Vision-Based Digital Controlled Systems Using Real-Time Multiprocessing

Real-time jobs such as vision sensing in motion control that have difficulty meeting time constraints because tasks take too long to compute can be run on a number of processors. Parallelism in task operation can be achieved by providing each task an individual processor. To accomplish the goal, a nonconventional vision system, known as FIVS, has been employed to demonstrate the concept. The paper illustrates the vision-based motion control with a project in a digital control course taught in the Intelligent Mechatronics Laboratory (IML), a graduate course in the Woodruff School of Mechanical Engineering at Georgia Tech. IML, which has been jointly initiated by the Computer Integrated Manufacturing Systems (CIMS) and the AT&T Foundation, is a campus-wide facility to support teaching engineers to integrate mechanical engineering, electronics, control engineering and computer science into fundamental design process. The course is taught in a lecture/lab format, with heavy emphasis on laboratory experience. The laboratory practice introduces C language, DSP-based real-time simulation and control, identification and estimation, multiprocessing real-time systems, and uses of digital vision system as a feedback element in control systems.