article · Applied Computer Systems
Calculating inverse kinematics for four-degrees-of-freedom SCARA robots can be difficult and computationally demanding using standard analytical methods. This study investigates the use of artificial neural networks to resolve and optimise these calculations while minimising mean squared error. The network performance was refined using three distinct training algorithms: Levenberg-Marquardt, Bayesian Regularisation, and Scaled Conjugate Gradient. These models were assessed against three generated datasets representing varied operational scenarios, specifically fixed step sizes, random step sizes, and sinusoidal trajectories. The comparative evaluation reveals that neural networks provide heightened computational efficiency and precision, making them well-suited for real-time robotic control and trajectory planning tasks. The findings offer guidance on identifying the most effective training configurations to improve control systems in manipulator robots where conventional analytical approaches encounter limitations.
Industrial manipulator robots rely on fast and accurate mathematical calculations to move precisely. Traditional calculation methods can be slow and computationally heavy. By demonstrating that artificial neural networks can reliably handle these calculations across diverse movements, this research helps support faster, more responsive automated robotic operations.
This work could enable faster real-time trajectory planning and control systems for industrial automation equipment, particularly four-degrees-of-freedom SCARA robots. Potential users include robotics manufacturers and automation engineers seeking alternatives to computationally heavy analytical models. Based on the abstract, the research represents early-stage development, as the models have been validated on generated datasets rather than deployed and tested in an operational industrial setting.
AI-generated from the published abstract. Always read the original work before citing.
Abstract In the pursuit of enhancing the efficiency of the inverse kinematics of SCARA robots with four degrees of freedom (4-DoF), this research delves into an approach centered on the application of Artificial Neural Networks (ANNs) to optimise and, hence, solve the inverse kinematics problem. While analytical methods hold considerable importance, tackling the inverse kinematics for manipulator robots, like the SCARA robots, can pose challenges due to their inherent complexity and computational intensity. The main goal of the present paper is to develop efficient ANN-based solutions of the inverse kinematics that minimise the Mean Squared Error (MSE) in the 4-DoF SCARA robot inverse kinematics. Employing three distinct training algorithms – Levenberg-Marquardt (LM), Bayesian Regularization (BR), and Scaled Conjugate Gradient (SCG) – and three generated datasets, we fine-tune the ANN performance. Utilising diverse datasets featuring fixed step size, random step size, and sinusoidal trajectories allows for a comprehensive evaluation of the ANN adaptability to various operational scenarios during the training process. The utilisation of ANNs to optimise inverse kinematics offers notable advantages, such as heightened computational efficiency and precision, rendering them a compelling choice for real-time control and planning tasks. Through a comparative analysis of different training algorithms and datasets, our study yields valuable insights into the selection of the most effective training configurations for the optimisation of the inverse kinematics of the SCARA robot. Our research outcomes underscore the potential of ANNs as a viable means to enhance the efficiency of SCARA robot control systems, particularly when conventional analytical methods encounter limitations.
This page summarises published work. The authoritative version sits with the publisher.
DOI: 10.2478/acss-2024-0004
Is something wrong with this record? Report it or request removal.
Discussion
Have you built on this work, tried to replicate it, or seen it applied in practice? Share what you know. Verified researchers and MARATTO™ domain experts can open a discussion, and any member can reply. Contributions are reviewed before they appear.
No discussion yet. Open the first thread.
New to MARATTO™? Create a free account.
