Controle servo visual aplicado a manipuladores robóticos

Abstract

In the last few decades there has been a considerable raise in the use of visual sensors for guiding robots in their tasks. These sensors increase the flexibility of the robot and make it possible for it to act in situations where the environment is unknown or when it is not possible to ensure the position of a target object. Visual servoing is a strategy that uses information from a camera to guide a robot to a visual reference. With the theoretical framework of Classic Control Theory, the insertion of visual servoing can increase precision in tasks as well as its reliability, besides being a low cost alternative when compared to the usage of other sensors that would do the same thing. There are basically two great strategies in visual servoing, one based in position and the other based in image. The first uses a visual reference to estimate the robot’s pose, calculate the error in relation to the desired pose and then calculate the control signal. The second extracts the positions of some characteristics in the image and compares them to the positions the robot should see when in the desired position. Its application in manipulator robots has gained great importance with the increase in processor capabilities of computers in the last few years and the advance of microcontrollers allowed this technique to be introduced also in mobile robotics. This work proposes to apply and evaluate the results of a visual controller based on image on a manipulator robot ABB IRB140, using a low cost camera. The approach based on image is robust in relation to the variation of its parameters, and lower computational cost, since it does not require reconstruction of a tridimensional image. The main goal of this study is to evaluate the performance of the system when using a proportional controller, as well as ascertain the difficulties of implementation in practice. To evaluate the results, a C++ computer vision routine was elaborated to extract the characteristics of the image of a target object. It was captured by a webcam placed directly above the robotic arm. The control calculations were made by an external computer and sent through the ethernet communication to the robot’s high level controller, programmed in the RAPID language. Tracking scenarios were analyzed for reference for approaching and withdrawal from the target object and response to disturbances. The performance was satisfactory for all cases, because the capacity to follow the reference was established for the technique applied. For small gains, the response had a long settling time and a soft trajectory. These characteristic are changed as the gain is raised, decreasing settling time in order to make the trajectory mildly oscillatory.