My buddies on the Yak Rover project are all much further along. Maier in Israel already has a spider-bot wandering around in his garden, controllable over the internet via commands to a bot on our Discord, and with a camera feed.
Conversation
But in my defense, I’m the only one trying to build maximally from scratch 😎. Of the 5 others in out build party, 3 are using off the shelf robots and focusing on software, 1 is using an ESA open-source design but focusing on a RISC-V board, and one is using a kit robot.
1
9
We’ll see. Place your bets: what are the odds that I will achieve my drive test goal for the year: an autonomous rough figure 8, including climbing over a 30-degree slope obstacle. I give myself 22.3%.
1
7
Nice trick
Quote Tweet
Replying to @vgr
I like to put electrical tape on needle nose pliers (as well as the jaws of soldering third hands). The tape provides some friction and cushions parts from (potential) damage from the jaws.
1
8
Rhett has a super-minimalist build going
Quote Tweet
The varying approaches have been fascinating as we all approach "Rover" with whatever part of the problem seems most interesting. I of course couldn't help myself from building a Flight System from scratch.
github.com/rhettg/stubborn
Show this thread
1
2
Continuing rover adventures, since my main robot design, Nature is Murder, is a complex 6-wheel rocker bogie, I decided I needed a very simple test chasis just to work on the motor control and programming in parallel. I decided to call it... Accessory Before the Fact 🤣
1
3
Unlike NIM, ABTF is not a production robot. It will have 4 driven wheels, each with an SG90 steering servo attached to a wheel unit with an N20 drive motor (4 corner holes), and a Beaglebone Blue board (4 posts)
1
3
It’s gonna be a cable management mess
1
2
It should be able to drive around fine, enough to test motor control, but probably can't handle significant obstacles or hold much by way of sensors and things... the mounting posts are to raise the board above the chassis, leaving room for a battery below
1
1
Printed and tried a partial ABTF assembly. Wanted to use 4x M3x20 flanged bolts and nuts to secure the board to the posts, but had to use unflanged for 2 because the mounting holes are cramped and right next to some components that don’t allow a flange or washer. Works though.
1
8
Next up: designing the drive motor housing and assembling the whole 8-motor beast. After which some janky JST wire harness assembly and adding a 7.4v lipo battery pack.
Then the Everest... actually programming the Beaglebone Blue. I’m frankly intimidated by that.
Replying to
👀 Eyeing this Oak-1 camera for my rover. Wasn’t actually planning to have a camera but this looks like it might make it doable. ht
3
4
Side quest: misadventures in learning to crimp.
Quote Tweet
Little thread on what’s one of the harder hands-on engineering things I’ve done in my life. A real test of dexterity and eyesight: crimping a wire to put in a connector. A JST PH connector to be precise. What’s crimping? Here is an expert tutorial iotexpert.com/jst-connector-
Show this thread
1
4
We make progress, designing the real rover... this is like 6 hours of CAD work :D
1
5
The trick to efficient OnShape cad design, I’ve realized, is to design each component in an assembly spatially in place using offset planes etc. That way alignments are visible as you build. You can check assembly view as you work on the part view, to make sure holes etc line up.
1
6
It’s interesting how much this is an REPL like in programming. I literally had calipers handy and was measuring COTS components as I went, like the 2s lipo battery (2 blue cylinders), turnbuckle link, and ball joint between rocker hinge and body, to make sure it would fit.
1
4
You discover design constraints/needs as you do this. For example, that white connector on the battery is supposed to plug into the 3-pin header on the board near it (this is a Beaglebone Blue robotics board). But it’s too short to come around from the bottom the way I’d planned.
1
2
So either Plan A: add an extension cable (which needs room) to keep the battery on the lower deck or Plan B: put the battery above the board, raising the center of gravity and making the rover more top-heavy than it already is.
1
Two more. In assembling-while-designing, I realized my original plan to have a completely covered top deck wouldn't work. The BBBlue antennae need to stick out there, plus need top access to gpio headers to connect stuff on payload deck. So ended up with a nice big hatch hole.
