Robot Learning By Visual Observation

The book presents the methodology for robot learning from observations. The focus of the book is on trajectory level of task abstraction. The content is divided into several chapters, with each chapter addressing the methods for tackling individual steps in the observational learning problem. The book describes several methods for mathematical modelling of observed trajectories, such as hidden Markov models, conditional random fields, Gaussian mixture models, and dynamic motion primitives. Methods for generalization of the demonstrated trajectories and for generation of a trajectory for task reproduction are presented. Furthermore, the authors describe methodology for robust execution of generated plans for task reproduction, which employs vision based robot control.

Aleksandar Vakanski is a Clinical Assistant Professor in Industrial Technology at the University of Idaho, Idaho Falls. He received a Ph.D. degree from the Department of Mechanical and Industrial Engineering at?Ryerson University,?Toronto, Canada, in 2013. The scope of his research interests encompasses the fields of robotics and mechatronics, artificial intelligence, computer vision, and control systems.