I have started assembling a new four legged robot made out of Lynxmotion erector set pieces. I had a number of 1/4 scale servos and brackets from a previous robot and I wanted to try something new. I was really inspired by innerbreed’s robots and I wanted to try to build a quad with the legs under the robot instead of sticking out as an L shape. At first I tried to use the T-Hex as a chassis, but it was a little short for the length of the legs. Now I am just using an aluminum channel so the body is very minimal.
It will probably take a couple weekends for me to get this walking, but I think it could be really cool. My goal is for it to go straight on rough terrain and have it be able to have 5 cm of ground clearance.
It has been really cool how the Lynxmotion brackets have made it possible for me to experiment and reuse parts. I just put up a few pictures at:
backyardrobots.com/kudu/kudu.shtml
Hi, wow thats a big quad. i would love to have the chance to build one using the larger servo bracket range.
i have a few concerns one the joints sharing the same linear plane. i understand you are only looking for forward loco. well this should be fine, but why not take advantage of full mobility. it seems imperative that you get this 5cm ground clearance. is there a main reason for this. the lower the better.
anyway, ill be following this so keep us updated.
Thanks for the mention. highly appreciated. happy to inspire. 
Well this would show why 5 cm of ground clearance is important, and why the goal would be to go straight 
youtube.com/watch?v=1ihNdEiki78
I made the middle joint have a long bracket, because without that I couldn’t get the 5cm gap between the leg straight and the leg folded. Even with the long bracket it is not clear that it really will have that ground clearance in practice. Eyeballing it I thought it would be very inclined to tip over, but there is one strange advantage to using those servos-- they are heavy. If I can get away with a 10 ounce LlPo battery (unknown) the center of gravity is going to be very low. The feet will weigh more then the torso, which makes it really hard to tip over.
I have seen the gaits of your robots and I can understand why you think it would be better to have other degrees of articulation. My thought was to try and see how simple of a thing I could get to work and then try different things. What I really like about the brackets it that you can experiment.
looking at your video and understanding your reason for the clearance and for the forward locomotion, i still think some form of turning or terrain adaptation will be needed for it to accomplish that task.
as shown in the video it has a slight problem going straight due to the terrain its walking on.
on the other hand the Ziggy robot does it quite well. i know this has 6 legs but also the feet are less obstructed by the uneven terrain due to their thinness and precision of the feet.
again thinking about your design. you might be able to adjust the course direction by over stepping with the flighted legs on one side. say if you over stepped with the left side legs then the robot will “steer” right.
but then again im guessing this is an autonomous robot.?
Yeah, the walking robot competition is autonomous. There is a huge advantage in using a hexapod because it is less likely to pivot on uneven terrain. Ziggy does also go off course sometimes, I think the videos you see are when it did the course going straight.
I don’t understand what you mean by “flighted legs”. I don’t understand that term.
If I put a touch sensor on the foot I could probably guess when a foot was stepping on an obstacle. A tilt sensor also might be useful in a similar way. You could detect when the robot was tilted or going up an incline. Last year I tried to keep the robot going straight with a gyroscope, but the rocking of the robot was too much to filter out.
I think it is kind of crazy that they give us no navigational reference at all. I would love it if they would put a line on the side of the track that I could detect with a color sensor, or a plexiglass wall I could detect with a Ping sensor.
Last night after work I taped on a micro-controller because I just couldn’t wait until the weekend. I have just begun to power it up but it does seems like it is able to stand very steady on two feet. I may get it walking sooner then I thought because it seems very promising.
by flighted leg i mean the leg being lifted.
Oh- I get what you mean. I don’t think it will be hard to give this robot tank steering: where one side goes forward and the other side stays in place or even goes backward.
The way I have been controlling the motions of quads lately is with a sort of central pattern generator. One saw tooth wave is fed into functions for each joint. The functions bend the joint with a sine function (well a look up to a sine table) with different amplitudes and offsets. With the standard pattern fed into the function the feet move in a circular motion. If you speed the saw tooth function up the legs circle more quickly. Run the sawtooth function backward and it does the whole thing in reverse and goes backward. Run one side forward and one side backward and you can spin in place. Also, with bit masks on certain joint flags you can dampen or increase movement along certain axis which can let one side be still in a horizontal or vertical axis. I have never tried it with this leg lay out, but in theory it should work the same. It’s not as robust as IK, but it is not computationally intensive either.
