I design a new foot for added a foot degree of freedom in order to creating toes. The connection is made by means of springs that can support the weight of poppy in this configuration:
For the moment, I don’t added sensor. In the second time, I think to use a force sensor maybe these sensor. I think to put 3 ou 4 force sensor, and I would change the mecanics to improve contact. I would add the encoder near the spring for mesuaring the angle of rotation. I don’t fix my choice so if you have any sugesstion, I take it
I think the double rotation foot is needed to improve stability but do not forget we also need to measure the pressure under the foot to stabilize the robot and also to detect the foot contact on ground during walking. It is very difficult to have a robust technological solution. I think the mechanical design should help to reach this design… I am working on it and keep you in touch.
Which sensor ? force sensor using variable resistivity or simple contact switch ? Torque estimation in motor or torque estimation in line?
The double rotation foot of this page is a good base.
I have started to work on the foot shape. To do so, I’m currently trying with the use of clay. I first thought to give the shape by hand modeling but then I decided to try before just by putting clay directly inside an appropriate shoe.
I see you want to reduce foot length. I advise you to take care. A test to check foot design is to put the robot on one foot, stance leg vertical and swing leg slighly folded to avoid hurting the floor.
With the tight motor make the swing leg balance as quickly as possible.
The inertial moment applied by the swing leg along vertical shall be lower than the moment applied by ground on the foot. On the contrary, the robot will twist and it will not be possible to make any stable stride. (Twist stability)
I presume it is the reason why romeo has very long feet because its weight is relatively low.
Indeed but it is more a foot-sole friction problem than a foot length problem. So we should take care of having enough friction to counter-act swing leg inertia, right ?
Yes but when you have a long foot the twist friction is better. Moreover, when you just put the heel on ground, you shall have a good twist friction, like an aircraft which lands on the runway. The shape of the heel is then hard to design.
And it means also that the twist friction/surface of the foot is linked with the total weight of the robot and the maximal speed of the swing leg.
On Poppy foot length IS a problem, mainly because of the backlash and non rigidity of the upper body, which makes the CoM easily drift out from the support polygon even in static.
But you are right, we should also see what happens with a high speed swing, if we are able to make it stand on one foot…
For the friction, our recent test indicates that the naked foot are very slippery. With the shoe it is much better.
@Steve From a limit cycle point of view, the static standstill phase is an attractive point of walking where kinetic energy is 0.
For foot design, I thought a long time about it. I converged around several solutions. The first one is that the design of the best “gripper” is the shoe. But the best shoe is the ballerin. It is light, flexible and gripps a lot.
After, to get a good twist stability, a good way is to have two main pressure surfaces at the toe and at the heel. It is like the foot wanted to “grisp” the ground. Moreover, the area of the two surface shall be vast. Some foam between the shoe and the plastic frame seems to be the best.
About the foot frame, The ankle shall be fully compliant below the stance leg. Moreover, the centre of pressure shall not leave the foot. To measure it, the only way is to have force sensors under the foot.
The ideal foot for me is a pyramid with the sole as basis, two corners at the heel and one corner at the toe. Each edge of the pyramid shall have a force sensor to know where is the center of pressure.
To do it, a monobloc 3D printed pyramid with a foot shape seems to be a good start. Along each edge of the pyramid, a notch shall be done to insert a Force Sensitive Resistance (FSR). An FSR shall not be in direct contact to the ground. The force interface shall be normalised to insure robustness.
The deformation of monobloc pyramid will pinch the three FSR.
Ankle compliance is something we tested and it seems to be a good feature to have.
For the measurement of the center of pressure (which is always on the foot by definition) it is more difficult.
FSR are not very precise and as you said, a clever mechanical contact is needed.
Your idea about the “pyramidal” foot is interesting. Right now we are experimenting something with a kind of sole pushing on the FSR. We will see if it’s better…
The ankle compliance is needed for high speed. But maybe for low speed, the ankle moment can be used.
I agree that fsr are not precise but maybe it is not important to be precise. I think that knowing center of pressure is on toe, on heel, on the middle, on the right or left is sufficient.
I ever tried to put directly the fsr on the sole. But the heel strike was difficultly detected. But if you added an interface i wish it will work. I will send a photo of my last version of foot this evening.
And here is the foot without shoe. You can see the two parts of the foot linked at the toe with two plastic screws (they are isolated).
Each metal part of the foot is linked to electronic. If there is a heel strike, the two parts are in contact and a short circuit is established. Heel strike is detected.
