This study aims to address the high cost, inter-axis error amplification and circular verification defects of traditional Stewart parallel robot calibration, to meet SMEs’ low-cost and high-precision calibration demands with dual-camera three-dimensional imaging.
Fixed transformations are acquired via rigid molds, vernier caliper measurements and medium-angle visual verification. An inter-axis adaptive weighted least squares strategy is designed, and a dual verification framework combining repeat positioning accuracy and assembly success rate is built. The method is validated via theoretical derivation, simulation and physical experiments.
Post-calibration position/attitude error RMSE reaches 0.5435 mm/0.0163 rad, outperforming mainstream low-cost algorithms by 38.0–41.0% and 29.4–36.6%. The assembly success rate hits 93% with hardware cost merely 5% of traditional schemes.
The method has a slight systematic error (=0.02 mm) in fixed transformation measurement and is suitable for scenarios without strict absolute accuracy requirements.
It enables low-cost automated calibration, providing a feasible solution for SMEs to implement robot automation and promoting intelligent manufacturing popularization.
This work proposes three innovations: a high-end equipment-independent fixed transformation acquisition method, a variance-based weighted optimization strategy and a noncircular dual verification framework, achieving a systematic integration of theory, simulation and experiment.
