6DOF robot arm controlled by speech recognition or EMG
Construction of the arm
The arm is made from 2mm ABS sheet. This material is very easy to cut with a carpet knife, can be very well glued with superglue and be easily bent by warming it up with a hot air gun, so you can make your own brackets, etc.
As the Tower Pro SG90 is a little bit weak to act as a joint to the gripper I increased the locating surface to stabilize the servo horn:
Complete upper part assambled:
Turn table using casters for furniture as support :
Robot arm mounted on the base. The base is made from a piece of floor panel and then "coated" with 1mm ABS sheet. Floor panels are heavy, easy to cut and do not deform:
Power supply design for the arm. The LM138 will need a heatsink. It is important to use 6V for the servos to squeeze out the biggest torque. At some movements the current consumption can easily exceed 3A, so it is not recommanded to use batteries.
Power supply, built on a perfboard:
Servos moving pretty fast (too fast) for a robot arm, which leads to big forces on some joints and parts, especially if the arm is fully extended. It seems to be a good approach to slow down the servo movements. Here is a simple code example how to do that:
for b0=1 to 50
for b0=1 to 100
for b0=1 to 50
For first electromyography experiments I will use this circuit:
The schematic diagram of the electromyographic (EMG) amplifier is borrowed from this site.
The electrodes are made from 0.5 RMB coins, using common sticking plaster to attach them on my biceps and elbow:
Two joysticks from an old, damaged RC RX added to control the robot arm manually (see first video). I am planning to store different trained movements in an eeprom.