center of mass

[cyberjeff]

Having watched the DARPA videos of quadruped robots, I'm working on a simple model.

At some point, depending on gait, a quadruped has one or two legs off the ground. If this were static, the robot might fall, for sure if it was on two legs. What prevents this, while in motion, is that the other legs move to stop the fall before it happens. The robot must be faster than the acceleration of gravity can take it down. That is doable  these days.

I see different sets of problems and equations. Let's say it is near stopped and on 3 legs. If the center of gravity is within the triangle of the 3 contact legs it will be stable. Tilting slightly one way, by  stretching or pulling a leg can make all the difference as to what is going to happen next.

I would think that a 3 axis accelerometer  and a gyro would tell which way the robot was headed and if it needed correction, Some of this data could be stored so that results of movements could be predicted by records of past actions.

So, it all boils down to software. How much of this already exists and is open source and in which language? Where are the resources?

I'd rather not have to write all this, although I can see path toward doing that.

[mklrobo]

  Hello!
Yes, it is a daunting task, getting the mechanics of making the robot walk. THEN, you have to program
the servos and sensors; which is in a world of itself. Your robot programming to that body will be
specific, and unquie. I am hoping that, sooner or later, everyone will have a few generic robot
bodies to use, and then design off of them.(Open Source Robot)
I would offer an opinion, but I would need more info. I have tried to use 4 legs, 2 legs are turned
inside. The spacing of the steps allow a triangle of support for the robot, while a leg is moving; THEN
the whole body shifts forward placing weight on a triangle of legs, giving another leg a chance to move
without threatening the stability of the robot. the legs fold underneath the robot, to allow wheels to take
over when necessary. I am trying to "cheat" whenever I can, to take full advantage of the mechanics.
After all, the human body does not have "cheats", but the robot does.    Good Luck!! 

[cyberjeff]

My HS physics is poor, but it looks like you can solve for reaction time if you know the velocity forward since g is known. Doubling the velocity, doubles the time.

 A servo that needs .125 sec to take a "step" needs to be going forward at 2fps to just equal gravity. So it seems to me that sets a lower boundary. Although 2 feet per second is rather slow, it is probably faster than most hobby robots are trying for.

It seems to me that when you are stopped, all 4 feet go down and you are stable. Otherwise you need enough speed for enough momentum if you will have more than one foot off the ground.

There is a gait called ambling, this is what gaited horses do when they are traversing rough ground. It is statically stable. If you have a 4 legged table and you lift the front left it leaves the table in a precarious balance. However if you shorten the opposite leg it will tilt back that way and be stable. There are other ways of shifting the center of gravity but I think for a simple robot that is a good option.

I'm awaiting some servos from Banggood: http://www.banggood.com/4X-TowerPro-MG996R-Metal-Gear-Digital-High-Torque-Servo-55g-p-983215.html

I'm mulling over ways to determine foot fall ground contact.

What do you think?