PDA

View Full Version : [Project] Steam Automaton Groucho



DangerousThing
04-08-2014, 03:09 PM
Hi,

I'm announcing this in hopes it can speed up my design. :)

Part of the reason for building this is to give me a reason to learn ROS. Hopefully what I learn will appear on my blog once i set up a robotics area in there.

The base for SA-Groucho is a very used Jet Power Chair. My thought is to remove the upper parts and use the base as is (mostly). The upper part will be constructed and will go approximately where the chair went.

Much of the exterior will be decorated with 3d printed steampunk items. This will be the first bot I've built that I want to paint.

The head will be small and oddly shaped, though I will do my best to put Groucho Marx' trademark features as much as I can, renvisioned for steampunk. However, the neck will be a robotic arm as long as I can make it, though that depends on my budget at the time. For now it will just sit in his neck like a normal head.

Most of the computing power will reside in his torso. I have ordered a NUC and will use some BeagleBone Blacks for some control areas, along will some various Arduino clones I have. Perhaps if the Arduino Tre is inexpensive enough I could find a use for it.

I plan to build him in stages so that I don't take on too much at any one time.

SA-Groucho will also be the head of a Troupe of smaller bots, including Zeppo, Harpo, and Chico. Most of these will be wirelessly controlled from SA-Groucho.

His purpose is mostly entertainment and art. I do plan on being able to take him for fairly long walks in my area just to see the reactions.

Currently I have most of the parts except for the power system. Due to cost I'll probably use large amounts of SLA batteries. The ones in the wheelchair are a pair of 12V 75Ahr SLAs that are quite dead. I'm debating on replacing them with SLAs or with LiFePo4s. The one is lighter, the other cheaper. Plus the SLAs are heavier which will lower his center of gravity.

Once I get some pics I will put them up here.

The rest of the Troupe is even less well defined, though I intend for both Chico and Harpo to be Hexapods, while poor Zeppo will be a simple rover. All the bots will have sound, though Harpo will only be able to communicate through music and cartoon sound effects.

jwatte
04-09-2014, 11:16 AM
From my experience, I'd recommend LiFePO4.
Sounds like a fun project :-)

DangerousThing
04-09-2014, 03:43 PM
Question: Would you run the electronics off of a 12v source or a 24v source?

Also, would you use two batteries (one for the motors, one for the electronics) or a single battery for both? If I use a single battery it makes charging easier. However, with two batteries I can choose the most convenient battery for my electronics. I think +19 and +5 should handle most of the electronics.

Given 24 volts, though, how is it best to get high amperage 12v for my MX-64s for the arms and neck?

tician
04-09-2014, 08:13 PM
There are plenty of dc-dc converters that could drop the 22.4~28.8V of an 8S LiFePO4 down to a pretty well regulated 12V. There are TI 12~22V/6A converters at ~$23, Murata 12V/10A and 12V/17A converters at ~$50 and ~$70, respectively, and Delta 12V/17A at ~$63, but getting non-'board mount' converters pushes the price up a lot. There is an even wider selection of dc-dc converters to drop a solid 12V (or variable/noisy 18~36V) down to 3.3V/5V for all the embedded controller and sensor needs. Stepping a 3S LiPo (10~12V) up to 19V for a laptop or NUC/ZBox is a little less common so often more expensive, but there are several adjustable dc-dc converters mentioned on other threads that can handle a NUC/ZBox reasonably well.

12V/5A is all that is needed for a DARwIn-OP with 20 MX-28 servos, but it will depend on which servos you choose, how many of each, and how heavily they are loaded. The servos are much less picky about the voltage tolerance of the 10~12V supply than computers and sensors with their 19V and 5V supplies, but any drops below ~9V because of over-current conditions in the power supply will cause issues.

DangerousThing
04-09-2014, 10:48 PM
OK, I'm thinking then of using two LiFePo 24v batteries. The motor battery will be a 60 Amp-hr, and the electronics should be able to make do with a bit less - maybe a 30-40 amp-hr.

