PDA

View Full Version : First steps for Super Mega Microbot



jpieper
06-14-2014, 08:04 PM
I posted a video of our turret a while ago, and now I finally have one of the mech itself!


https://www.youtube.com/watch?v=gvj75ICwHus

It is a quad with 3dof legs based on Dongbu DRS-0101 servos. The onboard computer will be an odroid U3 plus an AVR for I/O multiplexing and other miscellaneous functionality. In this video, both the turret and the odroid are not yet mounted. Instead, I've got an extra lipo wedged in place where the odroid goes to get the center of gravity to a more reasonable place.

jwatte
06-14-2014, 08:08 PM
I've been curious about the Dongbu servos. They came out just after I got my current set of Dynamixels. How are they working?

jpieper
06-14-2014, 08:30 PM
They seem to be working really well. I've done some characterization and they can hold 12 kgf*cm indefinitely without overheating, and can peak up to 17 kgf*cm. The serial protocol is reasonably sane, and the mounting system is convenient. In this mech, all the orange parts (and the one green one), are stock brackets from their humanoid kit.

The documentation has a few registers misnumbered from reality, which probably just indicates that it isn't that widely used by anyone programming it at the register level yet, but it in any case, it wasn't hard to figure out the correct layout. As far as I can tell, with the caveat that I have not used Dynamixels myself, they appear to be superior to the comparable Dynamixel in every way aside from network effects (i.e. you're unlikely to find anyone at an event you can swap parts with), with the added benefit of being slightly cheaper.

nvrdunn
06-15-2014, 08:42 AM
I've started a quad with the Dongbu DRS-0101 servos too. I have all the IK working, starting on the gait code.

https://ke6mto.files.wordpress.com/2014/06/photo-22-e1402839622581.jpg

jpieper
06-15-2014, 03:35 PM
I've started a quad with the Dongbu DRS-0101 servos too. I have all the IK working, starting on the gait code.

Are you starting on gait generation from scratch? If you're interested, I have a hacked up copy of NUKE which is slightly more general than the Dynamixel only one, but I ended up not using it in favor of writing something new.

Xevel
06-15-2014, 03:36 PM
Funnily enough, I just got some Herkulex servos 2 days ago, and I'm working on a derivative of the USB2AX for them.

Which registers are misnumbered? Have you found any other errors in the doc?
In your demo, how many orders do you use to drive the legs? do you interpolate yourself or do you leave it to the servos?

nvrdunn
06-15-2014, 03:46 PM
Are you starting on gait generation from scratch? If you're interested, I have a hacked up copy of NUKE which is slightly more general than the Dynamixel only one, but I ended up not using it in favor of writing something new.
Yeah I was starting from scratch, if you have some code the would be great.

jpieper
06-15-2014, 07:47 PM
Funnily enough, I just got some Herkulex servos 2 days ago, and I'm working on a derivative of the USB2AX for them.

I just got the RS232 adapter when I ordered mine, although a USB interface would certainly be more convenient than the separate USB-serial adapter I'm using now.


Which registers are misnumbered? Have you found any other errors in the doc?

The PWM value appears to be at RAM register 64 when the docs claim 62.
The only position value I've found is at RAM register 60, which I'm assuming to be calibrated position because I haven't found (or looked that hard yet) for the absolute or differential position. It did seem that those other positions weren't where they were documented to be.

I haven't been keeping a detailed list of documentation errors, but perhaps I will start.


In your demo, how many orders do you use to drive the legs? do you interpolate yourself or do you leave it to the servos?

For each update I'm using a single S_JOG command to command all the servos at once.

In that demo, I'm sending commands every 40ms with a play time of 80ms. That seemed to result in somewhat stable transitions from one command to the next with minimal loss in accuracy. I believe the control loop on the servo is running at 11.2ms; I might try sending commands closer to that period when I'm running from the onboard computer as opposed to a desktop PC as my gait engine is rate agnostic. The desktop isn't able to manage jitter well enough to warrant running at the higher rate.

jpieper
06-15-2014, 08:06 PM
Yeah I was starting from scratch, if you have some code the would be great.

