I am working on a Walking Robot Simulator in C++ that I hope to eventually use to drive actual robots, either running onboard or relaying instructions wirelessly. Throughout development I have been using a Hexapod model as this seems to be the most popular robot configuration, but recently I have opened the simulator up to the possibility of having different robot configurations, such as quadrupeds and bipeds.
Thats not to say that the simulator will only support these different robot types, in fact I’m looking for unusual robot configurations and designs to test out, such as robots with many degrees of freedom per leg, as well as odd numbers of legs. I have a few ideas of my own that I would like to try but I would like to hear your thoughts too.
Has there been any particular design you’ve had that would be cool to see animated? Things from sci-fi? Or perhaps ideas that have non-conventional leg placement?
A simulator that can animate any SES design configuration I have always thought would be useful. Such a design though would need to be very easy to use so that you can have a library of SES components and by clicking on any of the hole patterns, have the parts snap together so that the faces are matched up and at the angle specified depending on the hole patterns. This would allow users to experiment putting parts together without the need of a cad program with complex transform tools.
After building an SES bot, you could then animate the design to see how it moves. This is a very complex program and might not be possible to do such a thing, but It would be the ultimate virtual workshop.
I’ve had a bit of experience with this kind of system before. Back when I was doing my AI & Robotics degree I created a Modular Robotics Simulator for my dissertation. If you’re not familiar with modular robots then here’s a few vids.
The simulator allowed for the modules to be connected together by connection surfaces similar to SES mounting holes, with the physics engine animating all of the motors I specified. Unfortunately, after adding the Physics, Rendering, and AI (the robots learned to move) I didn’t have time to create a user interface for connecting modules together, I ended up just having the robots specified in XML files and imported.
I have continued with this project after finishing University and am now attempting to address the systems biggest problem, anything that resembles an actual robot (such as a Hexapod) proves too complicated for the AI to control, as it just can’t understand how to move all those joints in sequence. It has taken a while, but I have now got a kinematic system to place between the AI and joints to hopefully solve this problem.
So SN96, I agree that an Ultimate Virtual Workshop would be a very complex program, based on the efforts I’ve made and seen others do, it certainly isn’t impossible.
Anyway, enough about me and my robotics experience, back to unusual configuration ideas
Great drawing! I’m afraid most of us are not there yet!
How about 4DOF Quadruped simulations? Upright stance, not Splayed (or do both?). Several of us are working on 'quad robots and could use the help. the ID for the 3DOF splayed hexapod legs we understand, but I’m having a harder time getting my head around the upright stance.
I can’t take credit for the drawing, although I do agree whoever did it did a great job .
My practical knowledge of robotics is rather limited, its mostly theory and simulation, so I’m persuming that non-splayed is like the image from innerbreed below?
I just dawned on something. Out of all of the Lynxmotion projects I’ve seen on this forum there has been not one tripod project. I guess this would be because it would be rather unstable when taking a step, but maybe with a splayed leg design some sort of tripod gait could be made.
Once I convert the model into a format my simulator can use I can start playing around with the IK required and see what changes need to be made to my system (hopefully very few). It will probably be a few days before I start getting results, so keep the unusual configuration ideas coming in.
Well I’ve now managed to get the Quadruped into my simulator, and it seems to be behaving well, with the exception that it doesn’t understand feet (hence the way it looks). It is just treating the legs as sticks at the moment, so a bit like the ones on your Little Cat. The combination of Yaw and Roll seems to produce a lot overlapping solutions, but I have an idea of how that could be addressed.
Interesting! But I’m not sure how it correlates to Little Cat’s legs! Yours have a more sprawled (they face sideways) construction, it would appear.
The roll axis can possible be treated as an adjunct, and FK calcs done for it. Or so I’m told.
I’m anxious to see more!
Alan KM6VV
P.S. I believe the pix I posted has one leg slightly different, I was experimenting with how the roll joint was attached (C bracket up or down).
I think it would be great to build a “Christmas Bush” as was described in the Science Fiction book Rocheworld by Robert Forward. It was based on the research done by Hans Moravec of Carnegie Mellon university. I have a hard time finding pictures other than what was in the book:
Wow , thats absolutely brilliant!!! I can’t even imagine what kind of tasks you could achieve with robotic limbs like that. This does make me wonder, has anybody actually tried this, at least on the small scale? I mean has anybody put some AL5A’s on the end of a bigger robot arm and tried to manipulate objects with them? Would definately be cool to simulate :mrgreen:.
That model was more like two scouts connected together. This is more like your Little Cat.
The model does work in my simulator, but the yaw and roll caused some odd leg behaviour (switching quickly between two different IK solutions). In that picture I’ve disabled the yaw motors until I can think of a way of supporting it.
It is interesting that you should say that, as with the yaw servos fixed my simulator has no problem making the quadruped walk, making me think that only 3DOF is needed. Thats not to say that I don’t have ideas of how to support 4DOF quadrupeds, I just need to figure out how to program them.
I’d be interested to here what other people have to say about this, and unusual configurations in general. The reason i’m interested is because I would really like to make my simulator support all of these different things, so that people can design cool unconventional robots in simulation before splashing out thousands on the necessary servos and SES components.
Exactly! With my roll servos fixed, I basically have a 3DOF leg, which is enough to get walking!
I suspect, although I haven’t thought it through, that fixing the yaw servo works similarly, although the leg then operates in a different plane, not the sagittal plane that I intend my legs to work in.
I basically added a roll servo to 3DOF leg. I used an LP bearing to drive the yaw axis, something I hadn’t seen before. My chassis (a pair) can of course take a more conventional servo mounted between two chassis plates, and swing in the horizontal plane.
Converting from invertebrate to vertebrate “construction” of robots has its challenges!
I’d certainly like to hear more discussion on this topic!
On a related point, has anybody ever considered having an odd number of legs (beyond the tripod shown earlier)? I’m curious as I don’t think I’ve ever seen a 5, 7 or 9 legged robot (particularly made from SES), and can’t think of any real reason as to why not, other than nature likes even numbered legs.
.
I stand corrected. I didn’t think think tripods would be possible.