As for the motors, there will be:

1. A pair of wheelchair motors, 24v
2. A "waist" motor to rotate the torso.
3. The neck servos. This will be very similar to an arm in construction, except a bit longer. I'll be using MX-64 servos.
4. Eventually there will be two spindly arms, but I'll only start with one.

I can use 24v motors for the waist and use a separate 12v converter for each set of servos.

This also simplifies the charging station, because I will have, in addition to the chargers, 2 24v supplies to make the electronics and smaller motors happy during charging.

The electronics are a bit less thought out at the moment.

1. The main brain: an Intel NUC.
2. (Deleted for being too silly even for me)
3. A BBB to deal with the head itself.
4. A BBB to deal with each arm.
5. A wired/wireless router to enable the various parts to communicate via Ethernet.
6. The wireless part will allow Groucho to communicate with the rest of the Troupe without depending on my home's wireless network.
7. Various sensors including two or three webcams and a boatload of Sharp IR Rangers and sonar units.
8. Some microcontrollers and LEDs for various decorations.

Plus other things I can think of. I'll probably have some pistons that actually seem to do something.

jwatte
04-10-2014, 11:03 AM
If I were using step-down converters from 24V to your 19V and 5V targets, I would not bother separating the batteries. I would put a high-power TVS (5KP26, 5KP28, perhaps even 15KP26 or -28) across the input to the step-down converters, and perhaps a big electrolytic at each (1000 uF or more) and call myself good. A single, bigger, battery, greatly simplifies charging, and makes best use of my money to get the overall runtime up, and balanced between electronics and motors.

If you use Robotis or Dongbu servos, those are controllable by serial bus. Thus, you could use a single USB2Ax or similar adapter from the NUC to talk to the servos of both arms, no BBB(s) needed.

DangerousThing
04-10-2014, 04:31 PM
I can do the larger battery, but it is more of a pain. On the other tentacle I would have probably had to redo the battery tray anyway. The advantage of using a large battery is that it forces the CoG lower which makes going up and down my driveway more than a bit hairy. However, SA-Groucho should mass less than myself and I can make it down the driveway on Lee's powerchair.

tician
04-10-2014, 04:52 PM
Forcing the CoG lower to the ground (preferably below the wheel's diameter) decreases the chance of tipping over even with a large, unbalanced load on top. Overall weight is what affects traction and motor loading.

DangerousThing
04-10-2014, 09:49 PM
Thanks, Tician

Unfortunately modern powered wheelchair wheels are fairly small. And I'd like to keep the battery tray about the level it's at, though I'll probably move it to the back for more stability. For this, I may have to do some math.

I'm trying to balance the torso which sticks straight up. During normal travel, it isn't much of a problem, but our driveway is around 12 degrees, which doesn't sound like much until you take it in a power chair.

DangerousThing
04-17-2014, 03:28 PM
I would like to make a clarification. I've heard some people are wondering about how many projects I have and that I may be posting more than I should. I've been asking questions because I've been out of touch in robotics for a decade when I totally dedicated myself to taking care of Lee (my wife). I'm still her full-time caregiver but I realize I have to take time to do the things I need to do to keep sane and keep myself healthy as well.

My main project currently is building one robot (SA-Groucho). Groucho is part of a set of four robots, also including Zeppo, Harpo, Chico, and maybe Gummo; I will be probably switching back and forth between these robots as my finances and interests imply. In addition there are other small projects that aren't really directly a part of a robot, but associated. This would be the laser harp I've been asking about; after all, Harpo Marx needs his harp...

I'll admit that I'm going in a lot of directions all at once, but there *is* a madness to my method. Right now I'm in the design stage, so I'm still trying to work out what's going where. Sometimes I'm probably going to post in the wrong place. I do try to help people when I can. I also tend to ask questions where the forum seems to direct them. I do ask your forgiveness when I do things incorrectly. I will try to ask questions specific to Groucho in this project thread, but more general questions are likely to go anywhere.

