Control por guiado gestual para un robot antropomórfico basado en técnicas de visión artificial.

Abstract

In this project a gestural guidance control was developed for the anthropomorphic robot Raplim of the University of Pamplona UP from the capture of the movements of an operator through a structured light camera, allowing the three-dimensional reconstruction of the movement to program trajectories taking into account the kinematics of the robot. From the gestural commands, the capture of movements is initiated and completed using artificial vision techniques, these movements are transformed into trajectories by applying kinematic control and these trajectories can be simulated in the robot's workspace by applying augmented reality. For the implementation of the project, optimization algorithms were developed to identify the intention of the trajectory that the user wants to perform. The applied research methodology is of experimental type where some forms were tested based on the trajectories that were simulated in the real environment through augmented reality with a virtual robot giving validity to the developed algorithm showing that these trajectories are similar to the captured movements. The results of the tests for each model have a better representation in the XY plane due to the fact that the capture made by the structured light camera presents variation in the Z axis in the identification of the point of the joint of interest with which the path is made to generate the trajectory. Thus, since the trajectory must be generated in three-dimensional space, filters were developed to reduce the effect of the variations in the Z axis of the capture. Having defined the above, it is evident that programming trajectories for a robot from the gestures or movements of an operator can be a very complex task that was simplified by applying optimization algorithms and filters to reduce the error presented by the capture with respect to the sample model.