Around may I finally will have enough time to start a new robot-project. I have chosen the Scout although Bipeds appear slightly less popular than for example hexapods. It seems to me, that one of the reasons could be, that in general a walking hexapod looks more elegant than a walking biped. Due to balancing reasons, bipeds appear to walk a bit clumsily. I want to change that with the help of the biped-community, this is why I started this topic.
The main goal of this topic is to provide and discuss ideas of how to make bipeds in general and the Scout in particular walk more elegant and beautiful. Every idea is welcome. Beginning from may, I will document my Scout-project and of course provide images and code.
As for my hardware: I own the classic Scout-Kit and will use HS5645-Servos. For powering I will use a lab power supply which is able to provide currents up to 40A. In this way I will save the weight from the batteries. And this is also the first thought for improvement: the high center of mass from the Scout due to the batteries and stuff favours a mechanically more or less instable behavior. Other aspects for improvement could be acceleration sensors, gyros, …
Actually the basic boundary condition is, that the resulting force from the robot’s weight, coming from its center of mass, should always be above the contact from robot to the surface it is walking on, when standing on one leg…
I’m not sure if you find this interesting but I did work on a biped project called Archer a couple of years ago. I can’t say it walked beautiful though, but I find it interesting to use IK combined with a good walking gait algorithm instead of using predefined sequences. Basically I used the same Phoenix code (written by Xan) also for controlling the biped.
I did made a little video about it too. One issue you’ve to deal with is the gear-slope on the servos (I did mention it in the video).
Dynamic gaits look way better. But i noticed with my scout that it walked better when standing taller. I found that it needed to lean 20 degrees to the side to lift a leg and stay balanced. If you lower the c of g, it would have to oean farther and look more clumsy.
@Zenta: Of course I have seen the video already you posted. There aren’t too many Scout-videos, so I couldn’t miss yours. But this time I watched it and paid more attention of what you said and not on the movements of the biped. Thanks for the hint.
@Xzvir: Of course you are right, the advantage in lowering the center of gravity by making the Scout crouch lowly would make it necessary for the robot to lean over in a higher angle. This would result in a Charlie Chaplin like movement, which is quite cool, but not exaclty beautiful. Thanks for the comment.
A pity, that May is still sooo far away. I am eager to work on that task. Anyway, thinking and discussing this problem is the best option to fullfill my robotary needs at the moment.
Oh, the other problem I had when I tried to make it walk low was the legs sort of binding up. But that was largely because of the gear backlash in the 5645’s and if one were able to fix that, it would help tons.
I’m currently teaching myself kinematics using the book Theory of Applied Robotics and an al5d arm.
A long term project of mine (probably about 1 year away) is a humanoid biped with realistic walking.
I have limited access to internet (weekends) but here’s what I’ve found about realistic biped walking. This started out with information from a paper called “Human-like Walking of Humanoid Robot Based on Biped Kinematics and Captured Motion of Human”.
I don’t understand all of this stuff, but adding toes to the feet, and forcing the robot to walk with the knees stretched seem to be very important. Toes decrease torque on ankle and increase step length by up to 80%.
With the knee stretched, walking it is more realistic and minimizes how much the center of mass moves up and down.
Then worry about balancing robot. Popular methods are 3D linear inverted pendulum and Zero Moment Point. I think all of the math needed is in the book I have, but I haven’t gotten that far.
thank you very much for your comments. I also did a bit of literature already, focusing on bipedal movement. If you like, I can send you the PDFs I found. I also have my lecture notes somewhere on my robotics-lecture. If I can find them (I have been moving this week so nearly everything is packed away), maybe we can compare approaches and exchange thoughts.
Toes: I also think that toes are very important for a fluent movement. I already thought about pressure-controlled fluid-cushions under the Scout’s feet but I think there are many other options to be considered first and realized more easily.
Reading your post, I thought about a turnable toe to make Scouty more stable in the sideways-direction. Something like on the feet of chimpanzees. Check out this linked picture: williswissensweb.homepage.t- … ageP8B.JPG
Chimpanzees usually are broader than humans so their toe is directed more sideways. This is exactly one of the problems with the Scout. I love nature, because she has so many good ideas.
At the moment I still have to finish writing my PhD but I think I cannot stand it much longer without doing something robot-related so I will start to put my Scout together tonight. As this post is ment for brainstorming, I will start a new one to document the Scout-project.
I did not comment on all your input in order to prevent my answer being too long.
Thanks again for the input.
Feel free to send me your pdf’s. I’ll compare them with what I have and could send you some that you don’t have and would be useful.
As for toe joints I will probably run with a passive toe joint. Maybe spring loaded.
Most humanoid/biped robots I’ve seen have 2 dof at the ankle joint. One of them is in the x-y plane rotating about z axis. This might work to implement your turnable toe idea.
Zenta mentioned using IK vs predefined sequences. In the paper I mentioned in my 1st post, it appears that the motion of 1 knee is predefined, and the other knee motion is calculated. The predefined motion seems necessary because ik solutions will avoid singularity (knee joint angle approaching zero degrees), which is what is needed for realistic/efficient walking.
project-start is shifted a month or two. Surprisingly I discovered two brandnew aspects for my PhD and I have to add them before I can finish it and start something new (the biped-project).
In my opinion it is a bit unfair to have such an interesting world and only one life to live it - it is even more unfair that one sleeps approximately one third of this precious time. But life is no pony-farm.
Hips.
Its really that simple.
When a human walks, our legs are actually going up and down thanks to the motion in hips/hip flexors. That is the reason that so many bipeds wobble so much when they walk, they are simply moving their legs forwards and back.
I propose in order to make something like this work, the legs would have to be attached to basically teeter tottering hip with some form actuators to move it both up and down and forwards and back. Think like a crank shaft on an internal combustion engine. Something to actually lift the whole leg and move it forward with its motion, not having the legs do it all.
Walking requires hip motion too, and in fact its rather hard to walk with out using your hips.
I might be a complete newbie to robotics/programming/anything-of-this-sort-what-so-ever, but I have been cycling and running for a while, and this is what I have noticed.
