The difficulties of controlling a robot stem from the complexity of the dynamics of the robot manipulator.
To achieve desired motions for given path-dependent jobs, the learning control algorithms which make the robot motion closer to the desired trajectory by repeating operations were proposed recently.
However, the learning control algorithms which require, even conventional computed torque algorithms, the dynamics calculations and torque type servo controller were proved only by computer simulation except a few cases.
In this paper, the learning control algorithms were implemented to the SCARA (Selective Compliance Arm for Robotic Assembly) type robot to show the feasibility and effectiveness of such algorithms.
For this purpose, the dynamic equations of the robot manipulator are calculated and converted to the equivalent dynamics with respect to the actuator and finally added to the actuator dynamics.
Other important terms such as viscous friction and coulomb friction, which can't be obtained using the above method, are found by the experiments.
The learning control algorithms were implemented to the SCARA system which consists of a SCARA type robot itself, torque type servo controllers and MVME 110 single board computer to calculate the dynamics and the control laws.
The control inputs(torques) generated from the dynamic equations are applied to the torque type servo controllers and the responses of the robot manipulator which are measured by the sensing devices are used for generation of the control inputs.