I am with a group that has been tasked with designing a 3 foot (or as tall as possible) humanoid robot with a budget of under $3k.
Our main controller will be a RoBoard RB-110
We will model our robot in Webots, and use that to design preliminary control software while construction is underway.
Our main problem is to physically design something that is strong and powerful enough to cope with the joint forces generated, and not spend too much.
The budget puts high grade $500 dynamixel servos out of range, so I wanted to see if anyone had creative, cheap solutions for generating the huge torque needed. Considering the robot will carry the batteries on its back, knee joint torques will be expected to be over 1500oz-in.
We have access to CNC equipment for free.
Some ideas- Use large, heavy (3lb) windshield motors for leg joints, build external magnetic absolute encoders into the joints (pwm output). Use separate H-bridge control board to power these motors. These motors are cheap >$100, and can generate over 1500oz-in peak torque at over 60 RPM. This will make beastly, heavy legs, but they will be strong, able to lift the robot, and will lower the center of gravity.
Another idea is to use servos with less torque, about 700oz-in, but build springs/rubber bands into the joint to store energy while the knee is bent, allowing it to lift more when it is straightened. This will probably cause stability/control problems in practice.
Couple higher speed DC motor to a threaded rod, and have a load bearing-pivoting nut nut on a lever out from the leg joint. This would be lighter weight, built in gear reduction, but not very efficient, and require more complex physical design.
Anyone have experience with building powerful joints on a budget?