Mínimos Cuadrados Recursivos para un Manipulador que Aprende por Demostración

In this work, a control system is developed to allow a manipulator to learn and plan trajectories from demonstrations given by the hand of an user. The input of data is acquired by a sensor, and its behavior is learned through an automatic learning algorithm based on the recursive least squares. A trajectory profile of interpolators to stretches is used to avoid the impulsive jerk on manipulators motion. Direct and inverse kinematics analysis is done for obtaining the joints variables values of the manipulator. A dynamic model is created using Newton-Euler formulation. A proportional derivative control is applied to the system. The monitoring and control systems are implemented in an embedded platform for testing purposes.