New LIMS with Improved Performance! LIMS3-AMBIDEX

LIMS3-AMBIDEX
[Improved Properties]
- Increased payload
- Exceptional resilience to the impact
- Robust and simple mechanism aiming at commercialization
[Specifications]
Payload: 5 kg in the entire workspace
Torque: 4 times higher torque than LIMS2-AMBIDEX
Cont. torque: 42Nm, Peak torque: 119Nm
Total mass: 9.1 kg per arm
Moving part mass (under shoulder): 4.50 kg
Impact resilience: collisions with objects greater than 5 kg at 5 m/s.
% Supported by NAVER LABS

[@user-kv5qx1zu8w]

로봇 친구와 인간 친구 모두 잘생겼어요 Both robot friends and human friends are handsome 👍

[@MechTools]

This is amazing.....keep up the good work team 👌

[@MelloCello7]

Incredible incredible incredible work! 💖

[@yermekfun8203]

this. is. sooooooo coooool!!!!

[@sanches2]

this is amazing!

[@dannygarcia407]

Great work!

[@naon4588]

私が今まで見た中で一番美しいロボット、これは参考にしたい

[@xaytana]

LIMS has been such a long series of projects, someone should put together a video going through an in-depth design history with the benefits and downfalls of each design iteration, and with current plans for future iterations. I'd also be curious if LIMS will eventually adopt the ABENICS joint, it'd be useful for the wrist, and if scalable useful for the shoulder as well. A wire-driven ABENICS would also be interesting to see. Though I'm not sure of the payload capabilities of the joint or where the weaker link is, the motors, the drive coupling, or the sphere coupling.

[@IRIMLabKoreatech]

We think that ABENICS design is a very interesting mechanism. It is a brilliant idea to combine unique gears to implement 3-DOF. By the way, our goal is to minimize backlash and friction for highly dynamic force and impedance control as well as high precision manipulation. Thus, we don't prefer friction-causing mechanisms which have sliding components or transmissions that have backlash between the meshing. If we can find a way to implement the ABENICS mechanism with low backlash and low friction, it will definitely be worth a try.

[@xaytana]

@@IRIMLabKoreatech I would imagine the original design of the ABENICS is low backlash already, for accuracy purposes; if not, a low backlash variant can be developed, just as there's low backlash standard gears. As for low friction, either high lubricity materials or an enclosed gearbox with a lubricant should suffice. I presume the original design used worm gears for a smaller package, holding torque as they're not backdrivable, and speed reduction on higher speed motors. I also think in this use case the worm gears wouldn't be needed, which means simplicity and neater cable routing in a cable actuator setup, not to mention the potential of backdrivability but I have no idea if the system is backdrivable based on how the rotating monopole gears operate. If you have access to a resin 3D printer, for dimensional and surface accuracy, this could be rapidly prototyped as a proof of concept to test the worm-less design. I'm also curious how well these joints operate at different scalse, what their joint strength is (how prone the non-standard gears are to sheering/breaking), holding strength, etc., and I'm also curious if the parameters scale linearly or not; as in if you scale from the original model to a wrist/ankle unit, to a shoulder/hip unit, and to a simple single pivot torso movement unit. I'm also curious about efficiency at scale compared to typical joint structures, especially shoulder/hip units. If these ABENICS-like mechanisms scale well, they could have so many uses in the wider field of robotics.

[@IRIMLabKoreatech]

@@xaytana We agree that the ABENICS-like mechanism has clear advantages and proper application. We think that it can be compared with other gears and transmissions such as harmonic drives, cycloid gears, and various types of gear transmission, and eventually, it has to compete with them. We vaguely imagine that the backlash and friction of it may be rather similar to gears, which might be insufficient for some precise tasks or tasks with impacts. We hope ABENICS will continuously evolve, and we will be glad to watch it.

[@trimagos]

It look so smooth

[@huanovai]

This is exciting! :)

[@user-uk9ez3yc8z]

모두들 고생많으셨습니다

[@leonardochiruzzi7642]

it's still "fantastic", even for this reload !!

[@hoodio]

the wire pulley system is just beautiful to look at

[@IRIMLabKoreatech]

To simplify the mechanism but maintain the high stiffness of the joints, we used thick steel cables. Designing a multi-DOF mechanism with thick cables gives a lot of constraints. We put a lot of effort into maintaining a sufficiently big pulley/cable Dia ratio and minimizing the frictions.

[@hoodio]

@@IRIMLabKoreatech huge congrats, it looks like a very capable system🦾

[@uthvfyrekbnm6008]

@@IRIMLabKoreatechWhy you dont use dyneema rope? It used in many cable driven robots.

[@IRIMLabKoreatech]

@@uthvfyrekbnm6008 There are many good polymer wires such as Dyneema, Spectra, Vectran, and Kevlar. They have great flexibility and a small wire-pulley diameter ratio. But, some of them have nonlinear stiffness and creep, and they are easy to wear out. Thus, we used steel cables because of the linearity and predictable elongation, and long lifetime.

[@uthvfyrekbnm6008]

@@IRIMLabKoreatech Which structure of cable you use? 7*49 or something another.