Thanks for listening,

D. Jay Newman

DangerousThing
04-17-2014, 05:01 PM
Groucho's Status:


Removed most of the extraneous parts of the wheelchair. If anybody wants the original electronics, I'd be glad to send it to them for the cost of shipping. I still have to get a big honk'n LiFePo4 battery.
Built the torso structure. I still need to break this 2 foot cube into layers. For this I plan to use drawer slides: either buy some with ball bearings, or print some nylon ones.
My NUC arrived, but I exchanged it for the same motherboard in a larger case so I could put a notebook SSD as well as the mSSD card. This extra space will mostly be used as a shared disc by my Troupe of robots. Unfortunately the new one won't be here until the end of next week. And since we'll be away, it may mean until the following Monday.
I have most of my basic sensors including a Kinect and a bunch of sonars and Sharp IR Rangers. I also have GPS, accelerometer, magnetometer, and gyroscope. There are still some i need to get.
I have some MX-64s on the way for one or more arms or necks. I haven't done the math on the arms yet.
I am still awaiting my 3d printer for some of the mechanical stuff as well as the steampunk decorations.
All of the programming needs to be done. I may try to install ROS on a BeagleBone Black so I can start leaning it if I have time before I get my new NUC.


Some hints on the other members of the troupe are:

Harpo: A hexapod, probably a Phantom.
Zeppo: Just a bit more than your basic Hello World robot. But not much more.
Chico: A two wheeled balancing robot. Perhaps even a one wheeled one.


Other projects I'll be working on in between are some musical instruments, starting with the laser harp, and continuing with a small piano made from a very young child's piano (think Schroeder from Peanuts). I might make it from scratch and make it a steam organ instead.

DangerousThing
04-20-2014, 03:53 AM
I found my container of sensors that I bought and used up to 2003, and I found that most of what I'm using is the same sorts of things that I used then. A bunch of SRF04s, a lot of Sharp IR Rangefinders of varying distances, and 16 SFR08s (the I2C version of the SFR04). I'll take pics later!

And I know there are more to be found in my basement. Scarey but true, I haven't unpacked the basement in four years. Now that would be a good productive use of a robot!

DangerousThing
04-28-2014, 05:18 AM
Here's Pi in your eyes Groucho.

I was planning on using webcams, but I'm thinking of going a different way. A RasPi can easily connect to a camera module and Ethernet. Also, the RasPi should be able to do some preprocessing of the image before my main computer gets to it. Yes, I would prefer to go for a really high quality gigabit Ethernet camera, but I'll take what I can get.

So the main head will have three RasPis and their camera modules (two setup for stereo and a third one for IR vision). But first I'll just try with what I've got which is a single RasPi and camera module. I'm curious what I can make it do if I twist it's tail a bit.

One the RasPi Compute Module becomes available I should be able to make the heads smaller. Yes, this means I want to have multiple heads. I figure the main head will have the three eyes mentioned above, and two heads on either side with a single eye apiece. Each head will have an accelerometer/gyroscope combo to aid in helping the heads to know their their position; as well as the information from the servos controlling the necks. This should be enough to keep the cameras in alignment as long as I double-check the data from the images to make sure the positions are consistent.

The positioning should allow for a wider stereo view if I need it. Or to have multiple objects being watched. Or perhaps just to have some non-verbal communication and emoting.

In addition, I may have some cameras elsewhere to see if I can supplement traditional odometry with vision.

And since I'm thinking about vision, I'll just throw this idea out there. The normal human eye has one area of the eye that is high in receptors, called the fovea. The are is really where we do most of our sight, but it is very small. We just move our eyes around without noticing it normally.

To duplicate this, take one camera module for the center. Surround this module with six other modules for the next layer. These can have about half the detail of the center, but their fields of vision should each overlap the center and the ones next to them. Then put a third circle containing around 12-18 modules around the second circle. This would probably contain IR modules, and the detail could be lower than the second circle.

So for each circle, there are more modules, but less detail, so less computation per module. Of course, right now I couldn't imagine paying for this, or even how I'd go with linking the computers controlling the modules.

But i do think its an interesting idea. Maybe someday. Or somebody else will do it.

tician
04-28-2014, 05:57 PM
http://disp.duke.edu/projects/AWARE/

I'm kinda partial to designs combining multiple bare sensors in a contoured array illuminated by a single wide-angle lens assembly (panomorph is pretty awesome) similar to Kepler's ~95MP array. Although there are gaps in the composite image caused by the spaces between the sensors, it is simpler and much more compact than the optical assembly required for accurate alignment of the independent camera module images in gigapixel cameras.