I still think with this configuration I am going to have a really poor turning radius for a walking robot unless I add some more joints. I love the articulation you have on your robots but I am always trying to get a good result from fewer motors. One of the things I was thinking about was just to add a single servo that bends the robot at the waist.
Getting Kudu to walk has been much tougher then I expected. It’s a big jump from the style of quads I have been making. But I am getting it to stand and alternate feet in a 2:2 gait that I think is close to the one I am looking for. The knee timing seems off in a way that keeps it from going forward properly.
I put video of this at:backyardrobots.com/kudu/kudu.shtml#march_13
I am still really excited about this robot, but I have a long way to go to get it to walk well.
I set this project aside for a few months to prepare for Robogames and work on a different quad robot. This weekend I came back to it and I put in a jumper cable so the robot didn’t need to be plugged into a USB cable to power the controller. I also taped the servo controller board (propeller) down with duct tape and tied the servo cables together with twine. With just these changes I got the following results:
youtu.be/IDn_HkXa6p4
It seems to be closer to walking then I remember. It also actually seems to do better going faster, because it sways less. It also has reasonable ground clearance for this point in development. I could get it to navigate brick, carpet, and the transition from wood floor to carpet. (Every quad I have worked on has had the problem of dragging a foot when it does a 2:2 gait.) But it is still very unsteady, which is why I set it aside before Robogames. Now I am back on it and I am thinking what to do. This forum has the experts who can guide me!
I am considering:
- Making a wider frame so that I can set the legs further apart. Currently the feet are only ~3" apart (center to center) in width and 11" in length. I am thinking 8" width, 12" length. It will be heavier though at the top of the robot.
- Making the feet wider. I am pretty sure wider feat will reduce tipping. Seems like the easiest win.
- Put a counterbalance weight servo like a tail on it. I tried this before on an earlier quad and the results weren’t great. But that robot was a different geometry and swayed less in general.
The one thing I am sure I do not want to do is go to more servos per leg. I want see how well I can get the robot to work with four servos. I do think I will eventually replace the short brackets with longer ones for more ground clearance.
Advice?
Here is the robot with a ripple gait. This helps it not tip over with its narrow rail chassis. It’s slower than a 2:2 gait, but with the 2:2 gait it often tips over at obstacles. I am planning on making the chassis wider, which should make it less tippy and easier to turn. The way it is now turning is really hard because the feet are so close together.
I fixed the youtube link. 
It’s walking much better…
Thank you. It’s funny because I was looking at Quads on You Tube and I saw an early video of Innerbreed’s Dark Wolf and it had almost exactly the same gait as the ripple gait I was showing.
I got a new piece machined for a larger chassis, particularly to make it wider. I think wider will help it be able to turn better and make it less prone to fall on its side. Of course having more joints in the legs the way Innerbreed does would also solve the problem. Maybe that is what I will try next if turning continues to be a big problem.
yes its walking much better. i too had problems walking on carpet. i fould that i was able to remove the feet and just walk on the back side of the bracket. shown here: (although no walking is performed)
youtube.com/watch?v=TZPWQS5f_xk (no need to embed this one. )
i accentually ended up removing some of the servos in order to get more movement. i know this sounds strange up the finial idea ended up being 4 dof per leg. and not 5. to side step you will need to add another servo. with your smooth gait you might even benefit from having a rotate servo on the Y axis of the leg, allowing it to turn rather than step to the side.
I love that you got Dark Wolf to sit and stand up again. Eventually I would like to get Kudu to recover from falling on its side but first I need to get it to turn better
It’s interesting that you went to 4 DOF legs, 5 DOF was obviously working pretty well but I saw a movie with 4 DOF configuration that looked just as good. I noticed that the feet got smaller. Was that that to accommodate the legs changing their angle?
I would really like to stay at 3DOF if I can make it work but I don’t see how I could ever get it to turn half as well as Dark Wolf does. I like your suggestion of adding a y-axis servo to the legs. You could actually just do this to the front or back legs and get some turning. Another idea might be to add a single servo at the waist in the y-axis. I always want to try and keep the servo count low.
Here is the new frame piece I built to widen the stance and provide mountings for the electronics.
http://www.backyardrobots.com/kudu/kudu_frame1sm.jpg
I am going to need to get longer servo cable to get this working again, so it may take a while to get back to walking. But I think it looks cool! It is very exciting for me to learn how to make parts with CAD.
Sorry, I meant Iron Wolf. It must have been because the videos are dark
It’s walking much better… 