Optimization-based dynamic human gait prediction

A new methodology, called predictive dynamics, is introduced in this work to simulate human walking using a spatial digital human model. The digital human model has 55 degrees of freedom. The resultant action of all the muscles at a joint is lumped and represented by the torque at each degree of freedom. In addition, the cubic B-spline interpolation is used for time discretization and the well- established robotic formulation of the Denavit- Hartenberg (DH) method is used for kinematics analysis of the mechanical system. The recursive Lagrangian formulation is used to develop the equations of motion. The ground reaction forces (GRF) are obtained from a novel two-step active- passive algorithm. The problem is formulated as a nonlinear optimization problem. A unique feature of the formulation is that the equations of motion are not integrated explicitly, but evaluated by inverse dynamics in the optimization process. Besides normal walking, several other cases are also considered, such as walking with a shoulder backpack of varying loads, walking at different speed, walking with asymmetric step lengths, and walking with reduced torque limits.