One of my lurking grounds was invaded by quotes from yet another IDiot claiming the human eye to be a supremely designed work of perfection that could not possibly be the horribly flawed p.o.s. we know it to be. When confronted by the flaws we know exist, it then claimed those flaws were due to our lack of understanding of some unknown tradeoffs we may discover in the future that will prove the design to be perfect. When tracked to the source, its friend was one of the few people I've ever seen claim that light does not travel and that vision is caused by rays leaving the eye, thus completely refusing to accept reality. ...someone please stop the planet, I want to get off.

DangerousThing
04-28-2014, 08:20 PM
http://disp.duke.edu/projects/AWARE/

I'm kinda partial to designs combining multiple bare sensors in a contoured array illuminated by a single wide-angle lens assembly (panomorph is pretty awesome) similar to Kepler's ~95MP array. Although there are gaps in the composite image caused by the spaces between the sensors, it is simpler and much more compact than the optical assembly required for accurate alignment of the independent camera module images in gigapixel cameras.

Thank you for bringing that to my attention. I love the concept, but bit it's a bit big even for Groucho, though maybe if I were building Gigantot
:)

But it's wonderful for the uses to which it can be put. A simplified version would make a nice robot camera though.


One of my lurking grounds was invaded by quotes from yet another IDiot claiming the human eye to be a supremely designed work of perfection that could not possibly be the horribly flawed p.o.s. we know it to be. When confronted by the flaws we know exist, it then claimed those flaws were due to our lack of understanding of some unknown tradeoffs we may discover in the future that will prove the design to be perfect. When tracked to the source, its friend was one of the few people I've ever seen claim that light does not travel and that vision is caused by rays leaving the eye, thus completely refusing to accept reality. ...someone please stop the planet, I want to get off.

I understand. I ran into a thread where the OP believes that all creatures will evolve into humanoid form with human intelligence. Of course, the was on a role playing system forum, so a bit of fantasy is OK. Unfortunately misunderstanding evolution is one of my pet peeves. I make no claim that the human eye or any other system is even close to perfect.

tician
04-29-2014, 12:59 AM
The 'light field camera' is another rather concept that might be of interest. It is apparently not that difficult to implement (http://www.umiacs.umd.edu/~aagrawal/sig08/BuildingLightFieldCamera.html)with a printed mask placed over the sensor, and permits you to adjust the focus anytime after the image has been taken. It has also been used for creating 3D depthmaps, but there seem to be fewer freely available papers detailing that process.



How I wish I had cephalopod eyes with their movable lenses for focusing (like a camera, instead of deforming the lens like most vertebrates), receptor cells pointing the correct direction, and the ability to perceive the polarization of light. The eyes of the mantis shrimp would be a close second with its 16+ pigments for color and polarization, though the trinocular compound eyes mean lower visual acuity. Stuff the mantis shrimp receptors into the cephalopod eye, rewire my visual system to permit simultaneous perception of all receptors (instead of blending signals before the optic nerve), rebuild the nictitating membrane, and I would be "SQUEEEEE!!!!!"

DangerousThing
04-29-2014, 01:23 AM
The 'light field camera' is another rather concept that might be of interest. It is apparently not that difficult to implement (http://www.umiacs.umd.edu/~aagrawal/sig08/BuildingLightFieldCamera.html)with a printed mask placed over the sensor, and permits you to adjust the focus anytime after the image has been taken. It has also been used for creating 3D depthmaps, but there seem to be fewer freely available papers detailing that process.

Interesting. A very simple solution to create a 3d camera.

jwatte
04-29-2014, 12:21 PM
Light fields basically add two (!) more dimensions to the data that's captured, so your storage requirements do go up a fair amount.
The good news is that each of those separate dimensions are typically largely correlated, which can be used to significantly improve average compression ratios.

