Video

That's a very interesting question! Indeed, most of these objects could be seized with a pinch grasp, namely the distal phalanges. However, the resistance to slipping out is then weak unless the contact forces become huge since you only rely on local friction coefficients which also depends on the object and may vary a lot. Power grasps (compliant) on the other hand are generally better: they spread out the contact forces on a wider surface (thus less chances of damaging fragile objects) and are less prone to slippage. This is the same as with our own hands: for larger objects and a strong grasp we tend to grab with as many fingers and as many phalanges as we can. Pinching is an option when you have small objects (think of a needle) and enveloping is not practical but it is more difficult to maintain a secure grip.

Hi there, sorry for the late reply. The reason why the springs at the top are extending a lot is because: 1. they have a relatively low stiffness but most importantly 2. to produce the small tension required to keep the bottom cables reasonably taut a huge force is necessary in the upper cables. If you look at the projection of the top cables direction along the vertical axis, you get a very small vector component. Thus, to create a small force in this vertical direction a huge magnitude is required and this magnitude reaches values close to the maximally allowable tension. This phenomenon actually defines the limit of workspace of the robot. See the associated paper for more details on the computation of the workspace and numerical values. It is important to understand that this is a physical aspect that is intrinsic to the geometry of the robot, no control algorithm would be able to alleviate this issue. at least in static conditions. If the controller decreases the tension in the upper cables, either the lower cables will become slack or the moving platform will drop.

Merci Dmitri!

You are not far off... the main difference with the Chinese windlass is that our pulleys (drums in the windlass) are not round but have precisely calculated profiles to transform the linear force from the springs into a torque at the lever arm that balances the payload. But the arrangement with two drums/pulleys connected by a cable is indeed common to our mechanism and the Chinese windlass (and Weston's differential too)

That's a Schunk KGG 100-80

Hi Alex, We agree (mostly): we never said the encoder was defective. The source of the noise is indeed clearly not the encoder itself but the lines. They are long and unshielded which is bad (EMI). Another issue with the lines is that there are wires for motor power (PWM) alongside wires for the encoder. Again, this is not good practice. We built this setup with what we had at hand and very little budget. This is not an issue with grounding (aside the grounding of a potential shield) as we confirmed the noise with both the scope you see in the video and the tuning software of the motor drive. It is also not a power issue, the DC power supply we have is quite good and we have a huge capacitor on it (as per manufacturer recommendation). Where I disagree is that putting the capacitor on the line is pointless. It is not since it fixed our issue... It did not eliminate the spikes but reduced their amplitudes to a point that was good enough for the drive to work properly. Cheers.

@@PolyLabrobot The issue is not resolved. A lot of noise is clearly still on the line as shown on the scope and you've only just about managed to keep the noise within the thresholds. In industry, the cable is shielded and they use differential signalling. This capacitor 'fix' is a bodge - change the parameters of the system and the 'fix' completely stops working.

@@Shocker99 The issue *is* resolved since the motor is working correctly now. You can never completely and perfectly remove noise from any real signal. Filtering only allows you to reduce the magnitudes of unwanted harmonics and as you said keep it within thresholds.That is the point. We agree that using shielded cables and differential signals would be a great solution but in this case we are aiming at a quick and easy fix at the lowest possible cost, not industry standard compliance. This is prototyping, not production. Cheers!

Very often yes and I do love compliant mechanisms a lot! However, there are situations where you do need a rigid link gripper: e.g. in some industries the tools are getting seriously abused and/or work in harsh (corrosive) environments and in that case, steel or stainless steel with lots of material is a must. Cheers!

It's possible quite easily by moving the attachment point of the spring (lower right) with an actuated slider or arm. That's beyond the simple mechanism we intended to demonstrate here though.

Well the "truest" answer is we did it this way because we found it fun and interesting! In the lab we explore how to do complex stuff with the minimal number of actuators and relying as much as possible on mechanical intelligence rather than electronics. See our other works about adaptive gripper design for example. Now in this particular example, if you have numerous winches you need to control them synchronously to get the right trajectory. It's not very difficult but still, more work than for us: this particular mechanism does not rely on controllers or sensors at all. There is just a simple DC motor to which we bring a voltage, no feedback or pulse sequencing is required. By design the robot *cannot* deviate from the trajectory so there is by definition zero position error. Cheers!

