A versatile parallel architecture for vision based control applications

Abstract

The authors discuss a hardware and software architecture for implementing vision based control approaches applied to robotics. In such approaches, the sampling rate of the robot's closed loop control is directly given by the image processing time. That implies strong time constraints to the vision system. In a previous paper, the authors presented the design phase of a new machine vision based on a parallel architecture concept and first simulation runs concerning its expected performances were also shown. The complete vision machine is now available and the authors detail both its hardware and software aspects. They also present the first results obtained for two real vision based control applications.