Force-based centre of mass estimation technique for rigid objects in robotics applications

Abstract

This project aims to develop a method to estimate the 3D inertial parameters of an object using a Kinova robot arm and force/torque sensors. Inertial parameters, such as the centre of mass, are crucial for achieving stability in grasping and manipulation tasks. The current methods for estimating these parameters require specialized hardware, making assumptions about the object or having extremely controlled environments. The proposed force-based method involves attaching the object to the end effector and moving it through a trajectory. The forces and torques measured along this movement are used to regression the centre of mass coordinated in 3D space. The scope of this research is limited to estimating the centre of mass of rigid, non-deformable objects. The experimental procedure involves using a 3D printed rectangular hollow box with calibration weights screwed into different locations to vary the centre of mass. An extensive calibration of the force/torque sensor used is also undertaken to minimize the prediction error. The error of the measured centre of mass is evaluated against the expected centre of mass obtained from the Solidworks model of the box. This research has significant implications in producing meaningful centre of mass estimation and loadcell readings that can be used for improving the stability and efficiency of robotic manipulation tasks, in particular box tilting tasks for warehouse automation.