Human musculoskeletal models have revealed the general patterns of muscle recruitment during daily activities. Nonetheless, the consistency
of dynamics calculations from different musculoskeletal simulation packages is not well understood. The objective of this study
was to understand the effect of the simulation solver and simulation model on the musculoskeletal simulation results using lower limb
models in OpenSim and AnyBody modeling system. Matched musculoskeletal model and generic model in both systems were simulated
using the external forces and joint kinematics measured at the Fourth and Sixth Grand Challenge Competitions to Predict In-Vivo Knee
Loads. Muscle activation levels in lower limb were compared between the packages, and against the electromyography signals from the
aforementioned competitions. The muscle activation levels were very similar between the two packages when matched models were
simulated, indicating high consistency between the solvers. In the generic models, the root mean square (RMS) difference in the muscle
activation levels was high at 0.15 and 0.19 for ideal force generator muscles and modified Hill-type muscles, respectively. The RMS and
phase differences were high between the muscle activations and electromyography signals. Comparisons will help understand the similarities
and differences between the musculoskeletal simulation packages and the effects of the model differences on simulation results.