Category Archives: Visual Servoing

Visual Servoing -part I( Real-Time Pan & Tilt Tracking System )

Visual Servoing (VS) is a control technique in which a vision sensor provides feedback information to control a robot movement. In this project we utilize VS as an approach for some robotic platforms.

In the first part, As a common example of VS we implemented a fast Pan & Tilt tracking system. The aim of such a system is to keep it’s look on a certain moving object. In this manner, first a camera detects the object then using two servo motors, camera is rotated about pan and tilt axes so as to bring the object position -described in image frame- to the center of the image.

The mentioned system uses two Dynamixel servo motors and a high frame rate camera to perform sufficiently fast (120 fps) in order not to miss fast movements of the object even when the object is falling down.In addition, Machine Vision and Image Processing algorithms are implemented under QT framework by making use of fast C++ library for real-time purpose called CMVison.

(hbazargani{at}engcast{dot}com)

Visual Servoing –part II( Planar Arm Trajectory Planning )

In the second part of VS project to implement a two-link planar arm and to make it interact with other objects such as ball, obstacles and goal, we need to estimate object position in real world coordinates. It is done using curve (surface) fitting techniques to map from pixel frame to real world Cartesian coordinate.

For the purpose of making the arm move from current to desired position with fixed velocity or acceleration we need to solve Trajectory Planning problem. For this purpose we consider three approaches. Position Control is the first one in which we define several via points and solve Inverse Kinematic equations.The second approach is Velocity Control that utilizes Inverse Jacobian to maintain end effector velocity in proper way. In the third approach that is a combination of previous ones, the linear velocity is fed to Inverse Jacobian so as to compensate position error and avoid loss of efficiency near the singularity points.

(hbazargani{at}engcast{dot}com)