RE: [beam] 12 Motor Beam Hexapod (Long with Pictures)
- That's some good work. You could try using a 6 Nv loop as a main walking cpg to control your hexapod instead of the the 4 Nv Loop, or perhaps two linked tricores. I've managed to get the standard tripod gait, the ripple/ wave gait out of them .You will need at least a 4Nv chain for each leg, so you'll have 1 hc14 for the cpg, and 6 for the legs. Seems like a lot, but there will be 2 spare Nvs for each leg which will allow you to either make another cpg or neural neurons for sensor processing, or just to use as motor neurons for upgrading to an 18 motor tripod . You can have a finer, more specific control of each leg this way, more so than your current layout. It's a simple matter of injecting the correct number of processes into the loop to form the gait that you want. You can change the cpg, and even the individual motor neurons for great effect. You can use the tripod gait for walking forwards and back, while reversing a side will allow you to turn on a dime, and use an attenuated version of the ripple gait to do a gentle turn. The part of your design I like the most is the use of unmodified servos. I haven't looked at your circuit for driving them yet, but will it allow for the servos to hold their position (sorry, Queen's night, head very fuzzy)? This is key in getting your hexapod to walk well. Else you'll have to rely on springs. As for the leg layout, I would advise you to offset the pitch servo from the yaw servo instead of joining them at the case. This will give you a much more stable gait.Keep working at it, I like the ambition.Best Regards,Amit
Date: Wed, 25 Apr 2012 09:50:15 +0000
Subject: [beam] 12 Motor Beam Hexapod (Long with Pictures)Alright, so after my last bout of fun with walkers, the crawler, I decided that two legs just wasn't enough. So naturally I went straight to six legs! Alright, so that's no so much of a logical jump, but I've always had a fascination with hexapods. Plus, you can get a really easy 3 by 3 gait. I had this idea growing in my head of how to achieve a 3 by 3 gait simply using nothing but analog ideas. Now I've seen 2 motor hexapods, 3 motor hexapods, 9 motor hexapods and the standard 18 motor hexapods and they're all genius in their own right! But one thing I haven't seen a whole lot of are 12 motor hexapods. This seemed like the most logical choice for the goals that I had in mind though, so I busted out my breadboard, ordered a bunch of servos and got to work!
This design went through a huge number of iterations and each one helped me jump on to the next one! The design started as just trying to figure out how to get a simple nervous network running.
From there I started bolting things on to the design until I had something that I could scale out.
Until I finally got to this design!
Alright, so that's a big mess of wires, I should probably explain whats going on here! Let's start with the nervous net. It's a simple 4 neuron network that propagates a signal across. I used Wilf Rigters PNC uCore design only with the diodes flipped so I can source current instead of sink it. I then use a set of diodes to feed off of three of the neurons so I have essentially two signals that follow the pattern below as the original signal propagates along the core.
Sig.1 Sig. 2
I use this pattern as a base for my lift, move, drop, move sequence to get my legs moving just how I want them to. I then feed the output signal through a buffer to make sure I get good strong clean signals. The buffer is flipped on or off depending on a signal from an OpAmp hooked up to an LDR. The idea is that whenever the bot finds the darkest spot in the room, it stops moving and just chills.
From there, the signals split, one half goes straight to the servo controllers (more on this in a second) and the other half feeds in an X-Nor circuit. This X-Nor circuit uses two inputs to produce a single output for the servo controllers. The second input comes from a Sharp distance sensor running through a darlington transistor setup into an OpAmp. What this does is when the sensor kicks off, it revers the signal from the nervous network. This only applies to the moving servos, so my lift actuation still runs in exactly the same pattern. This allows half of the bot to reverse, essentially dodging any obstacles in its way.
Finally, all of that madness feeds into 4 pulse width modulators built from 555 chips. Each one of these splits into 3 and feeds through a buffer for a clean signal to the servos.
Hopefully, this will give me a working circuit for a 12 motor hexapod. So far, on the breadboard it works surprisingly well (although there's a bit of noise coming from the fact that the servo power source isn't completely isolated). There's still a lot of work to do, but today I mocked up how I was thinking to organize the chips.
And, while I was bored at work, I sketched up a design of how I kind of want it to look. I'm going for a very similar look to my crawler, only with a lot more motors and circuitry.
That's as far as I've gotten so far! I'm working really slowly on this because I don't want to push myself into making a design decision that I'm not sure. I'm sure my design and circuit needs a lot of work and it'll get constantly tweaked as I start to assemble everything! As more happens I'll post it up! Plan at the moment is to get the main circuitry built and working, then build my frame around it and finally attach the servos.