Seyed Mojtaba Karbasi

This paper investigates the potential of the intrinsically motivated reinforcement learning (IMRL) approach for robotic drumming. For this purpose, we implemented an IMRL-based algorithm for a drumming robot called ZRob, an underactuated two-DoF robotic arm with flexible grippers. Two ZRob robots were instructed to play rhythmic patterns derived from MIDI files. The RL algorithm is based on the deep deterministic policy gradient (DDPG) method, but instead of relying solely on extrinsic rewards, the robots are trained using a combination of both extrinsic and intrinsic reward signals. The results of the training experiments show that the utilization of intrinsic reward can lead to meaningful novel rhythmic patterns, while using only extrinsic reward would lead to predictable patterns identical to the MIDI inputs. Additionally, the observed drumming patterns are influenced not only by the learning algorithm but also by the robots¡¯ physical dynamics and the drum¡¯s constraints. This work suggests new insights into the potential of embodied intelligence for musical

ZRob is a robotic system designed for playing a snare drum. The robot is constructed with a passive flexible spring-based joint inspired by the human hand. This paper describes a study exploring rhythmic patterns by exploiting the chaotic dynamics of two ZRobs. In the experiment, we explored the control configurations of each arm by trying to create un- predictable patterns. Over 200 samples have been recorded and analyzed. We show how the chaotic dynamics of ZRob can be used for creating new drumming patterns.

Intelligent robots aimed for performing music and playing musical instruments have been developed in recent years. With the advancements in artificial intelligence and robotic systems, new capabilities have been explored in this field. One major aspect of musical robots that can lead to the emergence of creative results is the ability to learn skills autonomously. To make it feasible, it is important to make the robot utilize its maximum potential and mechanical capabilities to play a musical instrument. Furthermore, the robot needs to find the musical possibilities based on the physical properties of the instrument to provide satisfying results. In this work, we introduce a drum robot with certain mechanical specifications and analyze the capabilities of the robot according to the drumming sound results of the robot. The robot has two degrees of freedom, actuated by one quasi direct-drive servo motor. The gripper of the robot features a flexible joint with passive springs which adds complexity to the movements of the drumstick. In a basic experiment, we have looked at the drum roll performance by the robot while changing a few control variables such as frequency and amplitude of the motion. Both single-stroke and double-stroke drum rolls can be performed by the robot by changing the control variables. The effect of the flexible gripper on the drumming results of the robot is the main focus of this study. Additionally, we have divided the control space according to the type of drum rolls. The results of this experiment lay the groundwork for developing an intelligent algorithm for the robot to learn musical patterns by interacting with the drum.