This study compared the solutions and computational performance of three different static optimization algorithms used to predict muscle forces associated with elbow actuation during a simple throwing motion. The motion was simulated using an arm model with three degrees of freedom (shoulder pitch and roll, and elbow pitch). Elbow torque ranged from -1.36 to 3.07 Nm. Elbow torque and angle time histories were input into three independent static optimization routines, namely, constrained linear optimization (CLO), a genetic algorithm (GA), and linear least squares (LLS). These routines computed force values for the long and lateral heads of the triceps (TLg and TLt) and the biceps (B). For CLO and GA routines, different performance criteria were minimized, including sum of stress, sum of stress squared, sum of stress cubed, sum of force, sum of force squared, sum of force over maximum force, and sum of force over maximum force squared. Solutions found using CLO and LLS required far less computational time than GA. Muscle force predictions from CLO and GA were essentially identical, but there were substantial differences in muscle force predictions among the different optimization criteria utilized. In particular, the distribution and continuity of muscle force curves over the first half of the motion were found to be most affected by the power of the exponent in the optimization criterion involved.
展开▼