Piano actively playing is a advanced sensorimotor endeavor involving eyesight, audio, and touch. Most prior tries require a specific style or guide ordering. A new paper on arXiv.org indicates a reinforcement discovering algorithm so that an agent could find out straight from a device-readable music score to participate in the piano.
The strategy exploits tactile sensors for manage and utilizes the corresponding notes created as a reward. The piano-actively playing endeavor is formulated as a Markov determination system. A multi-finger allegro hand equipped with tactile sensors and a compact piano keyboard are produced to conduct the endeavor.
The experimental success display that the strategy can prepare a robot hand to participate in the piano with correct notes, velocity, and fingering. The research also displays the possibility of eyesight-based tactile sensors to increase piano actively playing, especially on the fingering