[@jbrownson]

Wow

[@ulforcemegamon3094]

Is Nice to see such improvement ! 3kg was a load that was too small , sure 5kg isn't that big either but is quite the improvement

[@IRIMLabKoreatech]

We still have a long way to go. But, we are satisfied with this torque and payload improvement for a moment. Payload 5 kg in the entire workspace is not that easy for robot arms. Many cobots with the payload 3~5 kg weigh 17~25 kg. At least LIMS3 is better than me. I cannot keep stretching my arm forward with a 5 kg weight for a long time:)

[@ragnarok700]

@@IRIMLabKoreatech 5 KG is a lot for many tasks where such robotic arms would be great for taking over repetitive actions that need high precision! I also wonder, how is the repeatability of motions with a 5 KG payload vs without a payload?

[@vietlong5957]

Good work, the motion is very like to a human. But what are the pros and cons compared to the industrial articulated robot arm that uses harmonic gear? Like cost-effective, performance?

[@adamfilip]

hope to one days your design in a robot thats filling my dishwasher lol ;)

[@BoyFromMa]

🤯

[@challacustica9049]

Great to see such substantial refinement in the design. What are you using to actuate the cables, servomotors?

[@IRIMLabKoreatech]

We make our customized motors by using frameless BLDC motor components. As one of our goals is to make the reflected rotor inertia as small as possible, we have to carefully make our own motor. Minimizing the rotor inertia and choosing a properly low gear ratio are both important to achieve human-like effective mass at the end-effector.

[@challacustica9049]

@@IRIMLabKoreatech thank you for the explanation! Have you considered using peano-hasel actuators as well?

[@IRIMLabKoreatech]

​@@challacustica9049 Thanks for introducing the peano-hasel actuator. Using artificial muscle is our long dream. It will make most of the problems disappear such as the reflected rotor inertia problem. But, as far as we know, the performance of electro-magnetic motors such as linearity, precision, and power-weight ratio outperforms the others. Maybe we need to take a look at the recent achievement of artificial muscles.

[@challacustica9049]

@@IRIMLabKoreatech thank you for the explanation!

[@AlexandrVoinov]

Добрый день а какие использованы сервоприводы!? Good afternoon, what servos are used !?

[@liuzzivfx9413]

Curious what is going on with that right angle transfer in the forearm? @ 1:21

[@shamsulchowdhury8240]

Elbow looks really human like.. Is it possible to get the mechanical drawing or cad drawing of the ELBOW design? i am working on prosthetic elbow joint as a hobbyist, cant find a realistib elbow joint.. thanks..

[@cosmopolitansoul2001]

Any chance to this version getting Open Source? I loved the design! I'd love to make 3D printed! congrats all!

[@joshmouch]

I saw a video where V2 was patented, so I don't think so.

[@AnkushKumar-fj1ez]

can this be used to draw shapes on board. like pictures etc

[@bluesir2210]

Great work! Is there any related paper about this work available ?

[@IRIMLabKoreatech]

We haven't published the papers about LIMS3 yet. For the previous versions, please refer to them: ieeexplore.ieee.org/document/8016639 ieeexplore.ieee.org/document/8594301 www.sciencedirect.com/science/article/abs/pii/S0957415820300787

[@williamhuang5329]

Hanzhen harmonic drive gear , strain wave reducer, robot joint , over 30 years experience

[@sY-yv6uh]

사용된 모터들에 관한 정보좀 알 수 있을까요?(제품명이나 브랜드 부탁드려요)

[@IRIMLabKoreatech]

We are using several frameless motors from the T-Motor company.

[@janlietaer9046]

Great robot : How does the motor to wire mechanisme work? How many motors are there in the arms?

[@IRIMLabKoreatech]

LIMS3 has 7 motors per arm which is the same number as the human arm's DOF. All the motors are placed at the proximal parts to reduce the distal mass.

[@IRIMLabKoreatech]

Actuating the high-DOF distal joints, especially the 3-DOF wrist joint, is extremely challenging. We came up with the unique Quaternion joint for the 3-DOF wrist. LIMS3 has the new version of the Quaternion joint. Utilizing the property of the rolling contact joint of the elbow, we could completely decouple the wrist motions and the elbow motions even though all the actuators of the wrist are at the upper arm. Finding and implementing these was a really long and exciting journey:)

[@user-jl6yt9ji8s]

새로운 버전의 ambidex도 대단하네요. 와이어 구매처를 알 수 있을까요??

[@IRIMLabKoreatech]

3.2 phi 케이블은 평범한 7x19 cable입니다. 튼튼하지만 bending radius를 작게할 수 없으니 설계할 때 신경을 많이 썼어요. 1.6 phi 케이블은 작은 bending radius가 필요해서 특수한 7x49 케이블을 썼습니다. 미국의 SAVA Cable사 제품입니다.

[@j1doc672]

Is this commercially available?

[@hikolanikola8775]

question: Why use cardan joints instead of CV joints??

