Faster Better Stronger Stringman

[Nathaniel Nifong]

Self parking and gripper changes

I spent some time working on a parking, un-parking, and auto start routine for Stringman and documented the process along the way. This is an absolutely necessary feature for livability. But there is still a long long way to go to “reliable appliance”.

Designing a Self-Parking and unparking feature for Stringman

The marker display box above the gripper has also had a slight redesign around small carabiners so that it will be easier to disconnect all the lines for storage.

The gripper also now incorporates a small cooling fan, which requires some holes in the face plate. Which design do you like more?

Constant force gripping

When stringman grips an object whether automatically or manually, it now seamlessly transitions between finger movement and constant force gripping based on its finger pressure sensor.
This means you can grasp things delicately even if you’re not in the room with the gripper looking at the object for real time feedback. As you lessen the grip force (holding B on the gamepad) as soon as it crosses zero, it switches back to opening the fingers at the commanded rate.
This way it can also release objects of any size in a quiet and minimal way without just blindly opening the fingers to maximum extent.

Prototype of Improved AnchorsThe biggest outstanding issue with Stringman is limited torque of the pilot anchor motors, which requires staying under a certain function of speed, height, and payload weight. If this threshold is crossed, the steppers will miss a step and the gripper basically plummets to the ground, and the motor controller then recovers by reeling it back up to where it started. This behavior of course is totally unacceptable and I don’t want to be limited to picking up shirts and socks in tall rooms.

Ever since that issue arose, it was my intention to select a better motor. I think I’ve finally found the right candidate, and I have a prototype of the new anchor on my desk. It’s roughly 8x stronger, and due to some clever savings elsewhere, the BOM cost of the whole robot is roughly unchanged.
In addition, it replaced the delicate line tension switch with constant torque feedback. This should help a lot with keeping the lines tight at all times, which in turn spreads out the load and gives more headroom below the danger threshold. It also lets us do things like measure the weight of a payload.
As an added bonus, this motor is utterly silent, even in motion. Massive performance optimizationsAfter a brief foray into trying to use dedicated nvidia hardware to accelerate the tag detection and video transcoding, I decided their walled garden wasn’t worth the trouble and came up with my own optimizations for the CPU bound apriltag library
CPU usage of Stringman during operation is down 90% with ram use down by about a gig. It should now be possible to run it on laptops and mini PCs. Please consider sharing my videos if you want to help out. thanks everyone for the support.