300, Dollars? I should think so. But then you might not be state side!
We have Tech Shop here in the Bay area. You could cut your parts there.
The Band saw is your friend!
Also, I posted on the octopod thread on a chassis I designed. Jim has the plans. It’s really too big for my little CNC’d Sherline mill, even if I use multiple setups. I might get it cut at work (aluminum).
like i said earlier if i was only stateside and had access to real tools… SVG is what inkscape uses for a file system. i really dont know how to convert them.
Have a challenge presented to me… Keep the ping))) sensor. But switch to video based navigation. And keep all the processing on board. Chip will be the ARM Cortex M3. And of course i said sure why not. I know nothing about it should be fun to learn… Now i need to find a video source for my bot… Simple list to do to finish it just got a whole lot longer. I need to stop talking to my friends that actually went to college…
If you are without “real tools” and you are going to only make one of these, then depending on the material being cut you may be able to hand cut out the parts using a coping saw and dremal tool.
i would have to order those in also… blah. instead i just had it cut from abs instead of acrylic and called it good. Mail has been absurdly slow lately so new progress other that code work on the AI. Which may just need a small amount of tweeking and should be good to good.
You may find doing real time image processing with only a ARM Cortex M3 a bit of a challenge to say the least. What demo/eval board are you going to use (if it has been selected)? I would also recommend using a camera that has some internal processing available, something like the oft mentioned CMU CAM 2 would possibly allow you to offload or at least scale the programming task for the arm7. Consider that you have no way to estimate distance with a single camera and it would be much simpler to supplement it with ultrasonic and infrared sensors for path planning and object avoidance than to add a second camera and try to process stero image processing to accomplish the same thing.
edit:
actually I was just digging around the cmu cam info and found the cmu cam 3 is only about $50USD more and has a whole lot more bang for the buck as it is already based off a LPC2106 ARM7 and has a whole open source development SDK with it. so it might be a bit more practical than I initially considered to use this as a starting point and modify the code to include IK and routines to drive the full hex through the ssc-32. you probably still want to add some near field ranging sensors but there are 4 analog inputs on the ssc-32 that could be used to read those sensors, albiet slowly. I would consider placing one on each of the front legs since as you walk they should each get a good scan of +/- 30 degrees or so to each side facing forward.
Sensors: 1 Ping))) ultrasonic sensor. 4 IR Sensors (was only 2 but your right and for some reason i checked my email before ordering).
I have been seriously considering the CMUCAM3 for vision the more i research on it. If i go with a serial camera i may be able to only use 1 or 2 frames per sec max.
The plan is to keep the SSC-32 as the “walking” brain and have it only worry about that. The Camera and the Sensors will trigger interruppts for sudden obstacles etc. The AI will monitor the sensors and decide leg postions and movement. I also have gyro to add into the mix… do i have enough sensor’s… maybe… (after all i wish i could do 360 degree and hearing also).
So much for an easy dumb robot for the first one eh? But none of this would have been possible if not for the ease of using the SSC-32.
Hmm, I just want to make sure you are not making a more or less classic error for new users of the ssc-32 which is the missing realization that the ssc-32 is strictly a servo controller and not a programmable controller. yes you can store a number of canned sequences in it but it still needs something to tell it to execute those sequences. typically people use a basic atom pro in the projects and tutorials available through lynxmotion. certainly other controllers have been used as well so I mention the basic atom pro only for convenience.
what I was suggesting above in reference to the cmucam3 was that it could do the processing typically associated with the basic atom and send the serial port commands to the ssc-32 if you were willing to dig into its sdk and develop the required programming which is probably in C/C++ (as I am guessing at a gnuC toolchain/set).
oh no that is why its the waling brain only. All it needs to worry about is the servos. My AI brain will take the input from the CMUCAM3 and process the right command to the SSC-32. Really leaning towards the cmucam3 as i read more on it
I am working on it. The heat shrink is being a pita to get over the connector. But after i finish the prep stage i will be assembling the rest.
I was dead tired yesterday so i just got the test done along with 8 servos loomed but not heat shrink’d, so far it looks like today is going to be another long day
That power connection for servo power doesn’t look too good. Asking for trouble. The servos are going to require about 8 amps. You need to make sure the wires can carry that much. Also screwing the terminals down on header posts is a poor connection and is a bottleneck for the current.
I got 12 gauge wire and a traxxas male plug to fit the battery on the way. I am a mechanic, which means i have let the magic smoke out of lots of wires. Thanks for the suggestions tho. That wire was just used to test the servos to make sure there were no DOA’s in the bunch.
ok time for jokes. I lost a screw has anyone seen it? So I got tired of my nikon point and shoot and broke out the nicer nikon so i don’t have to deal with the flash. I would have gotten more done but i spent 45 minutes looking for the screw i lost. It also doesn’t help i need to use tweezers to put this together. So tomorrow i am going to switch all the screws out for black socket caps instead. Also i got all the left hips assembled and realized: No bearings. crap.