Exploring Creativity through Humanoids and Dance

Programming a humanoid robot to dance to live music is a complex task requiring contributions from multiple disciplines. A vocabulary of intricate limb motions must be designed to be stable and stylistically consistent. To produce coordinate these movements with music, the robot must be able to detect the appropriate beats within audio. One approach is a perceptual model of hearing, which can accurately determine beat locations, even for music without strong rhythmic content. The chosen implementation used an IIR feedback comb filter bank to determine tempo of an audio stream in real-time. A library of gestures was designed for a small humanoid robot to represent various dance motions. The software coordinates dance by choosing a random series of gestures, ensuring that a sequence won’t upset the robot’s balance. While playing the audio, the software initiates gestures that syncronize with detected beats. The final implementation allows user choreography and generative dance in a graphical user interface.