Wearable anatomical converter and real-time motion control of highly atriculated biomechatronic systems

Abstract

Motion capture has been a popular topic in robotics, animation, computer games, virtual reality and etc. On the other hand, application of humanoid robots in the entertainment field is a fast growing industry. Motivated by the idea of a virtual-reality game using entertainment robots, this project explores the systematic control of a robotic puppet through motion capture data. Dynamic models of the robotic marionette were developed from on Lagrangian equation. Feed-forward control and feedback control were studied based on the dynamic models with the illustration of block diagrams. A process of identifying the parameters appearing in the dynamic models was presented in details. The modeling provides better insight of the system, and builds a mathematical platform to research the control and simulation of the marionette system. Software was developed to capture human motions utilizing the bend-twist sensor system, and use the motion data to control the robotic marionette by either offline control method or online control method (real-time). During the software development, a kinematic model was proposed to map the human motion data into marionette motion data, and from that calculate how much the servo motors need to turn in order to achieve desired motions. Experimental results show that the motions of the marionette and the performer have good correspondence. This proves the proper of the proposed kinetic and dynamic models of the marionette system. The research has explored a great deal of further works for the motion capture-control system to become a sociable robot, which can communicate and interact intelligently with human beings.