View Full Version : Servo Control for perspective project
08-15-2006, 07:41 AM
I have recently just finished an engineering course where we got to play with the BOE-BOT for the class project. It was really fun, and easy to program because I had already and experience with TI-BASIC.
The robot I want to do is a build a spider. What I am wonder is how would i go about controlling the 58 servo's required to leg movement? 2 B.O.E. in asych to split the work or what?
The 8 long legs requies 6 servos to function.
The 2 short legs beside the fangs require 5 each.
I DO NOT CURRENTLY OWN a Stamp, B.O.E, OR the 58 servo's. I have to save up the money for this project. The high torque servos are rather $$, and that's not even including the hardware for the exoskeleton.
It will contain the following once completed:
2 6-Axis robitic arms
RF or Bluetooth
Piezo Vibra Tab(s)
I will post a 3D model fron AutoDesk Inventor once I have completed my sketches and begin modeling it.
08-15-2006, 09:36 AM
The Parallax Serial Servo Controller will handle 16 servos. The processor only needs to provide 2 I/O pins for a serial input/output connection. Two Servo Controllers can be daisy chained to control 32 servos. If you devote 4 I/O pins, you can have 4 Servo Controllers that can handle 64 servos. The advantage of off-loading the servo control is that the main processor only has to worry about what position the servo should end at and how fast it should get there. There's a "set it and forget it" type of functioning.
You've got a lot of sensors and other actuators on your list as well. You're clearly going to need a bunch of processing power just to service all the bits let alone any "making sense of the data". You might consider using the Propeller instead of a Stamp. Normally I'd suggest using the Propeller for everything, but you have so many devices planned that there's just not enough I/O pins available. You could devote a Propeller for each group of 28 servos, but that's overkill in terms of processing power. You could build your own I/O extender for use with the servos or you could just use the Serial Servo Controllers since they'd do the job well, are already developed, and are similar in cost to a Propeller/Board package (about $80 for 28-32 servos). You still might need two Propellers because of the number of sensors, but you might consider learning the SX-48 and using Parallax's SX-48 Prototype Board. They're only $10 and have 40 I/O pins available. You could develop your own version of the Servo Controller and develop a sensor board to monitor most of the sensors and provide reports to a "main" processor like the Propeller. The SX-48 really isn't suitable for something as complex as your spider controller, but it could sure off-load a lot of
the simple I/O processing with the number of pins available.
08-15-2006, 10:02 AM
Well, I plan to build it in stages. slowly adding functionality, and as cash allows me. This project will take a very long time before completely done. I do plan on using the propellar chip as the "brain" for the spider. My first goal is to build a prototype leg, and get it operational by christmas. If I could do that, then that would be a big relief. Also, the hardest part probably will be the walking algorithm, it will take a long time to get it down pack.
08-15-2006, 11:45 AM
You might start with the Wulfden Propeller Board because it's set up to drive servos. There's a nice "object" in the "object library" for driving up to 32 servos. Again, pin limitations restrict that, but you could certainly gain experience with the processor and its capabilities and begin to work out the layers of algorithms needed. There'll be some "canned" patterns of movement, perhaps with some variations, that'll simplify the overall algorithm. Good luck!
08-15-2006, 09:42 PM
Adding to Mike Green's post...
THe wulfden propeller board can easially be modifyed so you can daisy chain additional boards to increase your servo capability.
One thing you might need to consider is the power requirements of that many servos. 6 DOF is a lot of motion, and can alternatly be "simulated" with 4 servos, thus reducing your servo, hardware and power requirements. Just a point to think about is you don't need to 'mimick' all the motions of an appendage for it to work correctly.
For an example, look at the HEX and QUAD crawlers. Each leg has 3 servos, and it's mobility is very good. By adding a rotator servo at the sholder (first servo), you can mimic 99% of the movment of a true leg, with only 4 servos.
Just some humble views on this other wise awesome project... Good luck!!
Propeller + Hardware - extra bits for the bit bucket =· 1 Coffeeless KaosKidd
08-16-2006, 03:46 AM
You should definitely get several dedicated servo controllers for this. Not only do they only take one I/O line each, but they provide the ramping capablility that is otherwise impossible to do (with your # of servos) with just a microcontroller. It's as simple as telling each servo to go from position A to B at speed C, and the servo controller will automatically move each servo to each position at the speed designated for each servo. I'm basing this off the Parallax servo controller, I cant say for other controllers.
If you need high torque servos and are on a budget, I would suggest looking at Blue Bird servos (www.slickzero.com), or even contacting their distribution center if you plan on spending so much on servos. Blue Bird servos are just as powerful as hitec's high torque servos but perhaps not as "finished" or "refined." You may want to try some out first hand before you decide to get 50 of these.
08-16-2006, 07:47 AM
Well the idea behind this project it to fully mimic the spider as much as possible. Yes, I can use less servos, but i don't want to
I made a road map for the project. it's a very long one spand an estimated 7 years, but probably going to take 10.
I also think it would be cool to have it walk around the yard during Halloween, that will scare the crap out of the kids.
Goals for my Spider Robot
1 – 6 months from now:
1. Prototype leg assembled and working
6 months – 1 year from now:
1. 1 fully functioning leg
2. Basic layout of design
1 – 2 years from now:
1. 4 fully functional legs
2. Walking algorithm in place
2 – 3 years from now:
1. 10 fully functional legs
2. Complete chassis of robot and cosmetics
2. Complete walking algorithm
3. Speed control
3 – 4 years from now:
1. Radio Control
2. Ping Sensor integration
3. Ability to climb stairs
4 – 5 years from now:
3. Data logging
4. Motion Sensing
5. Environment Sensing
- Collision detection
5 – 6 years from now:
1. Live Data feed to PC
2. Lay “babies”
- Direct movement
- Build a “map” of its world to know where it can and can’t go
5. Sense when it’s time to “eat”, recharge
6 – 7 years from now:
1. Ability to pick up objects with 2 arms
I may have forgot some things, or some things maybe be added.
Estimated cost once fully completed: ~$20K
08-16-2006, 11:35 PM
In 6-7 years you will probably be able to buy one at Wal-Mart for $99.95 !!
08-17-2006, 12:25 AM
No because I plan to make the spide have a span of 4'
08-17-2006, 12:35 AM
I suspect you're going to want a general purpose computer for the heart of this system. There are many "embedded" PCs that use processors from a Pentium to an Intel PXAxxx series. Many of them will run Linux from a Compact Flash card which you can get now up to 4GB or so. You can use Parallax's Serial Servo Controllers to run the servos and one or two Propellers to handle the sensors. Most of the embedded PCs have at least two serial ports and usually a few general purpose I/O bits that could be used for an SPI or I2C bus to talk to the Propeller(s). With Linux, you'd have access to the tools you'd need for some of the complex decision making (like standard LISP). Although some of these embedded PCs will run Windows, I wouldn't recommend it. Windows is not designed for this type of use and Linux (Unix) is. Windows is lost for example without a graphics display for its GUI while Linux is very happy with a simple command line interface over a serial line.
08-18-2006, 11:02 AM
Well, i finally got all the dimensions for the legs of the spider, and body. Now I just need to start sketching what the legs will look like, and make them in AutoDesk Inventor before my 180 day Student License Expires, stupid expiration date.
Embedded PC's will be much later down the road. I have just begun my Computer Science Degree, so I'm trying to get that done. Plsu I want to learn as many languages as I can, I will always be programming(keeps the mind sharp, and alert) .
If anyone has SolidWorks 2006, can you please go here (http://www.lynxmotion.com/ViewPage.aspx?ContentCode=sesmodel) and download and convert the files to STEP files so i can use them in Inventor. Your help would be much appreciated.
I was able to get the files converted. http://forums.parallax.com/images/smilies/hop.gif
Post Edited (The Captain) : 8/30/2006 8:40:39 PM GMT
08-31-2006, 03:37 AM
Ok, well, I've been talking to people on other forums, and it seems that the best thing to do is to use Linear Actuators for moving the joints. Each leg is looking to be around 5-8 lbs (no actuators), and the servos dont provide enough torque. I'm currently working on finalizing the prototype leg in AutoDesk Inventor right now, so i will keep you guys posted on development. Once i get the prototype done, i will post some pics and the files for viewing.
The spand of the robot is looking to be about 10ft across. http://forums.parallax.com/images/smilies/freaked.gif http://forums.parallax.com/images/smilies/freaked.gif http://forums.parallax.com/images/smilies/freaked.gif
09-01-2006, 01:30 PM
Umm... are you sure that RC servos are going to be up to the task? At that scale, I might think about either hydraulic or pnumatic control (like maby those artificial muscles), and a small engine (like off of a leaf blower, or perhaps even a lawnmower engine) for power...
