Experimental data and analysis outputs for "Learned dynamics or absorbed calibration: attributing the tracking advantage of bilinear Koopman MPC on a differential-drive robot"

Experimental data and analysis outputs for the paper “Learned dynamics or absorbed calibration: attributing the tracking advantage of bilinear Koopman MPC on a differential-drive robot” (submitted to Robotics and Autonomous Systems, 2026). It contains open-loop excitation runs (spin tests, PRBS, chirps, multisine, etc.) and closed-loop tracking sessions on a lemniscate using model predictive controllers. The logs identify the controllers as C1 to C7; the paper maps them to descriptive labels: NMPC (C1), KIN (C2), KMPC (C4), KMPC_cmd (C5), KMPC_ridge (C6), and CAL / attribution baseline (C7), while excluding C3. The record includes raw telemetry, camera native poses, encoder feedback, metadata, calibration files, and analysis outputs. The raw data and folder layout match the code repository (Zenodo concept DOI 10.5281/zenodo.23324968). Copyright (c) 2026 Ashkan Jelodar. Released under the Creative Commons Attribution 4.0 International licence (CC BY 4.0). Contact: Ashkan Jelodar, [e-mail].