We’ve come a long way! We are now able to control the tentacle fairly well with the Myo Band.
This video was from a demo for Ben Shapiro’s Machine Learning and Human Computer Interaction class (where most of the team members are from).
Myo Band control is implemented via a machine learning interface (Wekinator) interpreting signals from the Myo Band. Training is accomplished by having the user move his arm to four ‘corners’ that mark the range of his or her movement. Wekinator takes those signals and interpolates signals in-between, producing a mapping between Myo band arm position and tentacle position.
The main current challenge is consistency–both in terms of producing reliable control signals from the Myo band, and in terms of the tentacle responding consistently.
The latest video posted below shows some of the progress we’ve made with the software.
We’ve been able to speed up the motors considerably as we’ve become more comfortable with the setup.
From the physical standpoint, the tentacle has been wearing out–it seems like the foam has a tendency to fatigue and break down. This leads to the tentacle collapsing under pressure, and in an attempt to compensate for that I have added a stiffening element–a bit of plastic conduit. As a stop-gap this works, but it takes away some of the flexibility.
To fix this, I’ve simply made a new tentacle with fresh foam. For a prototype, I think it’s okay to have materials with a limited lifespan. I’m confident there’s a manufacturing process that would yield the right material, but for now replacing the tentacle is sufficient.
Plan going forward is to concentrate on the end effector. Designs are being drawn up, and we hope to have something soon!