I posted my diffs to github. Beware, while they worked for me at one point, I'm not using them now so I can't promise anything.

https://github.com/jpieper/nuke-jpieper/commits/master

I'm planning on posting the gait engine I am using now and the development tool for it, but haven't quite gotten to that point yet.

nvrdunn
06-15-2014, 10:12 PM
I posted my diffs to github. Beware, while they worked for me at one point, I'm not using them now so I can't promise anything.

https://github.com/jpieper/nuke-jpieper/commits/master

I'm planning on posting the gait engine I am using now and the development tool for it, but haven't quite gotten to that point yet.

I will check it out.
I'm using a raspberry Pi right now for compute so all the code I wrote in python. I have the commands to control the serves done. Also got basic start sequence which clears the serves, turns the torque on and puts it into a default stance. Now been playing with a single leg swing to get a good idea for leg lift, swing etc.

KevinO
06-16-2014, 04:56 PM
I will check it out.
I'm using a raspberry Pi right now for compute so all the code I wrote in python.

My last robot Charlotte ran a raspberry pi. Great platform. Looking forward to watching your progress.

jpieper
07-13-2014, 12:09 PM
I will check it out.
I'm using a raspberry Pi right now for compute so all the code I wrote in python. I have the commands to control the serves done. Also got basic start sequence which clears the serves, turns the torque on and puts it into a default stance. Now been playing with a single leg swing to get a good idea for leg lift, swing etc.

The promised python gait engine and development tool is now posted:

http://forums.trossenrobotics.com/showthread.php?7021-legtool

nvrdunn
07-27-2014, 10:46 AM
The promised python gait engine and development tool is now posted:

http://forums.trossenrobotics.com/showthread.php?7021-legtool

jpieper, thank you for posting the app!

jpieper
11-02-2014, 05:10 AM
While we know there still isn't any large competition in the near future, we made up a "qualification" video just to prove to ourselves that we at least had everything working end to end. The long dimension in the room we were shooting in was 13ft, so only a little shorter than a full arena.


https://www.youtube.com/watch?v=K3CTdOI9ous

jpieper
06-22-2015, 04:56 PM
And now, some more upgrades in progress... a prototype gyro stabilized turret. This is just a proof of concept video, with a hacked up aluminum gimbal frame and an off the shelf Martinez gimbal controller with its code slightly modified to stabilize about the pitch and yaw axes instead of pitch and roll.

6058


https://www.youtube.com/watch?v=J3ka7KrGOYo

The upshot is, even while walking, it keeps the video plenty usable and keeps the turret aimed about as well as any of us can aim it against a moving target.

Now I'm tweaking the controller's firmware to use the Dongbu servo protocol, so that we can aim it instead of just stabilizing. If that works out, I'll probably build a more suitable turret bracket, and possibly switch to a custom controller that has two IMUs for even better stabilization.

jwatte
06-22-2015, 06:32 PM
keeps the turret aimed about as well as any of us can aim it against a moving target

OK, now I'm scared! Move-and-shoot was effective for your guys already. (As was your two-person piloting team set-up.)

jpieper
07-03-2015, 09:38 AM
Next up I'm working on a new chassis and an integrated odroid daughterboard to go with it.

6084

No one seems to go with mammal style bodies, but I wanted to experiment with it for other reasons. I also figured out the "right" way to mount the HerkuleX servos, so now there are zero washers, lock washers, spacers, or nuts in the entire assembly and half the total screws.

The odroid daughterboard integrates what was 6 separate boards/modules on the old mech:

The 5V regulator for the odroid
Power and serial data distribution for servos, fuse, and main power switch
Drivers for AEG and agitator
Shore power connector
Weapon power switch and indicator
The stock IO shield for the odroid (we were using the I2C level shifters and AVR, and had built on it serial level shifters, a laser pointer driver, and IMU)

It also has a pass-through JST-style USB connector on top for the turret webcam.

6085

6086

The daughterboard and all the main structural components mount to the top plate, which has slots where the connectors from the daughterboard for top-side things protrude. I think you should be able to take it apart in nearly any order without needing to carefully thread cables through like on our old chassis.

