Repetitive motion error compensation through iterative learning based calibration with application to flatbed document scanner

Moeed Mukhtar, Purdue University

Abstract

This thesis details the research performed for development of motion control algorithms which improve the performance of flatbed document scanners with stepper motors and belt drive systems. Specifically, high speed scanning with the best possible image quality are the two main concerns of scanner manufacturers due to market competition. The research, therefore, could be divided into two phases. In the first phase, a velocity synthesis algorithm based on time optimal command shaping was developed to reduce vibration and overshoot during the ramp-up velocity transient region. A detailed first-principles based dynamic model of the process was also derived and rigid body motion along with peak acceleration constraints were imposed to solve an optimization problem for minimum move time. Experimental results demonstrate the effectiveness of the proposed approach in reducing overshoot in ramp up velocity transients thereby shortening the initial velocity stabilization buffer by 8% and potentially reducing the scan time. In the second phase, a motion induced artifact called color registration error was investigated. First, to characterize the color registration error, a line pattern image based technique was devised. Using the measured color registration error signal, a single iteration convergent iterative learning control scheme was proposed to synthesize a stepper motor step timing command for flatbed document scanners with no real-time feedback. Reduction in color registration error was achieved by compensating for velocity fluctuations in the constant speed portion of the motion. Experimental results demonstrate the effectiveness of the proposed approach in reducing color registration error by around 55%. The iterative learning control algorithm was computationally intensive and required fast processing and large memory resource. Exploiting the repetitive nature of the color registration error, a hardware friendly compensation filter belonging to the IIR class of filters was designed by combining iterative learning control with quadratic optimal zero phase repetitive control. Asymptotic stability and error convergence analysis were performed and stability bounds on controller gains were also established. The proposed algorithm requires less computation and has comparable performance, as demonstrated by reduction in color registration error by around 59%.

Degree

Ph.D.

Advisors

Chiu, Purdue University.

Subject Area

Mechanical engineering

COinS