news

after breaking boston dynamics' patent for somersaults, a chinese humanoid robot screws in with its left hand and draws coffee with its right hand

2024-09-15

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

the robotics company that broke boston dynamics' somersault patent with its "electric-driven robot" for the first time in the world is no longer hiding!

more skills and techniques of its humanoid robots were put on display.

in fact, it can also do some delicate work:make coffee.

it is also the world’s first humanoid robot that can make coffee art autonomously.

he was seen shaking his wrist holding the milk jug left and right, slowly and calmly:

then you will get:

then he left after his work was done, hiding his achievements and fame:

in addition to making high-precision autonomous latte art in coffee shops, it can also distribute midnight snacks to factory workers and serve as a guide in indoor and outdoor scenes.

even walk the dog.

let me formally introduce the company behind it.magiclab focuses on general robotics and ai technologies, and i heard it is taking a global approach.

humanoid robots are just one part of their product line. bionic quadruped robots, industrial quadruped robots, general humanoid robots, etc. are all part of the company's plan (and are being promoted).

the targeted application scenarios are also very broad, covering home, industry, commerce, etc.

it is understood that the practical application of 3c scenarios is under development and is expected to be put into trial operation on the production line by the end of the year.

performing performance testing of cleaning equipment within the production floor increased efficiency by 30% while reducing the need for manual intervention by 70%.

self-developed servo motors are also used in large models

the magiclab team focuses on core technologies of general robots such as robot core torque motor joints, robotic arms, and legged robots.

the team believes that the ultimate form of general-purpose robots must be humanoid robots.

from a technical perspective, the team mainly focuses on solving difficult problems in hardware and algorithms.

hardwarethe team chose to develop its own servo motors and core components to extend its capabilities.

the torque motor joint module design is committed to improving the power density ratio and torque density of the motor, including harmonic module joints, linear module joints, and small and micro servo units.

the team currently has a series of motors that cover a torque range from 10n.m to 500n.m.

previously, the electric-driven humanoid robot was able to perform a somersault thanks to magiclab’s new joint design.

the self-developed d190 joint module has a rated output torque of 150nm and a maximum output of over 525nm. it can operate at a speed of 110rpm. the high-power joint is combined with a driver specially designed for somersaults, which can cope with the instantaneous high and low voltages generated by the back-electromotive force fluctuations during somersaults.

the robot's mechanical structure design also adjusts the robot's weight distribution.most of the weight is concentrated in the thighs, reducing the inertia during leg swinging.

during jumps and flips, the robot can quickly adjust its center of gravity, stay stable and land successfully, while maintaining agility and precise control.

another benefit of self-research is that it allows for overall consideration of the coordinated optimization of software and hardware, using relatively low-cost hardware to bring out the available operational control capabilities, continuously iterating and implementing, and reducing costs.

algorithmicallythe team applied a multimodal large model. based on the general knowledge of the large model, it was able to cope with multi-task physical representation perception and understanding of objective physical facts in complex scenarios and form human-like behavioral decisions.

in terms of the whole-body control model, the small operation model based on reinforcement learning and the small operation model based on the skill library can fully decouple the environment and objects, the self-executing body, and the physical characteristics of the operation to make human-like smooth, robust, and universal whole-body control.

finally, through the graphical interface, combined with the above two large models, users can easily design and deploy robot applications to achieve multimodal human-computer interaction, scene task perception and autonomous decision-making.

from a specific example, the difficulty that the robot has to overcome is not only to be able to make simple latte art, the team saidlater, i also want to enable robots to replicate the skills of coffee masters.

it needs to interact with people in open spaces, which means there will be many uncertainties in its spatial position and movement trajectory, and it needs tools to recognize different materials and sizes.

as a core component in the humanoid industry chain, the dexterous hand is also the hardware that the team focuses on developing. in this direction, the team has adopted a strategy of generational iteration:

the mature generation, the pre-research generation, and the reserve generation.

sub-millimeter positioning accuracy and several kilograms of load capacity can now be achieved.