DangerousThing
04-29-2014, 04:47 PM
I'm keeping the light fields in mind for later. Currently I'm planning on going with a three camera head (two for stereo, one IR for night), plus two smaller heads mounted on "arms" each with one camera (or maybe later one normal and one IR if the IR proves useful). This will allow Groucho to have a large stereo baseline using both auxiliary heads, or to examine alternate objects of interest.

Or, to be completely honest, because multiple heads look cool. :)

KevinO
04-29-2014, 05:11 PM
How much of the original Groucho you did in 2009 are you going to keep? The four deck design using the Zagros REX 14C base.

DangerousThing
04-30-2014, 04:58 AM
How much of the original Groucho you did in 2009 are you going to keep? The four deck design using the Zagros REX 14C base.

It was closer to 2003. Basically after Lee got that sick, I had to quit making. Now I realize that was a bad decision because without being able to build things, I go a bit crazy. So I'm restarting

Unfortunately most of the old stuff is packed up somewhere in the basement. However, I found the Zagros base a few weeks ago, but not the motors I had on it. I did pack up the sensors and I found most of them. Plus I was able to buy some more Sharp IR Rangers. I think I even found the motherboard I used. Unfortunately I have no idea where the other levels or the motors are. But luckily I did get a few things from eBay: a Kinect, some laser modules, and the range finder from a Neato.

When we moved into this house, Lee was in the hospital so I just had the movers throw everything into boxes.

And now, five years later, I'm getting around to unpacking things.

DangerousThing
05-06-2014, 07:56 PM
Arrrrrgh!

I installed Ubuntu 14.04 on Groucho. I can't installed the current version of ROS (Hydro) because it doesn't support 14.04. And I can't install the latest version of ROS (Indigo) because it's not released yet (Real Soon Now).

Oh well, just waiting I guess. Now I need to find an old web cam so that I can experiment with OpenCV. Do I go with the latest version from source, or do I use the older packaged version? Gack! The Linux I remember everything but the basics was installed from source. I even remember having to triple compile the tool chain: the first time with the old compiler, the second time recompiling everything with the new tool chain, and a third time to see if the second and third compilers were identical (if they were, then you were done with the compiler).

And most of the stuff I wanted back then had their own sites, like Perl, and Python and such.

And I just got a USB flash card writer to write disks for my RasPis, because my notebook's built in writer doesn't do it correctly. Here's hoping that the new gadget I got will do this. Then I have to figure out how to make a networked RasPi into a fake USB camera.

tician
05-06-2014, 09:03 PM
Indigo is available (http://wiki.ros.org/indigo/Installation/Ubuntu) for 14.04, it just has not been 'released' yet as a stable version for widespread consumption (there will likely be lots of updates to make it a bit more polished before the release date). The packages might be marked as 'beta' in apt or synaptic, but should not require any changes to the repo information to later upgrade to 'stable' packages when indigo is officially released.

There are a few ways to access the RPiCam, but none of them were ideal when I last checked. There is the 'userland' repo (https://github.com/raspberrypi/userland/) released by the RaspberryPi foundation that provides low-level access to the GPU for controlling the camera and retrieving still/video data, but it is a bit more complicated than a v4l2 interface. There was talk of a true v4l2 driver for the RPiCam, but I've not checked recently to know if it exists. There is a user-level driver that grabs images from the 'userland' raspistill or raspivid program and send them to a fake-v4l2 device, but it requires binary downloads from a place I'm not sure I trust. There is also a python implementation of the userland library named picamera (https://github.com/waveform80/picamera) that might be useful if familiar with python and its socket interfaces.

tician
05-06-2014, 09:18 PM
Installing Xubuntu-14.04 on the Zbox now, so will test the Indigo install once that is finished. Don't have any webcams or kinects on hand, so cannot test that yet.

DangerousThing
05-06-2014, 10:30 PM
I tried to install things last night (I had found the Indigo install page), but following the instructions exactly got me a package not found message. I tried to do the full desktop download.

I have found a bunch of web pages that give instructions for getting the Picam to play nicely with OpenCV but as you say they use v4l2. That makes sense.

I read that the details of the graphics co-processor was released last month. Hopefully that will make things easier eventually.

