Trajectory Tracking of Robotic Arms Using Learning Modified Transpose Jacobian

In this research, the issue of trajectory tracking in task space of a robotic arm with rotational joints has been solved. For this purpose, a non-model based controller the modified jacobian-transpose is used which it’s performance has been improved by reinforcement learning. Traditionally, the settings of the error thresholds for this controller will require trial and error, but in this algorithm, reinforcement learning is used to adjust these parameters online and according to the conditions. The simulations for the desired trajectory indicate that the improvement of the trajectory tracking achieved by using proposed algorithm. Also, the superiority of this controller in noisy conditions compared to the reference Jacobian transpose controller has been shown.