The daughterboard is sent off to macrofab now, the 3d parts to shapeways, and the chassis plates I'm sitting on sending out because they aren't nearly as long of lead time as the other pieces.

jwatte
07-03-2015, 03:19 PM
Speaking of Macrofab, I sent the files for an Onyx Fire II to Macrofab to test them out. Just got the envelope, so now I need to hook it up and see whether it actually works :-)

jpieper
07-19-2015, 01:06 PM
I got my first boards from macrofab back for a second rev of the scoring system a week or so back and they seem to be working out just fine.

Also, some minor progress to report on the mech itself. A prototype mammal style leg, with a 3D printed bracket to hold the final leg segment:

6103

And a 3D printed gimbal bracket that permits full range of motion vertically:

6104

jwatte
07-19-2015, 02:07 PM
Shiny!

The IK for that mammal leg seems interesting. You CAN reach a reasonable envelope, but the angle to ground will be... varying.
I'm wondering whether a good use of a fourth leg servo is to define the contact angle with ground, and then use wider feet to get less tilt while walking. This would make for a more stable gait so stabilization is less necessary.
Might need a fifth servo, for orientation, even...

(In other news: I got the Onyx Fire II back from Macrofab, and it did, indeed, work. Who needs Seeed now? :-)

jpieper
07-19-2015, 08:14 PM
Fortunately, the rubber feet I've been using seem to work pretty well even as the angle changes during a contact sweep even with pretty steep contact angles. Two well publicized, but larger robots, (BD's Spot and MIT's Cheetah), use the same 3dof geometry, albeit with much more performant actuators.

Stag, mentioned somewhere here in the forums, used a 4dof leg, with the 4th degree in the hip providing yaw control for each leg. It wasn't clear to me exactly what the benefit of adding that degree of freedom was if you had a foot which made contact at effectively one point, as Stag did.

jpieper
08-15-2015, 02:22 PM
More slow but steady progress...

The odroid shield came in from Macrofab, as did the mounting plates, servos, and feet supplies. I haven't brought up the daughterboard yet or wired everything, but did attempt a test mounting of everything but the turret today.

6130

jwatte
08-15-2015, 05:00 PM
I think yaw would allow you to clear the legs from each other (and the body) when taking long strides.

jpieper
09-13-2015, 09:29 AM
Now we've got things wired up, and have the our full software system running onboard:

6161
6162


https://www.youtube.com/watch?v=MSAmmHjaxR8&feature=youtu.be

The gait engine is running at the 11.2ms HerkuleX frame interval, with IMU data currently being logged at 200Hz.

6163

And we've moved our source repository to a public one at github, which includes all the software for the mech, the gait analysis tools, as well as the diagnostic tools like the plotting and telemetry manipulation:

https://github.com/mjbots/mjmech

Next up will be integrating the new turret with gimbal stabilization, then back to making it walk faster.

jwatte
09-13-2015, 04:36 PM
OK, that looks cute!

r3n33
09-13-2015, 10:29 PM
I like it! and can't wait to see more. :cool:

Zenta
09-14-2015, 08:13 AM
Hi. Great progress! Have you considered designing the body using vertical plates instead?

jpieper
09-14-2015, 04:47 PM
Hi. Great progress! Have you considered designing the body using vertical plates instead?

Thanks!

So far I've only thought about vertical plates for the lizard configuration, where the HerkuleX mounting geometry doesn't make a single flat plate very convenient (as evidenced by the first iteration of this mech). For this configuration, I don't see a lot of downsides to the dual plate style: Manufacturing is trivial, it leaves lots of room to mount extra boards as necessary, and assembly and disassembly is really easy too. As shown here, I don't even have to mount the bottom plate unless I want a convenient place for a battery, or need to support the full weight of a mech warfare turret.