A new question: is it better to use the desktop or the server version of Ubuntu for a robot?

jwatte
05-06-2014, 10:44 PM
The difference between desktop and server is mainly in what packages come by default. You can add a window manager to the server install with something like "sudo apt-get install xfree86 lxde"
If you won't be hooking mouse, keyboard, and monitor to your robot with any frequency, then taking up CPU/GPU/power/RAM with a desktop environment doesn't make sense, though, and you're better off using the server install that doesn't start X.

tician
05-06-2014, 10:49 PM
If the robot is the only system with ros installed, you will definitely want a desktop system hooked up to a monitor while you are learning it. If you are using multiple systems to learn, then you should be fine installing a desktop on the primary development system with ros-desktop and leaving the headless robot with ubuntu-server.

Took a while to purge all the ubuntu software center, upgrade and update notifiers, and annoying media programs from xubuntu. The various 'installation' metapackages are not appearing in aptitude, so probably will not be available until launch day. You should get most of the basic desktop install with the 'ros-indigo-ros-tutorials', 'ros-indigo-ros-comm', 'ros-indigo-rqt', and 'ros-indigo-rviz' packages. It does not include cameras, arm control (moveit), base navigation, or robot specific files, but it should be enough to run all of the tutorials.

Really wish I trusted SSDs enough to put one in my netbook instead of just the bot-bound zbox.

KevinO
05-06-2014, 11:10 PM
Indigo isn't released until later this month. Like, latter part of the month. :)

DangerousThing
05-07-2014, 05:43 AM
Really wish I trusted SSDs enough to put one in my netbook instead of just the bot-bound zbox.

I really like SSDs. I've never had a problem from them. I have two of the on my laptop and it runs like a charm. Groucho has the topside slot filled with one and has a 1 TB SSD on the bottom. I'm working on a better backup solution than I have now.

jwatte
05-07-2014, 01:33 PM
Really wish I trusted SSDs enough to put one in my netbook instead of just the bot-bound zbox.

And you trust spinny platters with micron-size magnetic matter more than flash chips?
The truth is, there's just as many buggy spinny disk product lines as there are buggy SSD product lines.
And when they work as intended, SSDs are much higher MTBF/durability.
The time of SSD has come, and spinny disks are relegated to long-term bulk online archives.

tician
05-07-2014, 02:46 PM
Spinny magnetic stuff not get broken by writing process as fast as non-spinny memory stuff. 8GB spinny thing in thin thinky thing from 2001 survive 12 years of torture by many users. 160/320/640GB spinny storage things from 2005/2007/2009/2011 and mobile 500GB spinny thing from 2011 all chug along nicely, while 8GB USB3 storage thingy from 2013 die in 6 months of little use. While 32GB non-spinny memory thing from late 2013 very fast, it used little because so small, and bigger non-spinny memory things not in reach of poor, learning flesh sack with broken thinky meat.

DangerousThing
05-07-2014, 08:18 PM
Considering that the flash memory used in USB sticks and SD cards and mini-SD cards are mainly supposed to be used to transfer files (such as images from cameras) rather than be used as hard drives, I'm surprised all those mini-Linux boards work for more than a few weeks on a card.

As for your brain, I think it must be composed of non-spinny stuff like the rest of us. Though sometimes it *feels* like mine is spinny. :(

jwatte
05-08-2014, 11:50 AM
sometimes it *feels* like mine is spinny

Are you sure that's not the Tequila shots from Cinco de Mayo lingering? :-)

I agree that USB sticks and cards are not super durable for writing. But properly wear-leveled SSDs from reliable providers (Intel, mainly, except for that one bad batch) have performed great for us, for many years, including in production database servers that re-write all of storage on a regular basis.

Yes, in the glory days of good mechanical engineering, some spinny disks were made that last a long time before the inevitable head crash. With increasing density, that MTBF is going down, and SSDs are better. Even so, for laptops and robots, I want no moving parts other than the actuators if I can help it.

Anyway, SSDs are coming down in price and up in capacity (8 TB SSD from Samsung due soon) -- the days of the spinny discs are numbered. You can't even get an 8 TB magnetic clunker.

