Hi guys,
I am building a bipedal walking robot that kind of look like this
see attached
I am running into two main problems, one mechanical and one electrical
How does a two legged robot make a turn?
I think that to walk you will need two extra motors to shift the weight from foot to foot before making a step
Two ideas so far on turning
a) Have a motor to turn the entire leg as it is taking a step, and the other leg along with the body follows
I have seen this form on other robots before, but I don’t like it much. It adds extra weight (which is already a concern), need extra space and is slow
b) Just have one leg walk faster (take bigger steps) than the other
The turning radius will be shit but I prefer to do it this way. But will this way even work?
Any other ideas?
To balance the robot and do some form of fall prevention do I just need a gyroscope or do I need a gyroscope and accelerometer?
I don’t need to know how fast I am going, but I want to make sure the body is always leveled.
If I do need a gyroscope what sensitivity should I use?
The star wars robot you show is walking much like a bird. This is MUCH harder than it may seem. Consider whay happens when a foot leaves the ground - no static stability.
do I need a gyroscope and accelerometer
Yes, but how do you use the data?
Functioning solutions generally require a lot of expensive servos and much software.
I would suggest starting off with big fat feet to improve stability.
Quick and dirty solutions based on using large overlapping feet have been around for years.
If you want to make something that actually balances itself on two legs that is a much harder problem, which large universities and corporations have spent millions trying to get working, with mediocre success so far. To attempt this as a DIY project, especially for a novice, is extremely ambitious.
This project, considering its forerunners, has probably cost a bit less than $100 million. Here is the U.S. no sacrifice is too much in the pursuit of military superiority!
Hi
thank you all for the quick response
yes this is a very ambitious project, i have no idea if it will work.
but isn't that what it is all about? to build things that you imaged and making it work.
it would be easier to make a big cross footed robot with short legs, but that was not the goal of the project
mechanical wise the foot will be much larger than that of the picture, and aligned with the center of gravity
my main concern is will i need extra motors for turning, or can i just make one leg run faster, this will decide design changes
I will likely spend month just to make it walk a single step but i need the frame work to be able to make a turn should i ever get to that step.
I am confident that with some hard work and your support i can do this project,
and i have to the two weapons that over come all things, money and free time
yaleguan:
my main concern is will i need extra motors for turning, or can i just make one leg run faster
Since you have a working prototype (yourself), you can try out different approaches. I think you'll find that you need a twistable ankle, though, in order to walk round corners.
What if you were to have motors connecting the ankles and the feet, so that way you could pick one foot up, rotate the robot on one leg, turn in the desired direction, and then place the foot down to continue walking? It only requires 2 more servos.
The only other option I can think of is having, like you said, one leg take a larger step than the other leg.
I also need a solution to this problem, as I'm working on my own biped project.
The thing about big feet on bipedal robots is that the center of gravity stays over a point of support for the entire stride. If your robot conforms to this, it shouldn't be too hard to keep upright. It's statically balanced, that is, if you stop the motion at any point in the stride, the robot will remain upright. With small feet, the robot's center of gravity will not be over a point of support much of the time and your robot needs dynamic balance. Not a simple matter. There is a book on this, Legged Robots That Balance, describes 3 robots at MIT which dynamically balanced and were forerunners of robots like Big Dog. But not only is the software complex, the power requirements for the leg actuators are pretty high, because while you could gear the motors on a statically balanced robot to provide more torque at a lower speed, the legs on a dynamically balanced robot have to react quick enough to prevent the robot from falling on it's face (part of the stride is basically a controlled fall). The robots in the book I mentioned were used pneumatic or high-speed hydraulic actuators and were tethered, receiving power from sources not on the robot.
Which hopefully explains why you don't see a lot of bipedal robots with small feet.
pharseid379:
(part of the stride is basically a controlled fall).
I had forgotten about this. I saw a very effective wire-frame (stick person) animation of human walking some years ago. It seems like ALL human walking is the business of catching up with our centre of gravity. The first part of moving forward seems to be the movement of our torso fractionally ahead of our feet and then our autonomous system tells the feet to keep up. Which is consistent with your warning about the need for lots of power for quick reactions. And of course the system must maintain lateral stability at the same time - and on uneven surfaces with heights that vary at each step such as a track through a field or a cobblestoned pavement.
When you start to think about it, if a biped is standing still it can't simply lift a leg and stretch it out in front of itself (as a prelude to a first step) without falling on its face. It must also adjust its c-of-g to compensate.
This also reminds me that I have seen videos on-line of one-legged (monoped?) hopping robots. That might actually be an easier place to start.
In the book Legged Robots That Balance, the progression was from a one legged robot to a biped and then to a quadruped. While that may seem to be going from the least stable to the most, it was apparently easier to develop the software that way.
This project, considering its forerunners, has probably cost a bit less than $100 million. Here is the U.S. no sacrifice is too much in the pursuit of military superiority!
That's our government at work. Spend $100 million instead of buying a $2000 mule. The mule does the job 5 times better.
pharseid379:
In the book Legged Robots That Balance, the progression was from a one legged robot to a biped and then to a quadruped. While that may seem to be going from the least stable to the most, it was apparently easier to develop the software that way.
Marc Raibert
Guess what company he founded (is it any wonder)? I remember reading (and keeping copies) of the SciAm articles about his work when I was in high school and interested in walking robots...
