The task function approach applied to vision-based control

This paper describes some concepts and results related to the vision-based control approach in robotics. The basic idea consists in considering a vision system as a specific sensor dedicated to a task and included in a control servo-loop. Once the necessary modeling stage is performed, the framework becomes the one of automatic control, and, naturally, stability and robustness questions arise. Starting from the task function approach, the general framework of the control is described, and some stability results are recalled. The concept of hybrid task is also presented and applied to visual sensors. Some simulation and experimental results are finally given.<>