The aim of this project is to build an electric drive for my daughter's manual wheelchair.
Okay, so I'm pretty stoked about this-- Triad Semiconductor finally has the castellated TS3633 modules in stock. These are the light -> voltage frontend photodetector modules that go with the Vive lighthouses. If you haven't read about how the Vive lighthouses work for precision tracking, it's a pretty awesome design. But in short, these are a key piece in putting together a crazy-high precision indoor tracking system. Because of the precision and fact that it's royalty free, I'm convinced we're going to see some awesome innovation coming out of this tracking in the next couple of years.
Anyway, I just got my order in and should have them in hand soon to start putting together the tracking piece for the autonomous chair.
There are a lot of strategies for indoor autonomous driving. SLAM, for example seems to be getting really popular. What I've really been hoping for is some sort of indoor GPS solution. I've been keeping an eye out for something like this for years. Until about a year ago, the best bet I could see was some sort of a system that used the time of flight of an ultrasonic signal from multiple beacons (or receivers) along with an RF communication channel to determine location. There are a number of solutions that use wifi and the like for positioning, but precision is on the order of meters. But, in the last year, high precision indoor positioning has come a *long* way.
Pozyx hit the Kickstarter scene last summer, and promised an UWB tome-of-flight solution that promised 10cm precision. I picked one up this spring, and when I first set it up, it was awesome to see my positioning moving around the room as I did. But long story short, it's pretty susceptible to multipath interference and just not reliable enough for what I need.
I'd also considered taping fiducial markers all over my ceiling and using a camera to determine location. While this may work, I never spent enough time going down this path go get it to work.
Then, the Vive came out. If you haven't read up on the tracking technology, it's pretty ingenious. And it sports sub-millimeter precision!!! Right now, I'm waiting for the ts3633-cm1 modules to be available. Once I can get my hands on some, I'll be able to move forward.
Well, it's been a long time since my last update, but progress is ongoing. I've cleaned up all of the electronics and enclosed them in a case. So far, the Shoe Goo has worked very well to increase the friction, and it hasn't peeled off of the ABS, which was one of my major concerns. As a note, I did prep the ABS with acetone immediately before applying the Shoe Goo, which may have increased adhesion. At this point, the primary goal of making a powered add-on for the wheelchair is basically met. But, when I look at my daughter's ability to use this, I see I need to take it farther. The chair really needs to be able to autonomously drive to specified locations. I need to configure this so she can pick a destination in the house using her eye gaze system, and the chair will take her there. More to come soon...
One of the issues I've run into is slippage between motor-mounted drive wheels and the tires on the wheelchair. ABS seems to have a pretty low coefficient of friction. My first attempt to address this was some liquid electrical "tape" I had lying around. I tried both a spray-on version and a brush on version. Both work kinda okay, but they don't look like they'll stand up to long-term stress. They'll just peel off of the ABS over time. For a more permanent fix, I'm thinking a good product to try will be "Shoe Goo." I remember using it as a kid to get more life out of some sneakers, and as I recall it worked pretty well. Since it effectively replaces the sole of a sneaker, it should be both quite tough as well as provide good friction.
I've also found that the screw shield doesn't seem to hold the phono plug wires very well. I'll need to figure out something better. I can easily use a proto shield, I'm just trying to keep the build as simple as possible, especially limiting the soldering to only the bare minimum required for someone to get one of these up and going.
I did get some male and female wire plugs in the mail from China a few days ago, and I've added them so there's now a nice battery disconnect. Still need to add a switch-- I'm thinking a standard household electrical switch will work well, and yet be very inexpensive.
Oh, and I've added a couple of pictures and a short video of the first trial. It's not on the video, but I did try riding in the chair myself, and it had no trouble pushing me around at 160 lbs (73 kg).
This has been a project I've wanted to do for a while, and finally dove into a few weeks ago. I've got a first iteration of the design that is functional.
The basic electrical design consists of an Arduino receiving input from a couple of switches, and driving a Monster Moto Motor Controller board, which is connected to two decent gear motors. Add in the 12V sealed lead acid battery and a 5V regulator and that's all of the electronics needed.
As for the mechanicals, the entire rig connects to the wheelchair via two small brackets mounted to the chair. The main housing slips into the brackets and rotates into place. The drive wheels mounted to the shafts of the gear motors contact the chair's tires. And rubber bands maintain sufficient pressure to ensure the drive wheels are pushed against the tires with sufficient force to avoid undesirably slippage.
The mounting brackets are mostly 3d printed, as are the motor mounts and the drive wheels that mount on the gear motors.