the basic version is designed with 11 degrees of freedom, of which 6 are active degrees of freedom. the thumb has two active degrees of freedom, which can be used for bending and sideways swing, allowing the dexterous hand to imitate various movements of the human hand, including clenching a fist, opening, pointing, pinching, and fiddling.

in addition, force control and fingertip tactile perception are introduced to achieve precise control through the body's torque sensor feedback. no matter what shape, size, or material the object is made of, even fragile, sliding, or deformed objects can be easily grasped.

the team then plans to introduce tactile perception and greater freedom of design.

defining the “m0-m4” of humanoid robots

after a conversation, the magiclab team's future plans and goals are still very clear.

in view of the current development stage and future prospects, magiclab divides the implementation level of humanoid robot scenarios into five stages: m0-m4.

let’s break down the criteria for dividing these five stages and discuss them separately.

first is m0

m0 is a laboratory product and is not targeted at specific scenarios.

from an r&d perspective, many people may pile up some software and hardware regardless of cost without any clear cost control. even if someone purchases it from a scientific research perspective, there is no way to truly implement it commercially.

arrived at m1at this stage, the humanoid robot is more like an intern, performing some local applications in specific commercial and industrial scenarios.

for example, let the humanoid robot pour a glass of water, dance, or make a cup of coffee with latte art.

from an r&d perspective, the m1 stage can already optimize products and costs based on specific scenarios, specific tasks, and other requirements; but its generalization is far from sufficient, so it cannot be called a truly practical product and can easily fall into commercialization difficulties.

m2 can be defined as a skilled worker.

products of this level need to be able to achieve a closed loop in specific commercial and industrial scenarios, which means they can truly do things for people.

purchasers can calculate the economic costs of using humanoid robot hardware and labor to determine commercial feasibility.

it was also during this stage that humanoid robots truly reached the doorstep of commercialization.

MagicLabthe definition of a humanoid robot at the m3 stage is "like a nanny"

at this time, humanoid robots can already enter homes and complete tasks such as companionship, housework, and elderly care.

"but we fully realize that before achieving this goal, we need to fully enhance the generalization capabilities of large models and hardware." magiclab r&d director zhou yang said that to reach this stage, the entire industry will have to spend a huge amount of time exploring.

the highest level is m4.magiclab calls it the siri of the physical world, a true universal robot.

one machine can be used for multiple purposes - how many uses it has depends entirely on the hardware level of the humanoid robot.

however, m4-level humanoid robots must be versatile and fully integrated into society, and perhaps they can go out to work and earn money in their spare time (???).

the magiclab team told us that the long-term goal is to produce m3 to m4 level products.

currently, the company's humanoid robots are in the transition stage of evolution from m1 to m2, and at the same time, they are working hard to promote the implementation of the m2 level.

bringing robots out of the lab

magiclab was established in mid-december last year.

so far, it has grown into a team of more than 100 people. the members come from all over the world, and more than 80% of them are r&d personnel.

the team focuses on robotscore torque motor joints, robotic arms, and legged robotsand other general robot core technologies, and some achievements have been made in the algorithm level, such asoperation and control algorithms, navigation algorithms, vision and ai algorithmswait.

at the same time, it has the ability to independently develop and produce robots, the ability to implement in multiple scenarios, and the ability to mass produce and deliver.

in the communication with magiclab r&d director zhou yang and chief strategy officer ivan, quantum位 learned that the team is wholeheartedly focused on one thing -

efforts are being made to enable robots to truly “replace humans” and achieve closed-loop implementation in specific business scenarios.

when we asked about the team's strengths, zhou yang also said:

in addition to deepening our technological research, we also value the power of technology and business, constantly exploring the application scenarios of technology and engineering our products.

we believe that with the mature mass production of universal bodies such as dexterous hands and humanoid robots, universal humanoid robots will also be able to obtain positive feedback in terms of commercial value.

the entire team very much hopes to combine robotics technology with industry to exponentially increase the value of robots.

in addition, because magiclab feels that being a mere robotics equipment company would be too limited in its imagination, they use “robot+” to position themselves.

just like the internet+ back then, it brings technology to thousands of industries and solves practical problems.