Efficient algorithms for articulated branching mechanisms: Dynamic modeling, control, and simulation

With advances in computational and mechanical hardware technology, a growing body of interest has emerged in the area of real-time dynamic control and simulation of branching mechanisms---systems of tree-like topology such as a humanoid robot. Application of such interest requires fast algorithms to provide intuitive, systematic, and efficient ways to model, control, and simulate the dynamics of these complex systems.; This thesis presents efficient recursive algorithms, using the operational space formulation, to model and solve, at the task level, the dynamics problems of highly redundant articulated branching mechanisms with multiple operational points. A modified spatial notation is introduced to model complex robot kinematics and dynamics in an intuitive and systematic way. Using this notation, efficient recursive algorithms are developed for branching mechanisms in order to control and simulate task/posture behavior dynamics, closed-chain dynamics, and contact dynamics. A basic underlying framework is developed to provide dynamic control of intuitive task-level commands and posture behavior. The application of these algorithms results in a significant increase in the interactivity and usability of the dynamic control and simulation of complex branching mechanisms. The computational complexity of these algorithms is analyzed and compared with existing methods. Experimental and real-time dynamic simulation results are presented to demonstrate the effectiveness of the proposed algorithms.