What are your ideas for how vertical body plates would be useful?

jpieper
02-01-2016, 09:14 AM
And now, finally, I have a working inertially stabilized gimbal! Neither it, nor the walking gait are necessarily tuned that well, but here's my youtube existence proof before I return to getting ready for Robogames!


https://www.youtube.com/watch?v=D4OAJK1vIN0

jwatte
02-01-2016, 10:43 AM
Walking and shooting. Scary!

r3n33
02-01-2016, 01:52 PM
Cool! :)

Slugman
02-01-2016, 07:23 PM
& with the addition of a single rangefinder & an IK calc you could even fix the vertical error caused by changing the distance to the target. OMFG! Humans are doomed..... Amazing work though. Scary, but amazing. :)

jpieper
02-01-2016, 07:53 PM
& with the addition of a single rangefinder & an IK calc you could even fix the vertical error caused by changing the distance to the target. :)

Our actual plan is to use the camera feed itself to keep locked onto a target during translation. We're doing the IP camera part ourselves, and our camera spits out simultaneously an h264 stream that we send over the wireless and an uncompressed stream for use on the robot. With any luck, I'll even get to it before the next event!

Slugman
02-01-2016, 09:11 PM
Ooohhh. Manually activated target-tracking.....
Yeah, still not reducing the scariness. :)

Gertlex
02-02-2016, 01:27 AM
I like it! Though I fear for the well-being of those servos :-/ Add some springs maybe?

jwatte
02-02-2016, 10:44 AM
We're doing the IP camera part ourselves, and our camera spits out simultaneously an h264 stream that we send over the wireless and an uncompressed stream for use on the robot

That's cool! I did the same thing for Money Pit 3 last year, on the Raspberry Pi (its camera/GPU integration is actually quite powerful, if you find the right bits to poke at it!)


still not reducing the scariness


https://www.youtube.com/watch?v=cR-YlZ9NdIA

jpieper
02-06-2016, 01:52 PM
And I've now got a more detailed write-up at: http://joshp.no-ip.com:8080/blog/robots/20160206-gimbal-stabilized-turret.html

MacroFab guilted me into writing it up... ;) They haven't paid me anything, but I figured I owed them some public recognition for their great work at incredible prices.

Gertlex
02-07-2016, 11:40 PM
Holy badassery.

jpieper
03-08-2016, 06:53 AM
We did a scrimmage last night between the two versions of Super Mega Microbot to check out how things were working. Here are the final two matches where everything was actually working.


https://www.youtube.com/watch?v=YVEcmF0pfRI

https://www.youtube.com/watch?v=JofAGUHCPrM

jpieper
08-28-2018, 09:12 PM
An update:

Our concept had been to have a moderately capable computer on board so as to do computer vision, and use 5GHz wifi to get HD quality video back to the control station. On paper that worked when we competed in 2016, but in practice it was problematic. We used an odroid U3 in the body of the mech, and connected a webcam and 5GHz wifi transceiver over USB. When it worked, it was awesome, however, USB being USB, meant that at the least opportune times, one of the camera or the wifi adapter would fall off the bus, which kinda made the thing pointless.

Fast forward 2.5 years, and the raspberry pi 3b+ came out, with integrated 5GHz wifi and a non-USB camera interface onboard. I've got a draft mechanical mounting of the rpi in the turret of super mega micro bot, all our software cross compiling for the rpi (it doesn't have enough RAM to compile it onboard like we did for the odroid), and a vision based tracker which without much tuning works reliably out to 15ft. This video only shows 10ft, but that's just what I could fit in the video frame.


https://www.youtube.com/watch?v=aeGUYlaC0Sg

jwatte
08-28-2018, 11:42 PM
Wow, welcome back Josh!

giantflaw
08-30-2018, 09:23 AM
Josh, love to have you enter again in 2019! Super Mega Microbot will raise the tech in the whole event.

GhengisDhon
08-30-2018, 10:49 PM
Good to see progress is still being made on Super Mega. We missed you at Robogames 2018. Hope to see you in 2019!

jpieper
10-20-2018, 10:23 PM
Another update... trying to see if I can build some more capable actuators for faster walking (maybe eventually running?): https://jpieper.com/2018/10/21/improved-actuators-for-smmb/


https://www.youtube.com/watch?v=H22vPpwXBJo

jwatte
10-21-2018, 06:01 PM
A fast high-powered brushless robot servo with built-in controller would probably be a nice product to sell as a niche market.
Just sayin' :-)
Also, without brushes, there's less that can go wrong, so it might be more robust, too...

jpieper
10-23-2018, 09:27 PM
Certainly could be more robust. Also a lot of new ways for it to get busted too... ;) I was putting 200W into this one for short periods and boy does that motor get hot quickly!

jwatte
10-24-2018, 01:23 PM
You have cold winters in Boston, so that seems like a benefit :-)

jpieper
12-01-2018, 09:21 PM
Slow but steady progress...


https://youtu.be/pVTEgMYdDoI

jwatte
12-02-2018, 12:33 PM
That's super awesome!