[@IRIMLabKoreatech]

The Cardan joint has a disadvantage of speed fluctuation. As we used two Cardan joints in series, there is no fluctuation between the wrist yaw joint speed and the motor speed. Two CV joints can be also used. However, in the new Quaternion joint of LIMS3, these two joints should withstand all compressive, tensile, and torsional forces. Thus, Cardan joints are more suitable.

[@uthvfyrekbnm6008]

Is the assembling of entire construction very complex?

[@IRIMLabKoreatech]

In general, assembling tendon-driven robots is more complex than other articulated robot manipulators. But, the assembling complexity of LIMS3 was significantly decreased compared to LIMS2. For instance, we can finish assembling frames without cabling. After that, we can assemble cables that already have end plugs. That means that we don't have to worry about the cable length during installing the cables.

[@uthvfyrekbnm6008]

How you tension the cable in big capstan?

[@IRIMLabKoreatech]

There are many simple and effective ways of tensioning. LIMS3 has some bolts to pull the end part of the cables. Thus, the tension can be easily adjusted using a hexagonal wrench.

[@inofid]

Какие тросы используются? Похожи на стальные. Толщина тросов очень большая. Не является ли проблемой большой радиус изгиба таких тросов? Таким тросам нужно очень сильное натяжение, и соответственно более жёсткая конструкция манипуляторов. От чего и масса тоже станет больше. Компактность и лёгкость не в приоритете? С такими толстыми и тяжёлыми тросами, использование тросового привода теряет смысл и является чисто принципиальным решением. What cables are used? They look like steel. The thickness of the cables is very large. Isn't the large bending radius of such cables a problem? Such cables need a very strong tension, and accordingly a more rigid design of the manipulators. From which the mass will also become larger. Compactness and lightness are not a priority? With such thick and heavy cables, the use of a cable drive loses its meaning and is a purely principled decision.

[@IRIMLabKoreatech]

We used two types of steel cables in connection: Dia 3.2 mm and 1.6 mm. The 3.2 mm cable was chosen to achieve high stiffness even though it is overkill considering the strength. The bending radius should be large enough, which was a big challenge in the design stage. Large cables do not always require strong tension. The pretension should be more than half of the maximum tension we use, not the tension related to the max breaking tension. We think that we can understand your point. There might be two ways of research; sticking to and improving the original idea and pursuing the optimal solution. Both of them are meaningful, and this time we chose the optimal solution. Having 4 times larger torque and better stiffness at the expense of somewhat increased weight was sweet for us:)

[@tianliangliu9872]

@@IRIMLabKoreatech May I ask that how can you test the pretension of the steel cable to ensure that it reaches more than half of the maximum using tension.

[@sawkabear654]

Where the music form cyberpunk

[@slevinshafel9395]

1:03 look more complex noe than older deasing.

[@IRIMLabKoreatech]

It depends. It has simplified cable routing and a more complex coupling link design. It'll be interesting to compare with LIMS2 as, kzbin.info/www/bejne/l33EooComrudb9E

[@oneproductivemusk1pm565]

This is really smooth and obviously has many degrees of freedom, why not sell the technology to Tesla? Looks like this could work better on the Tesla Optimus bot than that clunky shoulders and about to poop stands.

[@AdityaMehendale]

Why the delete + re-upload?

[@IRIMLabKoreatech]

There were some typos and some mistakes. But still, we have typos :( Thanks for visiting again.

[@AdityaMehendale]

@@IRIMLabKoreatech I ask, because I usually set (such) videos in topical playlists. If you delete a video, it disappears from my playlist without warning; essentially I lose the 'bookmark" forever. I luckily caught yours in time :) Great work! Side question, you likely use steel cables for their stiffness. Did you consider other organic materials? What led you to choose steel over the others?

[@IRIMLabKoreatech]

@@AdityaMehendale We have experienced polymer wires such as Dyneema and Kevlar, but not organic material. Does it mean the organic string such as tendons used for bowstrings? Plz recommend it.

[@AdityaMehendale]

@@IRIMLabKoreatech Indeed, I meant stuff like Dyneema, perhaps carbon-fiber. What are your findings? (and why did you ultimately choose steel-cables?) Your idea of tendon-like materials also sounds good; I hadn't thought of that. (Are such materials readily available? What advantage would they offer over Dyneema?) Thanks!

[@user-sl5yh2mh7w]

а зачем такие толстые тросы? они кажутся избыточными

[@IRIMLabKoreatech]

For strength, this 3.2 phi cable is overkill. We have used this thick cable to get high stiffness. Most of the tendon-driven mechanisms have low joint stiffness due to the elongation of the cables. However, low stiffness decreases the control bandwidth which could cause a crucial problem for dynamic control. Для прочности этот кабель 3,2 phi является излишним. Мы использовали этот толстый кабель, чтобы получить высокую жесткость. Большинство механизмов, приводимых в действие сухожилиями, имеют низкую жесткость соединения из-за удлинения тросов. Однако низкая жесткость уменьшает полосу пропускания управления, что может создать серьезную проблему для динамического управления.

[@user-sl5yh2mh7w]

@@IRIMLabKoreatech спасибо за пояснение)