Accurate robot calibration via cascaded adaptive momentum LM and B-spline interpolated PSO

To improve robotic positioning accuracy and enhance overall manufacturing precision, this paper proposes an adaptive momentum Levenberg-Marquardt cascaded B-spline interpolation particle swarm optimization (AMLM-BIPSO) algorithm for calibrating robotic geometric errors. Initially, a momentum term is incorporated into the traditional Levenberg-Marquardt algorithm to suppress overshooting and oscillation, thereby improving the preliminary estimation of geometric parameters. Subsequently, inspired by the concept of Knowledge-based artificial Neural Networks, B-spline interpolation is embedded into the standard particle swarm optimization framework to refine the final calibration. By cascading these two enhanced techniques, the proposed method achieves higher accuracy in parameter identification. Experimental validation on an industrial robot confirms that the AMLM-BIPSO algorithm yields substantial improvements in positioning accuracy and calibration reliability.