(Do you have access to internal company machining resources for making parts?)

jpieper
12-02-2018, 04:47 PM
Nope. The only machined part in the whole assembly is the piece of aluminum scrap holding the motors onto the guide rail that I butchered with a jigsaw, drill press, and dremel. Everything else is either off the shelf or 3d printed. All but the legs are shapeways parts, the leg parts I had done in Onyx on 3dhubs.

I'm torn how much I want to get machined to make enough for the first whole mech. I only have basic tools myself, so will need to send nearly everything out. There is only so much you can do with plastic. I'm considering trying out some of the fiber reinforced Mark Forged parts through 3dhubs to see how far that goes.

The big challenges are the legs, for which the Onyx probably isn't going to work once there is a 3rd axis and they have to support weight in the lateral direction, heat dissipation from the motor/controller, and the gears. Plastic is a terrible heat conductor and I can't really l leave anything exposed to flying projectiles. The off the shelf gears are a sub-optimal gearing ratio and awfully heavy.

tician
12-02-2018, 05:57 PM
http://www.etotheipiplusone.net/?p=4688

Onyx is plenty strong enough for full-combat beetleweight robots, so will likely be good enough for MechWarfare as long as you use proper wall thicknesses and infill settings and maybe add fiber if it's really needed. Would even work for the gearboxes if the teeth are large enough. Which reminds me that I still need to get my compound epicyclic gearbox printed - got put on hold a few months ago because of funding issues and my BuildOne is still likely several months away from delivery.

jpieper
12-02-2018, 06:54 PM
Yeah, I'm not worried about anything getting fractured by BBs with any of the materials, even shapeways nylon. I used Onyx for the legs because they were bearing load over a long lever arm. With at least the setup I have now, they are less rigid than I would like in the lateral dimension. I don't think I want to make them wide enough to achieve the same rigidity in both axes and ridges or webbing aren't really an option because of the link geometry.

tician
12-02-2018, 08:28 PM
Is the lack of rigidity a bad thing, given you have been making this thing jump about? Elastic elements are a very big deal when dealing with non-trivial masses/velocities and less than smooth motion no matter how nice you think your sensors and control algorithm are. A bit of flex or springiness is better than breaking teeth, snapping bolts, or deforming bearing surfaces.

Is the bigger issue not "bolting a leg to the face of a gearbox" instead of supporting it on both sides with bearings to prevent it ripping out the output stage of the gearbox?

jpieper
12-02-2018, 08:48 PM
In this case the gearbox is actually supported by two separate bearings internally on either side of a planet carrier, so it is plenty rigid and robust to torsional loads. However, the Onyx printed leg when fully extended can deform by +- 4cm laterally even when rigidly mounted on the top. My goal is not to make a series elastic actuator, but instead to have something that is accurately controllable with a high bandwidth while still being highly backdrivable.

And yes, my mostly non-existent sensors and control algorithms are very nice thank you very much! :)

jwatte
12-02-2018, 10:06 PM
Do you find a big difference between the "Onyx" black nylon versus Shapeways "Strong Flexible" in black? They're both nylon ...

Also, for SMMB I might be able to make some parts in aluminum for cost of materials + tooling, as long as they can be done in 2.5D + flipping and rush speed wasn't important. Send me an email if there's something particular that would be useful for.

jpieper
12-03-2018, 09:28 AM
Do you find a big difference between the "Onyx" black nylon versus Shapeways "Strong Flexible" in black? They're both nylon ...


They claim it is significantly stronger and stiffer than straight nylon. https://support.markforged.com/hc/en-us/articles/209934486-Onyx

Subjectively they certainly feel much stiffer than what I would expected from a normal strong and flexible print.

