For awhile now, I have been thinking about jumping in and trying to build a quad (or two or three)… So Recently I purchased a Lynxmotion Quad body (lynxmotion.com/p-435-quadrap … -mini.aspx) and I used 4 legs off my reconstructed CH3-R hex robot, at has on it an SSC-32 and a Botboarduino and currently a PS2.
Also I have received an Orion Robotics Quad kit that uses their Venom legs. Edit: This is now known as the Mantis (orionrobotics.com/Mantis-Qua … p_329.html) This one is setup with their DaVinci Arduino board with their Orion shield and uses one of their PS2 controllers.
For the Lynxmotion quad, I am using my Phoenix Arduino code base, that I have up on github.com\kurte, I have a quad fork of my Phoenix in parts project. I have most of the code in place, and for the first pass was starting with the body shift tables. But there are some other ideas bouncing around that I would like to integrate in. More details soon.
Currently for the Venom Quad, I am starting with their Venom Hexapod code base and I have been modifying their code base. Currently I have this one walking better, as you can see in my quick and very dirty video:
More details to come. I am hoping to document some of the different ideas that a few of us have been bouncing around. Got to go.
Kurt
Edit: As per suggestion from Xan - Changed Thread name Except fix spelling to imply multiple…
And extremely interesting video. You have that walking much smoother than I had been led to expect quad robots were capable of. I wish there was an easy way to run the video in slow motion - slow enough to really detect the sequences of leg parts in motion. Are your modifications to their code made with IK routines?
Also, pardon me for being confused, but where is the messy desk you mentioned? I can still see the desk.
I have two differ net gaits involved. The first is a wave type gait that only has one leg up at a time, which is consistent with several of the threads that are up on these forums. The 2nd is a simply gait that is like the 4 step tripod gait, except we only move 2 legs on the diagonal at a time. My guess is this only works when it is moving reasonably fast. That is what you are seeing in the later parts of this video.
For the more normal wave gait, next is to do some balance shifting. Many ideas to try out here. Some of these schemes may be specific to take advantage of the hardware capabilities of some of these quads. Example with the Venom or with the Arbotix, we may be able to take advantage of being able to query legs for how much force/torque that are on them. Example if we start lifting a leg, we may be able to monitor the force on the opposite leg and if it looks like it is starting to lift, we may be able to increase the body shift… My guess is that I will run out of horsepower on these lower end Arduinos…
Yes a Roboted pointed out… the desk is still visible. ? For that, im quite disappointed. Thought you were busier than that. Lol
I have already seen basic micro have uploaded videos of their range of hexapods. And so to see a modified custom quad is really nice. I think they have a good thing going on with regards to positional servo feedback with their kits.
I like the idea of a dynamic body shift using this feedback feature. The quad shifting tables I used worked quite well but your idea will allow for much better control.
I also notice those servos are very fast. How do they compare to the hitec 645mg’s regarding torque?
As for my desk, well I have company coming later today so I had to clean it up some Also I did not show the rest of my office.
I would also like to start off saying that the majority of work was done by or made possible by the works of others! This of course includes Kåre (Zenta), Xan (Jeroen), Nathan (AcidTech), Innerbreed, and several others from other forums like Trossen. I just have the fun of trying to help put pieces together and try to make it work.
I have the moving 2 legs at once Gait moving pretty fast as to not give it much time to fall down. I will try that one as well on the Lynxmotion Quad, but being that it is quite a bit longer in one direction, I am not sure it will work as well. But we will see.
Yes it is nice to see that they have uploaded some videos showing their new Hexapod robots.
As for the specifications of their digital servos: 220 oz/in at 7.4v (Lipo), 0.12 sec/60°.
As compared to 645mg: 133 oz/in at 6v and .20 sec/60°, Or 5685MH: 179 at 7.4v, .17 sec/60°
I very much like the idea of the servo feedback that these provide as do the Robotis Servos. I will soon be creating a thread up on their forums to talk more about the things that are specific to their platforms.
This is really great Kurt. You’ve done a lot of great code work to the robotic community. Fun to see some fast quad gaits too. Like I mentioned to you earlier, using the balance mode in a slightly different way would be fun to try. There are some parameters you can change for tuning the gait. Let me know if you need some help/hint.
This was inspiring. I might need to make a new quad one time. Orion robotics really did some great stuff to their new servos.
You (and Xan) deserve most of the credit for this!
On the Venom side, I have started to experiment with with playing with the balance mode… Still need to debug through it. Would always appreciate your help with this.
Likewise Xan mentioned about calculating the balance using a subset of the legs… Lots of different things to try out.
I don’t think they have uploaded the code yet to the server. Have not checked the website for a couple of days. I received zip files with the code for both the Venom and the Fire Ant as well as all of the supporting libraries by email.
I would like to jump in here. Just like Kurt I’m also building a quad. I will post some pictures this weekend since it’s to dark to make a picture now, but the setup is close to this image. but I will post more about that later.
Kurt and I where already exchanging ideas about the COG shifting logic. Instead of making a static location for each step with an byte table I would like to solve it with math. The rough idea is to predict the leg(s) which will be lifted by the gaitengine. Then use the balance mode to calculate the correction needed to place the body within the supporting legs. I guess that the correction will be to large so we need to reduce this factor somehow.
These quad developments sound interesting. Something id like to see. I did find a few problems while using the byte table shifting. If one wanted to quickly change direction then the body would "unnaturally’ shift to the required postion, in a fast motion. This shift would often disturb the overall balance, making things look rather jerky.
I like the idea of solving this using maths, and shift integration into the ik directly.
My 4dof quad worked rather well but again it did have issues due to the tables as explained.
Looking forward to seeing more.
Except fix spelling to imply multiple…
. The legs seem long though, how many inches are those ?