The quest for dynamic consistency: a comparison of OpenSim tools for residual reduction in simulations of human running.

Abstract

Using synchronous kinematic and kinetic data in simulations of human running typically leads to dynamic inconsistencies. Minimizing residual forces and moments is subsequently important to ensure plausible model outputs. A variety of approaches suitable for residual reduction are available in OpenSim; however, a detailed comparison is yet to be conducted. This study compared OpenSim tools applicable for residual reduction in simulations of human running. Multiple approaches (i.e. Residual Reduction Algorithm, MocoTrack, AddBiomechanics) designed to reduce residual forces and moments were examined using an existing dataset of treadmill running at 5.0 ms-1. The computational time, residual forces and moments, and joint kinematics and kinetics from each approach were compared. A computational cost to residual reduction trade-off was identified, where lower residuals were achieved using approaches with longer computational times. The AddBiomechanics and MocoTrack approaches produced variable lower and upper body kinematics, respectively, versus the remaining approaches. Joint kinetics were similar between approaches; however, MocoTrack generated noisier upper limb joint torque signals. MocoTrack was the best-performing approach for reducing residuals to near-zero levels, at the cost of longer computational times. This study provides OpenSim users with evidence to inform decision-making at the residual reduction step of their workflow.