Guess I should get those parts done at Shapeways too and compare.

jpieper
01-11-2019, 08:32 AM
Continued progress, maybe slower than I'd like, but progress none the less... ;)


https://www.youtube.com/watch?v=PHQ1jWSkDx8

jwatte
01-11-2019, 09:47 PM
That is freakin' scary!

GhengisDhon
01-13-2019, 08:21 PM
Jump Jets!

jpieper
03-07-2019, 06:29 AM
Continued progress... I'm probably about 85% likelihood that I'll be able to get a mech with these legs going in time for the Maker Faire. All the parts are either here or expected to arrive in the next week.


https://www.youtube.com/watch?v=9AXlkIq_FQM

jpieper
04-26-2019, 09:55 PM
Well, I'm working super hard to make this happen, but as is the way of things it is all coming together at the last minute.

I've got at least a little progress to report, in the form of actual walking:


https://youtu.be/xyNiaPdQuIU

Of course, it needs assistance to start or stop, is wobbly when it walks, has no battery, no turret, has unprotected innards, has an affinity for slippery rulers, and is awfully big so I'll have to be aiming down a lot, but hey, it is something! With luck, it will even be able to jump. ;)

jwatte
04-27-2019, 12:24 AM
That looks cool!

I wonder if you could crouch the legs more and get lower height? Or would that overheat the tibia motors?

Can you strafe? Perhaps mount the gun underslung from the body?

jpieper
04-27-2019, 07:31 AM
Yep, the legs definitely have room to crouch more, even within their current thermal and power limits. I did these first tests up high largely for stability. I still have direct drive motors on the upper and lower leg, which while they have great compliance and torque control, only have so-so position control bandwidth. Crouching more resulted in them sagging a lot when some legs were in flight, which made the walking not so successful. I'm going to just feed-forward the correct torque for any given static position, and am also going to eventually get gearboxes for all joints. The gearboxes are vastly superior, 6x better position bandwidth, 6-12x higher peak torque and 36-144x less power for the same torque, only minimally less compliant, and only twice as heavy. Mostly it just depends upon time management at this point. Once the last part on the slow boat from aliexpress arrives, I'll have enough parts for a full set of gearboxes, but they currently take me about a day of careful assembly time each, which is kinda cutting it close at this point. Besides the fact that I need to redesign the legs!

I'm figuring the more successful strategy at this point is to get all the things necessary for an actual match working in some minimal, if suboptimal way, then continue with iterating on performance improvements. At that point, feedforward torque on the direct drive motors is probably the best I can hope for.

Slinging the turret under the body is a really interesting idea. Unfortunately, because of aforementioned time pressure, I suspect I am stuck with my existing hardware, which wouldn't be super easy to convert from a top mounted position.

jwatte
04-28-2019, 08:24 PM
And I thought I was cutting it close ...

jpieper
04-29-2019, 08:34 AM
And fortunately, the new gimbal control board seems to have functioned correctly on the first try. I'm basically out of time for another revision, so that was good!


https://youtu.be/47oez2-oWc8

cire
04-29-2019, 09:16 AM
Everything about this robot is amazing. What changes did you put into the v2 of the turret?

jpieper
04-29-2019, 09:33 AM
Not much functionally in this revision of the board... it supports the new 5S bus voltage and the RS485 based communcations protocol. It no longer has USB pass through, as the new turret has an rpi3 mounted in the gimbal platform with camera directly attached. Also, it has a power pass through to power that rpi3 without needing a crazy wire harness and the weapon power connector is positioned better to allow more freedom of motion.

I think it ended up being bigger too, but that was only because of an aborted attempt to use the Intel Joule 3 years ago and I didn't want to spend the time to shrink the footprint back down again...

I've got a blog post queued up for later in the week with a few more details.

jwatte
04-29-2019, 12:14 PM
Wait -- jumping AND stabilized gun?

