Frequency-Response-Guided Iterative Learning for a Low-Cost SCARA with an M–N Tuning Guideline and Simulation Study

Inexpensive Selective Compliance Assembly Robot Arm (SCARA) manipulators that rely solely on proportional feedback often experience consistent tracking problems across multiple motion cycles. This study presents a frequency-response-guided serial iterative learning control (ILC) that utilizes a finite impulse response (FIR) learning filter, along with a practical M-N tuning guideline to ensure spectral stability. The primary objective is to evaluate the effectiveness of FIR-based learning filters within an ILC architecture through extensive simulation research employing MATLAB and Simulink. A miniature SCARA, equipped with brushed DC motors and encoder feedback, was designed to provide discrete-time representations for each joint. All investigations were performed only in simulation, employing a continuous band-limited pick-and-place trajectory replicated across multiple iterations to distinguish learning behavior from hardware latency and jitter. FIR learning matrices with dimensions 2×2, 4×4, 6×6, and 8×8 were assessed. The proportional feedback alone resulted in mean root-mean-square (RMS) position errors of 1.52°, 1.32°, 1.07°, and 1.61° for joints A, B, C, and Z. The application of ILC reduced the steady-state RMS error by more than 98% at the 6×6 matrix, yielding values of 0.023°, 0.020°, 0.014°, and 0.016°, respectively. No learning transients were detected, and spectral analysis confirmed that the spectral radius and the maximum singular value remained below one for all identified models. Smaller filters showed reduced convergence rates, while larger filters offered negligible improvement. The proposed proportional-plus-iterative learning architecture achieved sub-degree tracking accuracy in a purely simulated, low-cost SCARA environment.