An Environment-Adaptive Position/Force Control Based on Physical Property Estimation
Authors:
- Tomoya KitamuraTKTomoya Kitamura
- Yuki SaitoYSYuki Saito
- Hiroshi AsaiHAHiroshi Asai
- Kouhei OhnishiKOKouhei Ohnishi
Preprint
Abstract
The technology for generating robot actions has significantly contributed to the automation and efficiency of tasks. However, the ability to adapt to objects of different shapes and hardness remains a challenge for general industrial robots. Motion reproduction systems (MRS) replicate previously acquired actions using position and force control, but generating actions for significantly different environments is difficult. Furthermore, methods based on machine learning require the acquisition of a large amount of motion data. This paper proposes a new method that matches the impedance of two pre-recorded action data with the current environmental impedance to generate highly adaptable actions. This method recalculates the command values for position and force based on the current impedance to improve reproducibility in different environments. Experiments conducted under conditions of extreme action impedance, such as position control and force control, confirmed the superiority of the proposed method over MRS. The advantages of this method include using only two sets of motion data, significantly reducing the burden of data acquisition compared to machine learning-based methods, and eliminating concerns about stability by using existing stable control systems. This study contributes to improving robots' environmental adaptability while simplifying the action generation method.Keywords
All Related Versions
This publication has 17 references indexed in Scilit:
- Soft and Rigid Object Grasping With Cross-Structure Hand Using Bilateral Control-Based Imitation LearningIEEE Robotics and Automation Letters, 2023
- Learning to Shape by Grinding: Cutting-Surface-Aware Model-Based Reinforcement LearningIEEE Robotics and Automation Letters, 2023
- Reinforcement Learning for Robotic Assembly Using Non-Diagonal Stiffness MatrixIEEE Robotics and Automation Letters, 2021
- Motion Generation Using Bilateral Control-Based Imitation Learning With Autoregressive LearningIEEE Access, 2021
- Robots and Jobs: Evidence from US Labor MarketsJournal of Political Economy, 2020
- Survey of imitation learning for robotic manipulationInternational Journal of Intelligent Robotics and Applications, 2019
- Artificial Intelligence for Long-Term Robot Autonomy: A SurveyIEEE Robotics and Automation Letters, 2018
- Stability Analysis and Experimental Validation of a Motion-Copying SystemIEEE Transactions on Industrial Electronics, 2009
- Designing autonomous robotsIEEE Robotics & Automation Magazine, 2009
- Torque sensorless control in multidegree-of-freedom manipulatorIEEE Transactions on Industrial Electronics, 1993