalable delivery capabilities.
From RGB, binoculars, structural light, iToF to multi-camera synchronization, from visual acquisition, 3D reconstruction to target identification, spatial understanding and end-to-end decision-making, combined imaging with Vision for all to create the visual infrastructure of the physical AI world.
04
Seeing and feeling
Let's create a complete circle of physical intelligence
On August 18, Five Eyes officially unveiled its Intelligent Data Base at the Physical AI Magnificent Blueprint 2030 conference. As a jointly-owned subsidiary of Five Eyes and United Pictures, AperData is committed to building a neutral, open, soft-source data collection terminal + data governance engine + open data interfaces.
In the field of real-time, precision-sensing acquisition, Huawei Creative Science is a reliable supplier partner of shared imaging capabilities. The combination of both capabilities is not a simple superimposition of two types of sensors, but a more complete multi-modal perceptual system built around real-world tasks.
(Left one: Head of Joint Pictures Products; right one: Dr. Fishchen, CEO of Huawei Creative Sciences)
• At the level of environmental perception: vision is responsible for identifying targets, judging positions, understanding space and recording behaviour; cognition complements information such as positive pressure, lateral pressure, friction and slip resistance when collecting contact.
• In the action execution layer: hand-eye coordination combined with real-time multi-dimensional power acquisition, so that every detail of position, gesture, technique, intensity is accurately measured and recorded at the same frequency as the number of people or staff during each hand operation in the task of grabbing, assembling, grinding, moving, etc.
• At the product engineering level: cross-sectional research and development, synchronous labeling, edge computing and mass production delivery capabilities, in conjunction with high-performance sensing capabilities of Huawei science, driving the program from prototype verification to stable products.
For example, when a person picks up a glass, the ego vision system first identifies the glass body, estimates the distance and records the grabbing gesture; after contact, the multi-dimensional sensors in the umi perceive the actual grip vector and sliding trend; the system then identifies and cognizes each moment's behavior, thus creating a high-quality data set required by physical intelligence.
05
Strong and united
Shared acceleration of the landing of physical intelligence
Competition with physical intelligence is ultimately not just a competition between models and algorithms, but also a comprehensive competition between sensors, systems engineering, supply chains and scale manufacturing. Every breakthrough in underlying technology, every collaboration in industrial chain capabilities, brings robots one step closer to the real world.
In the future, we look forward to continuing to deepen our collaboration with Huawei Creative Sciences: expanding the perceptual boundaries of machines visually, enhancing the accuracy of machine interaction, shortening the distance from innovation to products in systems engineering and scale manufacturing, and accelerating the acceleration of physical intelligence in robots, smart manufacturing and more real-world scenarios.
It's about making machines see the world and making machines feel the world.
* This article is translated from Chinese Creative Science
—END—
Without filming, without performing, the company has maximized the practicality of industry-class quad-footer robots to the point that 10 companies such as Walking Robot/Synchronous Robot/Fao Robot/Shanghai Weiwei have won the award for pioneering China's smart welding industry - Industry Innovation Award.