Purpose

Solve 80% of the chores 80% of the time

I spend a lot of my time picking up after my kids. it's not hard work, but it's repetitive. put the toys back in the box, put the laundry in the hamper. Of course I teach them to do it myself, but I can't help trying to automate it too. 

Picking up laundry does not require a delicate touch. any two fingered gripper will do, and we don't need millimeter accuracy, just centimeters. Cable driven parallel robots offer an interesting solution to this problem. I hang a gripper at the end of a couple of fishing lines on motorized spools and program them to move the gripper around the room.

I think if Stringman can tackle one common chore in a good enough way while staying cheap, then it's a viable product you might buy.

Main Components

A gripper, with one smart servo (ST3215) for the fingers and a linkage to move them in a mostly parallel way when near closed. and another servo to spin a wrist.

A place to display the fudicial tags that help calibrate and locate the gripper in the room. After a few iteration I decided this is best placed above the gripper by about 50 cm and linked with a rigid pole. the pole gives the wrist something to push off, and the distance lets it get down beside chairs and beds.

Anchors. Four motorized spools. the motors have to be capable of precise low speed motion but also able to turn without limit in either direction. 1nm of torque is sufficient for most payloads. Reduction is basically unnecessary and would be a source of noise

I'm currently in the process of upgrading the anchor system from four anchors driven with MKS42C steppers to two anchors each having a pair of DAMIAO BLDC hub motors. This affords a lot of improvements I'll elaborate on later.

Cost

My aim with this robot was to target $1000, and that's roughly where I'm at. 
The pilot anchors + arpeggio gripper currently is $790 on Tindie. While the system with the new anchors are going to be higher, probably between $1200 and $1800.

In order to build the current design yourself, the total BOM cost before shipping and tariffs is about $600

Control Methods

In theory the control of a CDPR is simple, calculate the distance between each anchor and the platform and command every spool to reel to that length. But in practice you need a lot more active correction because it's impossible to perfectly calibrate it.

I've settled on using a Kalman filter that combines an estimate of the hang point from the motor encoders and the visually estimated location from the apriltags into a synthesized position.

I use velocity control rather than positional control because it gives more consistent behavior throughout the room and isn't as dependent on calibration.

I don't use ROS because it's over complicated and couldn't simulate strings well. in fact I've never gotten any physics sim to property simulate the strings yet, but I haven't tried that hard. it's just that what I've come up with works fine and I've moved on to applying to cleaning up clutter. I can always come back to simulation later.