1
The second one: while sketching this out, I hadn't factored in that the turnbuckle (red rod a few tweets up) is too long for the differential bar to be on the main body... so now I had to add a tail plane to hold the differential bar (on perseverance this is in the middle)
1
3
Whew, finally figured out how to do mate connectors correctly. Now I have this aluminum strut mated via a slider joint to the rocker hinge
1
4
My rover body is going to end up tiny relative to the undercarriage. This is because my 3d printed part dimensions are limited by my printer's build plate (120mm square, but effective max more like 100mm square)
1
2
Ran out of steam, but printed out a roughly to scale side view so I can sketch to visualize the rest. Not bad.
2
4
Hehe triple-hybrid test assembly: CAD printout, pen lines on paper, some actual components. This could be an art form.
1
10
The first complex thing I ever built was a model kit glider ~1988, that came with 1:1 scale drawings. This now-vintage one. I built it by using drawing itself as the construction template, with nails on a board,. It was all balsa wood cut-and-glue pieces. 1:1 is a great hack!
2
8
One minor problem I have to solve now is to make a passive roller out of this wheel. It has a 3mm d-shaft hole to press fit on the shaft of an N-20 moto, which is how I’ll use it for the 4 drive wheels, front and back, but the middle 2 wheels need to be passive free rollers.
1
1
First thought is to try and drill out the D into a 3mm circular hole, and put a circular shaft into it supported by a fork. But I don’t like destructive mods. Alternative is to use a round shaft with a D-end, and cantilever other end in a sleeve.
There’s just so many details.
2
1
Sketched out wire harness problem. I need to solve for A and B and make 4 of these.
Current solution: A and B both 9-pin DuPont wire-to-wire connector pairs. So would need to cut off the connector that came with the sg90. The red cable needs to route through aluminum struts.
2
Thought of using ribbon cable for this (initially thought 18 conductors together for 2 drives on one side). suggested IDC punch down connectors for ribbon end. Not sure what it would mate to on board end… wires soldered onto an IDC header?
1
thought this was overkill and just shove 9 separate wires through. He also suggested USB for the red interconnect, though it’s not meant for this use case.
1
One interesting thing… the orange stepper female 3-pin futaba connector is what’s needed at the board end. The cable is even long enough and fits through aluminum strut. I may be able to just make A/B a 6-conductor problem instead of 9.
1
And for the 4+2 N20 problem, the cables at either end are not long enough but together the should do the trick. Remember this sketch is ~ to scale. If so, no red cable/connectors may be needed at all.
2
I could just solder individual wires or use butt connectors like these to connect wires individually.
1
1
1
The main reason to try and create a “middle mile” interconnect (the red bits) is modularity and swappability. I could swap out drive units easily, exchange boards, mix and match undercarriages.
1
I do like the idea of each drive unit being a neat 9-pin connector.
2
Farthest out of my wheelhouse: rest of the wire harness for peripheral and payload function. There are 8 GPIO pins I can make good use of, but the bus options (CAN, UART, SP1, I2C) are black-magic to me. Where might a camera go?
3
1
I’ll admit the thought crossed my mind since D-subs are the only kind I’ve actually soldered in my brief, unglamorous grad school career in hardware circa 1998. Bulky so easy to solder.
Quote Tweet
Replying to @vgr
If you’re proposing a DB9 connector I have to leave.
1
3
For the Gen-Z usb-c kids who’ve never seen this kind of cabling...
1
5
Forking off a subthread on learning to use the Beaglebone Blue, which I suspect will take over this main thread if I let it. Will post main rover shit on this thread, BBBlue-specific stuff on this side thread.
Quote Tweet
Off to an inauspicious start with the Beaglebone Blue. The network drivers won’t install on MacOS Catalina. Apparently the packages, HoRNDIS and EnergiaFTDIDrivers (no idea what they do… ELI5?) do shady shit at system level so Catalina sez no.
Show this thread
1
Reprinted the ABTF test rig chassis with taller standoffs and now the battery fits and the short cable even reaches the board. But ordered an extension cable too.. this is too awkward.
1
As you can see, I had to add like 15mm because I initially planned on flat pouch battery but switched to this cylindrical one. Switch from red to white is because I got a 2 for 1 deal on white PLA on prime day and am saving my red PLA for main rover.
1
2
Show replies