I've been thinking about an RPi 3 for a long while, too. It even has solder spots for taking out a composite video signal, to use low-latency FPV video. That would let me overlay all kinds of interesting information on the display, as well as use the Pi to do target plate tracking. Really excited to see you go this direction -- really looking forward to it!

jpieper
04-29-2019, 12:19 PM
Well, like most tortoise and hare stories, you know how this will end. ;)

Yeah, I've got the pi3 doing aruco target tracking using whatever random technique is built into opencv. Without too much optimization, it can track over a few hundred pixel square image, which covers most of the HD feed I'm sending back.

Also importantly for the pi3, is that it has a camera connection, and built in 5.8Ghz wifi, neither of which uses USB internally.

jwatte
04-29-2019, 09:04 PM
It even has a GPU that can be fooled into doing some useful things (including compressing H.264 in hardware for storing for later offline analytics.)

jpieper
04-29-2019, 09:19 PM
Yep, I've got the hardware encoder compressing to h264 with low latency for streaming over IP...

jpieper
05-14-2019, 07:00 AM
Because my robot wasn't big enough already...

7521

20lbs and counting... Fortunately, both for everyone else and me, it is at best going to be a lumbering giant for the weekend.

Still have to get target panel mounting in the front and make it a bit more robust at walking.

Gertlex
05-14-2019, 09:39 AM
Now that's a tease! (There's nothing to see there other than a grey restricted icon thing, it seems)

20 lbs is insane sounding :D

jpieper
05-14-2019, 09:46 AM
Heh. Apparently I can't work the forum. Post should be updated with an actual picture I hope!

jwatte
05-14-2019, 11:57 AM
That's .... gotta be a battery drain :-D
Are you using the power tool pluggable batteries?
That means you'll have an easy time swapping!

jpieper
05-14-2019, 12:03 PM
It is actually relatively efficient at just walking. The motors are all sized for running (although there is no way I'm doing anything like that soon). Just walking, my power tool battery should be good for something like 2 hours.

jpieper
05-14-2019, 06:34 PM
Now I can walk around and shoot for 10 minutes, that means I'm ready right?

spoiler alert... 10 or so minutes of *really slow* walking interspersed with a few BB shots


https://www.youtube.com/watch?v=sZfvbiWGTc4

The sad/funny thing is that SMMB1 is still here, perfectly functional, 1/3 the weight, and also has the upgraded turret with auto-targeting. It can probably walk twice as fast. That said, I've put enough work into this new one that I'm going to bring it and have what fun I can!

FWIW, the biggest speed challenges I have now are that I discovered this morning that my 3d printed timing belt pulleys can't reliably support the weight of this new beast doing anything remotely dynamic. Next step there is to either machine my own out of aluminum, or post-machine one of the stock ones. Not going to happen for this weekend though.

GhengisDhon
05-14-2019, 07:03 PM
Looking good. Be interested in seeing the up gr... err down grades.

We've been discussing what you've been doing with your motors, if they are direct drive, have gearing, or what. The jumping was a bit scary. Can't wait to see it in person.

byi
05-14-2019, 07:05 PM
At least it sounds really cool?

Seriously though, very cool to see a different approach working well.

jpieper
05-14-2019, 07:29 PM
Looking good. Be interested in seeing the up gr... err down grades.

We've been discussing what you've been doing with your motors, if they are direct drive, have gearing, or what. The jumping was a bit scary. Can't wait to see it in person.

This version is all gearboxes, 6x reduction planetary. All the previous jumping videos were with only direct drive motors. The gearbox versions have probably 12x the torque and still can go *really* fast. If the legs held up, I think the gearboxes could probably make the whole robot jump a foot or two (or maybe even more) into the air. With a little more work on the thermal management (right now I have the hot things mounted directly to plastic), it could do so continuously.

In my "bonus" time tonight and tomorrow, I'm going to try and assemble one or two spare gearboxes and legs that can be for show and tell.

jpieper
08-22-2019, 12:26 PM
Not dead yet...


https://www.youtube.com/watch?v=E3QCTPOWj1c

This used *very* conservative parameters, as I don't really want to break things until I've done more testing and have more spares. I wouldn't be surprised if it can go 10-20x higher.

jwatte
08-22-2019, 01:15 PM
Nice!

Don't show it to my wife; she's afraid of big jumpy crawly things ...