Uncertainty and Sensitivity in Human Motion Dynamics Simulations

• Main Author: Wojtusch, Janis
• Format: Others
• Language: en
• Published: 2018
• Online Access: https://tuprints.ulb.tu-darmstadt.de/6590/1/Dissertation-Wojtusch-2018.pdf; Wojtusch, Janis <http://tuprints.ulb.tu-darmstadt.de/view/person/Wojtusch=3AJanis=3A=3A.html> (2018): Uncertainty and Sensitivity in Human Motion Dynamics Simulations.Darmstadt, Technische Universität, [Ph.D. Thesis]

Biomechanical dynamics simulations facilitate the investigation of fundamental principles and concepts in human motions. The simulation results help to explain experimentally observed phenomena and reveal underlying mechanisms. Due to unavoidable restrictions in biomechanical measurements and the determination of personalized model parameters, the simulation results always lie within a specific range of possible solutions. Since these uncertainties can have a significant influence on derived scientific conclusions and clinical decisions, this thesis provides a systematic uncertainty and sensitivity analysis of the common inverse dynamics simulation to assess uncertainty propagation and the contribution of individual uncertainty sources in the estimation of joint torques resulting from particular human motions. The analysis evaluates uncertainties and sensitivities in selected joint torque estimates of the lower limbs for three motion tasks performed by a female and male subject. It follows the procedure of the inverse dynamics simulation including the acquisition of biomechanical measurements, estimation of model parameters and realization of dynamics simulations with a parallel assessment of uncertainty propagation and apportionment. This approach ensures a systematic and consecutive evaluation of uncertainty and sensitivity with respect to the sequential nature of the procedure and existing dependencies between the involved uncertainty sources. The human locomotor system is modeled as a three-dimensional multibody system implemented within an efficient multibody systems library. The analysis employs a global method based on Monte Carlo simulations in combination with a quasi-random sampling strategy in order to explore the whole input space and consider nonlinearities in the biomechanical model. Potential correlations among the uncertain simulation inputs and model parameters are considered. The biomechanical measurements that form the basis for the uncertainty and sensitivity analysis comprise motion capture data and force plate measurements with an instrumented treadmill for the considered motion tasks. Associated uncertainties caused by variations in anatomic landmark identification, soft tissue artifacts, motion capture and force plate measurements are quantified and modeled by experimental investigations with the actual subjects and measurements systems or by suitable computational models described in literature. The variances in parameters of two comprehensive regression models for joint center estimation and one widely used regression model for anthropometric parameter estimation in female and male subjects are assessed and modeled based on the found uncertainties for biomechanical measurements as well as additional statistical properties from literature. For the anthropometric parameter estimation, a revised set of regression parameters is derived from the obtained results. The impact of the identified uncertainties on estimated anthropometric parameters is investigated in an exemplary sensitivity analysis. With incorporating the previously determined models for the individual uncertainty sources, the uncertainty propagation and apportionment in the actual inverse dynamics simulation are evaluated and discussed for a walking, running and kicking a ball motion performed by both subjects. The identified uncertainties at the individual levels of the inverse dynamics simulation allow to evaluate the credibility and accuracy in this and similar biomechanical simulations, while the corresponding sensitivities identify uncertainty sources with particularly high influence on the simulation results. These results give an indication of the expectable validity in biomechanical dynamics simulations, but also allow to enhance the quality of biomechanical studies by specifically approaching the identified problems. In addition to the uncertainty and sensitivity analysis, complementary research topics regarding a user-centered design methodology for active prosthetic and orthotic devices based on biomechanical simulations of human motions are investigated and presented. A particular focus is put on the analysis and implementation of serial elastic actuator concepts that ensure inherent safety and energy efficiency as well as the development of a personalized audio-visual simulation with respect to relevant psychological factors.