Image based visual servoing for robot positioning tasks

Visual servoing has become a popular para- digm for the control of complex robotic systems: this sensor based approach exploits the image informations provided by one ore more cameras in a feedback con- trol loop to drive the system to the desired configura- tion. Here authors will refers to a monocular system where the camera is mounted on the end effector of a 6-DOF manipulator. Among different Visual Servo- ing approaches Image Based Visual Servoing (IBVS) has been the most investigated in the literature because of its nice properties of robustness with respect to both robot modeling and camera calibration errors: in IBVS the control loop is in fact directly closed in the im- age; moreover IBVS doesn’t require the knowledge of the target/scene model (model-free approach). De- spite its advantages IBVS may be affected by singu- larity and local minima problems of the control law: these drawbacks arise especially when the initial and the goal camera images respectively corresponding to the actual and desired system configurations are very different (i.e for large system displacements). To over- come these problems an image path planning can be exploited to ensure system convergence. In this paper author presents an off-line image path planning that can be used to execute system positioning task also in presence of large camera displacements: planning trajectories has been developed such as to make the robot end effector move on a 3D helix, connecting the initial and the desired arm configuration, by generating fea- sible robot twistscrews and keeping the target in the image field of view. During control execution also 3D target informations are retrieved through an adaptive estimation law. Both simulations and experimental re- sults show the feasibility of the proposed approach.