Statistical modeling of heterogeneous robotic assembly time with Weibull regression

It is critical to model the relationships between robotic control parameters and the performance of the robotic systems. Existing modeling methods generally assume homogeneously distributed performance data. However, such homogeneity assumption may not be realistic in industrial practices. With explicit consideration of potential data heterogeneity, this paper proposes a Weibull regression and an EM algorithm to model the impacts of robotic control parameters on the time for robots to successfully complete a task. A numerical case study shows the high accuracy of the model parameter estimation. It demonstrates that the proposed method without homogeneity assumption is effective and can be applied in many real-world problems.