Surrogate modeling for time-dependent reliability analysis of robotic manipulator trajectories

The kinematic reliability analysis of robotic manipulators is crucial due to uncertainties such as joint variations, manufacturing tolerances, and external disturbances. Traditional methods often rely on analytical techniques that struggle with nonlinear performance functions and fail to account for trajectory-based reliability. To overcome these limitations, this paper proposes a novel surrogate model-based approach using Kriging to estimate the reliability of robotic manipulator kinematics while considering end-effector trajectories. The methodology begins with building an initial Kriging surrogate model to analyze reliability, effectively capturing how input uncertainties influence trajectory accuracy. This model is then refined through statistical sampling techniques, ensuring an efficient evaluation of manipulator performance against specified tolerances. The approach reduces computational complexity while maintaining prediction accuracy. Compared to Monte Carlo Simulation (MCS), the proposed Kriging-based method reduces the number of function evaluations by over 98%, achieving comparable reliability predictions with significantly fewer function calls, and enhancing efficiency in kinematic reliability analysis. The proposed method is validated on two 6-DOF industrial robots, including the UR5, demonstrating improved computational efficiency and accuracy. This work has practical applications in manufacturing and healthcare, where enhanced kinematic reliability leads to greater operational efficiency and safety.