Torque Measurement Using Regression-Based Calibration Approaches; A Mobile Robot Perspective

This study presents a methodology for calculating the rotational torque values of the DC actuators, which is then applied and evaluated in a mobile robot scenario. The influence of torque-sensing on robot applications is substantial. This study focuses on enhancing the accuracy of torque measurement by constructing a regression model that establishes the relationship between torque and the input voltage applied to the motor and the angular velocities of the motor and wheel. The model is trained using data from an electrodynamometer framework that utilizes a counterweight mechanism to generate linear torque. The input voltage is adjusted as a percentage of the entire load value. Relevant sensors are used to monitor the electro-mechanical characteristics, and this data is utilized to generate the model. The framework is compared to a second model that establishes a relationship between torque and current signals. The previous model exhibits superior performance in the regression analysis. The model's efficacy is evaluated in a mobile robot through a case study that predicts wheel slippage. This prediction is accomplished by employing machine learning algorithms that utilize torque signals as input features. Both models are utilized to compute the torque values, and the average performance of the learning algorithms is compared, with the first regression model demonstrating superior results over the second. Therefore, the suggested method of measuring torque by calibrating voltage and angular velocity through regression analysis is suitable for achieving both cost-effectiveness and accuracy in measuring torque in robot actuators.