Robot Learning Human Skills and Intelligent Control Design

Learning from demonstrations (LfD) is a method for robots to acquire skills from human demonstrators, enabling them to perform new tasks quickly. This book discusses advanced LfD-based learning and control approaches to improve robot dexterous manipulation, including transferring human arm variable stiffness to the robot, modeling motion and impedance profiles, learning correlation between signals, and using neural network enhanced control strategies. Examples of simulation and experiments are included to facilitate better understanding.

Format: Paperback / softback
Length: 174 pages
Publication date: 25 September 2023
Publisher: Taylor & Francis Ltd

In recent years, the expectation has grown that robots will become increasingly intelligent, capable of handling a wide array of tasks. Particularly, robots are expected to acquire manipulation skills from humans. To achieve this goal, numerous learning algorithms and techniques have been developed and successfully implemented for various robotic tasks. Among these methods, learning from demonstrations (LfD) stands out as a powerful approach, enabling robots to rapidly acquire skills by observing and mimicking human demonstrators.

This book presents the latest advancements in developing advanced LfD-based learning and control approaches to enhance robot dexterous manipulation. Firstly, it provides an introduction to the simulation tools and robot platforms employed in the authors' research. To enable a robot to learn human-like adaptive skills, the book explains how to transfer a human user's arm variable stiffness to the robot, based on online estimation from muscle electromyography (EMG). Next, the motion and impedance profiles can be modeled using dynamical movement primitives, allowing both to be planned and generalized for new tasks. Furthermore, the book introduces how to learn the correlation between signals collected from demonstration, such as motion trajectory, stiffness profile estimated from EMG, and interaction force, using statistical models like hidden semi-Markov models and Gaussian Mixture Regression. Several widely used human-robot interaction interfaces, such as motion capture-based teleoperation, are presented, enabling human users to interact with robots and transfer movements to them in both simulation and real-world environments.

Finally, the book discusses the use of neural networks to improve the performance of robot manipulation. By leveraging the power of neural networks, robots can learn and adapt to complex tasks, making them more efficient and effective in performing various manipulation operations.

In conclusion, this book offers valuable insights into the development of advanced LfD-based learning and control approaches for improving robot dexterous manipulation. It provides a comprehensive overview of the simulation tools, robot platforms, and learning algorithms used in this field. By leveraging the power of LfD, robots can become more capable and adaptable, enabling them to perform a wide range of tasks with greater efficiency and precision.

Weight: 453g
Dimension: 234 x 156 (mm)
ISBN-13: 9780367634377