I’m sure a lot of you thought of ideas to improve bipeds.
I had a few, let me know what you guys think.
Based on observations of both videos and a friends biped, i noticed it has to lean to step.
Has anyone consider using a counter weight on the top to balance the weight on one leg, without requiring the biped to lean?
It would allow for thinner, longer feet. Even more interesting it would no longer require the ankle/foot servos to lean.
Another idea was using small electro magnets connected to relays, which would allow the biped to move quickly on a metal surface.
How strong is a 12v 5A electro magnet anyways?
Not to mention you could use soft suction cups with a servo controlled release valve to walk quickly on any smooth surface.
I don’t have a bi-ped so i’m just imagining all this stuff, but you gotta start with the imagination right?
“Oh the stuff i could create with a box of various brackets and servos.”
I agree, bipeds do have some room for extra improvement. But they are also very hard to control so making them walk smooth will take some innovating piece of hard and software. Although i’m not into counter weights and magnets. I’m more into looking how nature solves it. By looking at a human walking you’ll see that we humans “fall” forward and try to put a leg underneat our body before hitting the ground. I think this is the method Boston Dynamics uses for there biped.
archieving this will need some pretty fast servos and a big IK and Dynamics study.
The brat does have limited DOF’s but the principle I think is still the same which is balancing the center of mass when taking steps. Using counter weights might get the job done, but not in an efficient way. The ideal way is to have a dynamic balancing system that works as the bot moves between taking steps. The brat could be made to use a more dynamic walking gait, but you would need to add more joints, and then you might as well call it something other than a brat at that point.
I don’t think this is the case with the BRAT. When I had MsBrat built up, I started to have some small success with a more natural human like walking gait. I eventually want to get back to this sometime and do almost have MsBrat back together now, less servos. I’ve had enough success with this that I do think it is possible with a stock BRAT configuration.
I don’t see how you can effectively manipulate the COG dynamically with a stock Brat. The entire bot rides on one ankle or the other. You can control how much it leans before falling over if you use an accelerometer to keep it within a safe zone, but you would need more DOF’s to shift the weight more effectively. I’m guessing I don’t understand what you have in mind, but to me, the torso needs to be able to stay level and move when there needs to be corrections to the COG. A “PETE” style biped can give you more natural human like walking gaits, but the Brat is more Penguin like.
Maybe there is a clever way of making the Stock Brat walk more dynamically, but I don’t see how beyond the amount that it leans to one side or the other.
All you would need to build a biped with as many dof as a human would be a 3dof hip, a knee joint, and a 3dof ankle. Has anyone ever tried this before?
My weight idea consisted of a small weight on the waist, that would move side to side by a servo.
A little leaning is ok, but from the videos of the brat, the thing leans so much it’s crazy. It just looks silly.
I was thinking maybe one day building a scout style biped from scratch, something that can move quickly, at least quickly compared to most other bipeds i’ve seen.
The weight in most of these cases could simply be the battery pack.
If you place the battery pack on a a couple slider bars that extend beyond the sides of the biped, then use a string/pully system highly geared up to a servo, or you can simply use a stepper motor, you can pull the weight from side to side. I’ll try my hand at MS-Painting the idea. I don’t have the equipment here to actually try something like this out. But maybe you can run with the idea.
Looks like you’ve been thinking about the same that I’ve done lately.
I do have a prototype/(proof of concept) biped that uses the method you are explaining here. I made this for about a month ago. And I know it do look kinda bizzare right now. My intention was to keep this a secret
But what the heck… Here is a little picture of my current project:
At the moment it has 6 DOF using 5645 and the new ARC-32 board. I’m planning to place the battery on top of the ARC board. Since this is my first biped I’ve zero experience and the project might end in the trash can very soon too. But its always fun to try some ideas that are spinning around in our robotic-minded heads…
I remember that I had MsBrat leaning just enough to barely balance on one leg and then I would run the other leg forward of the first leg. MsBrat would indeed almost fall at this point, but she was balanced just enough to allow the other leg to be moved forward without falling.
If I had enough servos to work with, I would fully reassemble MsBrat and strart tinkering with her again. Unfortunately, I don’t even have enough good servos to fully assemble the strut/leg sets I am building for WALTER, so MsBrat will have to stay on the shelf for now.
I would have replied sooner but i was to busy sketching up the idea.
I’ll post the picture anyways in case someone wants to see my 3D sketch of it. I decided against MS-paint.
With enough weight shifted to the other leg, you can keep a general center of balance without leaning to much, and if the timing is right, it can be used to assist the walking and it can help you fall into that next step.
You’ll no doubt need some rubber feet on the biped so it can grip the surface, you can’t have a natural gait if you’re sliding around, like i see done in most biped videos.
In my attachment, imagine a string between the 3 copper pulleys.
Wow, nice! I’m not into the shifting counter balance weights type of thing but your design looks really cool. One problem with weight sliders is that you have momentum build up from the weight sliding from one side to the other, so I would think that it has to be done slow. I guess you could do it fast, but then the legs would need to counter the tendency of the weight throwing the bot completely over. I think a better system is to use the servos to counter react vs. sliding weights.
I have confidence in you zenta. I think you can get this working at least as good as the quad you built.
Using a counter weight would be most effective in a fast gait, it would have to be used to force the robot to fall in a specific direction onto the next legs step.
It would shift the weight to the support leg while the stepping leg is moving, and as the stepping leg is coming down, the weight + a tiny bit of lean, would shift the entire bots weight onto the second leg, while the first one moves up for it’s next step.
It would require fast AND strong servos. If you have enough balance sensors and proper programing, you could probably even get the bot to “run”.
I could probably do the construction on such a bot, but i’d be in way over my head when it comes to programing. The programing would probably take 100+ hours.
You got a point there, I guess we’ll just have to try it out and see how it “go” . But I do believe the slider would work for a Mech style biped since it doesn’t have a flexible upper body or arms that can help stabilizing the body. As Fallentine pointed out, the main idea of using the slider is to reduce amount of sideways movement of the body while walking. One thing that worry me a bit is the mechanical slope in the servos gear train (I’m mean the servos used for the legs), I’m afraid the robot can wobble a bit.
To be honest I’ve not done much since I shared the pics with you. I’m soon going to work with the code and I’m planning to use the Phoenix code Kurt posted for Phoenix on the ARC-32 as a base.
Yeah, it would probably be something like that. I guess you need to do some fine tuning to get it right.
I look forward to see what turns out with the slider design. I really like what you have done. the leg design is almost the same design I plan to use for my next project, when ever I can get around to it.
I forgot that you can position the foot in different locations. I’m not sure what you mean about backward direction. Do you mean the foot? or the direction the leg is going to be moving? I will likely keep the foot in the center since I am not going to be using any IK.
You mentioned using the Phoenix code for your biped project, I think this would be a great addition!!! do you have a name for your new project? what about a name for the modified Phoenix code? Something like “P3b” for example (Phoenix v3 Biped). Just a thought.
I think he means reverse knee.
Reverse knee probably is the better choice.
When you walk forward it pulls the weight of the reverse knee under the biped, where as a forward knee, you’re pushing a lot of weight forward, and your step range would be much more limited.
Here is a little picture of my current project:
