When
Thematic Session 3: Modeling and Analysis (Tuesday, 09:40)
Abstract
In musical robotics, one can think of a robot that perfectly plays an instrument, or overperforms human musicians' abilities. Usually, the benchmark for evaluating the performance of a robot can be the performance of human musicians. The challenge here is to make a machine limited by the huge difference between the neuromuscular system of the human body and the electromechanical body of a robot. How can a robot find its own voice according to its own body and physical constraints? Can a robot exploit the dynamics of the interaction with the instrument to find its unique voice instead of replicating human behaviour? In this talk, these questions will be addressed in a real-world application using drum robots. The robots are designed and built to play drumming tasks such as drum rolls and are capable of adapting to different physical characteristics such as stiffness. The robots need to have some sort of embodied cognition in order to exploit the physical potentials of the body and the instrument. We will show how this can be developed and implemented.
Bio
Mojtaba Karbasi is a PhD fellow at RITMO Centre for Interdisciplinary Studies in Rhythm, Time and Motion, at the University of Oslo. His research is focused on robot control and learning, reinforcement learning, cognitive robotics, and motor control theories. He is currently working on drum robots which are designed to learn drumming tasks through interaction with the environment. In this work, he is exploring reinforcement learning methods and motor control models to develop an interactive robotic system for playing the drum.