... and DON'T forget the remote emergency stop! I don't wanna' turn on the news and see "Giant robot goes on rampage thru city, finally stopped by millitary A-10 attack plane..." http://forums.parallax.com/images/smilies/freaked.gif
09-01-2006, 01:53 PM
Well, I was looking in to the 5995's, but I would melt them under the amount of torque required to move the leg. Then I thought about a DC motor - plaetary gearbox - dual output shaft gearbox, but nothing created torque above 40lbs, and I need at least 50lbs of torque. So now I'm looking into linear actuators, each producing 165lbs of force. This is more than enough to move the leg. The legs need to be modified so that the actuators would fir, cause thier 7.3" retracted with 3" stroke. I'm thinking about using the 2" stroke actuators, but I'm still designing. Once I get a good enough drawing, I'll post them. So far, it's looking that the legs will be about 4' long if completely liaded out, but actaully won't cause of the actuators. There is now way I will have a "working" prototype my Christmas because each actuator costs $100. Also, I will have to rethink the power disturbution because each actuator requires 12v DC and draws 3amp each.
Maybe a R/C glow plug engine w/ alternator anyone???
09-01-2006, 08:33 PM
36 Watts per actuator X 58 Actuators = 2,088 Watts peak power if all actuators were maxed out. That works out to not quite 3 HP. Say you only need 50% because they won't all be at peak. That's still 1.5 HP, and a little bigger than a glow plug engine.
Some of the larger four cycle "hobby engines" may do the trick. Salvaging something like a leaf blower or chain saw engine may be more appropriate.
As far as the "alternator", use a DC motor as a generator, and you should be good to go.
For starting out, I'd think about sticking with "wall power" and just have the cord dragged around. Use a white cord, and call it a thread of "silk". You could use some weight to simulate the presense of the generator. Then when you have things worked out on walking, etc., worry about the generator at that time.
8 + 8 = 10
09-01-2006, 10:04 PM
Do NOT use a RC type glow-plug engine, because they're NOT built for long-term use, and also needs a lot of cooling. (not a problem in a RC-plane, with wind rushing by, or in a boat, with all the water around it.)
Use a leaf-blower or chainsaw engine. not only are they cheaper, but they're a biit more rugged, too.
(And you probably won't have any problem with the extra weight.. )
There's a bit of a problem with the power generation, though...
Even at 1000W, it equates to almost 80A/12V, which is 'a bit of power'. you need a rather hefty alternator to get that kind of juice, and if you want a bit of leeway, say up to 1500W, we're suddenly talking about 120A.
I suggest you invest in the largest car-battery you can get hold of and run it with a heavy umbilical at first.
(If you're working in a garage, you could just clip a couple of leads onto the battery in the nearest car... )
What kind of controllers will you be using?
(I assume the actuators are steppers)
In a project this size it may be a good idea to consider the placement and cooling of them already during the early phases of construction.
You may also want to place (some of) the motors/actuators in the body and use steel wire 'tendons' to transfer the power to the joints.
Don't visit my new website...
09-01-2006, 11:37 PM
That's what i planned. The project will an embilical cord for most of it's life. Right now i'm still in the idea's, and design stage with 3D modeling.
However, this is for the big Spider. I can actually use servo's for the minature spiders i plan on making. The will have a span of about 3' if fully laid out. These will be light because the frame doesn't need to be as thick to support the weight, and forces acted uponit.
09-02-2006, 10:53 AM
>> chain saw engine
Actually, most chainsaw engines are not designed for continuious use, either. I would look for a lawnmower or a big leaf blower (such as the backpack kind). Used lawnmowers can usually be had for a song, and some have electric start...
>> You may also want to place (some of) the motors/actuators in the body and use steel wire 'tendons' to transfer the power to the joints.
Very good idea. Also, longer actuation lengths are available. Check out http://www.surpluscenter.com/ if you haven't already; they often have this kind of stuff for cheap - both electrical linear actuators and hydraulic cylenders & pumps.
09-04-2006, 05:19 PM
Finding a motor with an intact electric starter...
Add a couple of decent batteries and stick in some power-management software, and it'll start the gas engine only when needed.
(The neighbours will like that... As soon as they get over the idea of a huge robot walking about... )
Don't visit my new website...
09-05-2006, 10:32 AM
Heres the pic of the actuated arm Im designing. In actuators go as follows farthest to closets.
200lbs of force - 4in stroke
165lbs of force - 3in stroke
165lbs of force - 3in stroke
200lbs of force - 4in stroke
165lbs of force - 2in stroke
It's reach is 48" long, and 20" high
Estimated mass of arm in picture ~12kg
09-05-2006, 03:52 PM
Almost 100Kg just in the arms...
Don't visit my new website...
09-05-2006, 09:48 PM
I never said it was gonna be a small project. Plus, that is estimated for the 3D model program I use. It probably will be higher. Once I can get the actual wieght of the actuators and plug them in, I will get a more accurate mass reading.