I give you news for the foot with two linear actuator. I change the design of foot and I change many fragile part (axis…)
For the moment, I haven’t tested the solution on Poppy, because I have just one foot (two actuator), but I have tested the speed and I think it’s pretty good for the walk.
@NicoX I did not progress with the solution, because there are many inconvenient.
The actuators are not compliant and the speed are low. I did not find any other actuator to respect the design of Poppy and the speed.
For Poppy, I choose this solution : New solution of ankle: Motor + spring
If you have idea or new actuator to respect speed and design I am interested.
Hi
I am not at all a robo expert Just a hobby builder.
Actually I was trying to explore cheap actuators for building, and liked your idea. But linear actuators are slow I guess. I am at no point to commit as I do not carry any experience.
I do not know whether this gives you some idea but I found this link below after googling. http://roboticgenome.org
if you just need positional actuators (to move something) versus force actuators (to really lift something), nitinol wire can be a solution.
what you need is a procedure leaving the ankle free when foot touch the ground, then you need several pressure sensor reading to check if foot is flat or not.(all sensors touching equially the ground)
then you can act on the nitinol wire to compensate but that 'a pretty complicated way.
if you want an active system instead a reactive system, you need an optical system that can read the foot position versus the ground configuration, like projectinf a grid on the floor and reading it with a camera to see deformation. it is commonly used by kinect and other sensor of the same type. again this is complicated because it is just giving info, then you need to do something.
another solution would be to do nothing , leave the angle in free mode to touch the ground, let the foot take the orientation and jut freeze the ankle when the body weight push really on the foot.
After all this is the way it works for real.
This could be done with ferro fluid or powder that you just solidify with a magnetic field.
I have already tried the nitinol wire. It is not possible to make control with these actuators since:
the connection between wire and mechanics is very complicated (as there is deformation)
no regulation possible since the delay between order and contraction is too high
The purpose is to heat the wire to force deformation… there is a very low efficiency.
No contact with plastic since there are risk of burning
No contact with hands (I know it a lot )
The robots which use these are on/off actuators
Therefore, this technology is not yet efficient for robotic domain… for the moment.
Concerning optical sensors, I wondered about these. There is a big issue about leg flexibility/slack. Therefore getting position of the feet on ground using camera is a good idea. The issue is that the technology which does this at 100Hz is easy. I thought about a laser running along the leg to detect bending…
personally i would try to leave the articulation free until the foot touch the ground then lock the articulation with an electromagnet. So you just need to detect the contact with ground, wait for the robot transfer the weight on the leg and lock the articulation.
the tricky part is to measure how the leg is extended, to know if the foot lands on a rock (so the leg does not full extent) or lands in a hole (the leg is fully extended and it is not even enough).
there would be several solutions like measuring two pooint between the legs, so would you know which leg is higer than the other.
Just a hobby builder.
)