Comparative Study of Machine Learning Models for Inverse Kinematic Prediction of a Flexible-Tube Wrist Mechanism in Robotic Charging Stations

In robotic charging stations for electric vehicles (EVs), precise plugging is essential for efficient electric power transfer. This paper presents a comparative study of various machine learning (ML) models for inverse kinematic prediction of a wrist mechanism equipped with a flexible tube, which can passively deflect in six degrees of freedom (6-DOF) due to external forces. The study also integrates Finite Element Method (FEM) analysis to simulate static deflection of the flexible tube under applied loads, complementing the ML predictions. ML models, including Feedforward Neural Network (MLP), Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), Random Forests, XGBoost, LightGBM, Support Vector Regression (SVR), Gaussian Processes (GP), and hybrid approaches were compared in terms of their accuracy in predicting joint angles (yaw, pitch, roll) given the end-effector positions and tube deflection. Using Root Mean Squared Error (RMSE) as the evaluation metric, MATLAB simulations shed light on each model's performance. The outcomes demonstrate the advantages of integrating data-driven and physics-based approaches for challenging inverse kinematic tasks.