DangerousThing
05-08-2014, 01:50 PM
Anyway, SSDs are coming down in price and up in capacity (8 TB SSD from Samsung due soon) -- the days of the spinny discs are numbered. You can't even get an 8 TB magnetic clunker.

The largest I've seen is 1 TB. I'm curious about the price of an 8 TB disk. Luckily I don't need one yet because I don't think I could afford it yet. :)

KevinO
05-08-2014, 02:28 PM
Groucho has the topside slot filled with one and has a 1 TB SSD on the bottom.. Why such a large hard drive for a robot?

DangerousThing
05-08-2014, 04:44 PM
Because Groucho will store data from the other robots of the Troupe (Harpo, Chico, and Zeppo). Also i plan to do some visual experiments and I may want stored feeds from all the cameras so I can figure out what went wrong.

Groucho will be large enough that he can store enough batteries to keep a lot of things. He's an art work as well as a robot. At this point I just set up his brain as a small computer. Right now I don't really know what I'll want to store. And I can always get a smaller drive if I'm not using this one to its fullest.

A long while ago I was the computer geek for my unit as Penn State, and I made a recommendation for hard drives based on what I thought was a reasonable increase in usage. We had to buy extra hard drives for everybody because I made a mistake. I find that it's better to get more than I need right now if there is even a possibility of needing it.

jwatte
05-08-2014, 07:12 PM
I found that if you capture live streaming video from a HD stereo camera at 30 fps, available disk space gets eaten in a hurry...

DangerousThing
06-18-2014, 01:53 PM
Sorry, I've been away and not working on robots. I'm hoping to help set up a robotics club for children and young adults, but that's up in the air now. I've ordered more Groucho parts from eBay, including another wheelchair frame, motors, and wheels. I then used up most of the rest of my budget and bought a laser rangefinder from RobotShop.com, a Hokuyu URG-04LX-UG01, as well as getting some basic woodworking tools.

I should be getting a bunch of packages (see above) starting today.

We took a lightning hit that fried our computers and the DSL modem and parts of the entertainment center. Luckily Groucho's brain wasn't effected. So now I have to put together a NAS to backup important files before I get to work on robots. Then I am going to have a go at fixing Lee's computer. Once I get a second computer I'll be using that, until then I'll be using a BBB. The NAS will connect wirelessly because the lightning came from the DSL modem through the ethernet wires.

And then I have to clean house. Sometimes I wonder why I got married. :)

More robots. Real Soon Now!

DangerousThing
07-07-2014, 08:16 AM
I now have two different wheelchair frames and a battery that doesn't fit either of them. So now I have to figure out how to modify one of the frames or else make a new frame from some 8020 aluminum extrusions and plywood.

Modifying a frame will probably leave its strength intact.

Building a new frame has the advantage of looking like I want it to look, and I think I can make it strong also. I'd also need some curved areas which I might not be able to do with my current knowledge and skills. On the other hand, I've rarely let that stop me before.

Decisions, decisions. I think it will come down to how easily I can attach the motors to a plywood box. I *really* wish that I hadn't let Lee talk be into getting rid of my CNC mill. It would be very useful right now.

DangerousThing
07-29-2014, 05:47 AM
Quick update: I am redesigning SA-Groucho to be a two-wheeled balancing robot (AKA a Mobile Inverse Pendulum Robot). This will allow him to fit better in our home.

However, before I go all crazy with a 5' tall behemoth powered by wheelchair motors, I'm going to be making a prototype Mini-Groucho with some smaller motors using only 12 volts. Mini Groucho will only be 2-3 foot tall and have a lot less weight. I'll be powering MG with a 12v 10 amp-hr battery, and for the electronics I'll be using a cell-phone charger battery for the electronics.

I'll probably be using a Teensy 3.1 for the brains.

This is mainly to test the theories regarding control of a MIPR. I have the parts for this I only have to construct the motor support from thin sheet plastic cut and glued together. I'll also have to cut some scrap 8020 extrusions to fit. So a couple weeks work and I'll have a bot that can fall down easily. And yes, I will be making a blooper reel. :)