Industry generally wants products to be as cheap as possible. Why spend money on two axes when only one will do? Furthermore, it's an additional axis will require more maintenance and there are more things to go wrong. If a product is only doing the same operation then why build in flexibility that will never be used? There are many machines in manufacturing that use a similar motion that repeat 100+ times a minute and they will do that exact same motion for the entirety of the machine's life. I like this different take to perform this motion with only a single motor. Normally, i see it with a cam and a 4-bar linkage.

@@Shocker99 I only meant that he should mount the meter in different orientations on his shaker to test if it is more susceptible to vibrations in some axis than others.

Thank you!

I do know this product very well as I am the original inventor of its kinematic structure. If you're curious enough, you can find my name in the "patents" section of the 2F-85 and 2F-140 user manual. Robotiq licensed my patent on this kinematics. The Pacome fingers are an evolution and improvement of this previous patent. These fingers are entirely passive whereas the 2F-XX grippers need an actuator to rotate their lower link and close. The Pacome fingers can be attached to pretty much any gripper you have (hydraulic, pneumatic, electric, whatever) and they use the motion of this gripper as the actuator. They are simpler (=cheaper) and more versatile. There are also some more subtle advantages with the Pacome fingers but it would get quite long and technical to explain that here. Cheers!

@@PolyLabrobot are you selling this one or licensing the patent to a manufacturer? I'm looking for a good robot end effector for a unique application

@@imeautomation Yes and yes. Reach out to me at lionel.birglen -at- polymtl.ca and I'll give you more info.

The simulations starting @3:09 were made using Ansys. See www.ansys.com/products/structures/ansys-mechanical

@@PolyLabrobot ah thank you, the simulated 3D structure is really impressive anyway!

Thank you!

That's a very good point and you are absolutely right. At the rising and falling edges of the signal, that capacitor is essentially shorting the input pin and that could create current surges. I believe it still works out here because I have a very large line resistance due to long wires and multiple connectors along the path which all act as the R of the RC filter you rightfully suggest. What leads me to believe this is that I actually measured approx 2.7V for the square wave at the motor driver input instead of the 5V supply of the encoder. That tells me I actually have a voltage divider where the line resistance (+output impedance of the encoder) is of similar magnitude than the input impedance of the motor driver.

Thanks! Having a look at the original paper might help you: asmedigitalcollection.asme.org/mechanismsrobotics/article-abstract/10/6/061001/367087/Differential-Noncircular-Pulleys-for-Cable-Robots If you don't have access to the ASME repository, an open source preprint of that work is available at: publications.polymtl.ca/3544/

Thanks

@@va7795 I haven't read the paper but my intuition is that the shape of the cam stretches the springs to match the gravitational force exerted by the mass, so that the two forces balances out at all angles. Ie. The mass exerts 10N of force downwards, and so the cam stretches the spring so that it also exerts 10N of force but in the opposite direction, therefore the mass doesn't move because the forces are equal. The cam is designed so that the force produced by the springs always matches the force exerted by the mass. The mass will always rise if it smaller than what the cam is designed for and it will always fall when the mass is greater.

Hi! Sorry but I do not think we have the program we used to record this demo anymore. What I can tell you is that it was a simple playback of a sequence of prerecorded waypoints. It was indeed run from ROS but the robot is not detecting the objects by itself. The objective of our video was to demonstrate the fingers we designed rather than the robot capabilities to detect objects and compute grasp poses.

Hi Tim. We actually have already built several compliant versions of this design and indeed it works out very well. We did not put anything online yet as we are trying to publish a scientific paper about that first. Hopefully, we could let out more info this year or the next. Good luck with your copy of the design!

Indeed, that is the main limitation of our robot: you cannot change the trajectory of the robot with this design. It is essentially a tradeoff: you only need a single motor and the control is utterly simple (just rotate the axis, the trajectory is mechanically programmed in the shape of the pulleys) but deviating from that trajectory is impossible without updating the mechanism, e.g. changing the pulleys, as you rightfully noticed. Cheers!

PolyLabrobot thank you for the response.. What program do you use for computer simulation?

Matlab.