A Methodology for Pneumatic Actuation via Visual Feedback

Pneumatic actuation systems are widely used in industrial automation due to their high power-to-weight ratio, fast response, and cost-effectiveness. However, achieving precise position control in pneumatic systems presents certain challenges. This is due to the compressible nature of air and the nonlinear characteristics of the system. Therefore, this thesis proposes a novel experimental setup that integrates visual feedback with pneumatic actuation. The system achieves high-precision one-dimensional position control of a 9mm diameter steel ball within a horizontal transparent tube. The experimental setup uses an industrial camera for real-time position tracking and feedback. A Raspberry Pi 5 serves as the system control core responsible for image processing, PID control algorithm execution and signal coordination. A five-way three-position solenoid valve provides pneumatic actuation, and pressure sensors monitor system pressure. Two controllers are implemented: static PID controller with constant gains and dynamic fuzzy adaptive PID controller with variable gains. The machine vision system can achieve sub-pixel accuracy of ±0.1 pixel (±29 microns) in position detection. Experimental results demonstrate that both controllers successfully enabled smooth ball movement. The actual position curve highly overlapped with the set position curve, representing excellent control performance. This integrated vision-pneumatic control system shows significant potential for applications requiring precise positioning in manufacturing, automation, and robotics.