Multibody Dynamics and Predictive Simulation of Human Movements

点击翻译按钮获取中文摘要

A surgeon has a new idea for a procedure that may help an injured patient walk again. The procedure involves re-routing some tendons and ligaments. But how does the surgeon test out their idea before using it on a patient? How can an engineer design and test a new prosthesis or rehabilitation device, and have confidence in its performance and safety? Computer simulations can help answer these and other questions, but the simulations must be accurate and fast. Above all, the simulations should be able to predict human motions before treatments, without relying on post-treatment experimental data. The US Food and Drug Administration has recently embraced this idea and recommended the use of computer simulations to design new medical devices. However, there are no computer simulations that can quickly and accurately predict three-dimensional (3D) motions of a human, and the corresponding muscle, joint, and contact loads. Therefore, the goal of this research is to invent new methods for predictive "what-if" simulations of complex human motions, including 3D arm-reaching, walking, and running. To achieve this ambitious goal, the applicant will build upon his previous work in multibody biomechanics. Using "what-if" simulations, new designs for products and processes can be quickly evaluated; hence, multibody dynamics has been widely adopted by the robotics, aerospace, and automotive industries. However, recent attempts to extend multibody dynamics to the unique features of 3D human motions have encountered a number of challenges, including: human joints have complex geometries that cannot be represented by simple mechanical joints; human tissue has very nonlinear properties, especially during contact with hard surfaces; bone, joint, and muscle model parameters are difficult to obtain for individual subjects; the optimization criteria used by our brain and nervous system to coordinate our movements is not well understood. To develop fast and accurate predictive simulations in this research, new subject-specific models of muscles, joints, contact dynamics, and human motion controllers will be incorporated into the applicant's multibody models and simulations. Symbolic computing will be used so that the system equations can be viewed, shared, and automatically converted into highly-optimized simulation code. Symbolic derivatives will support sensitivity analyses and the identification of subject-specific model parameters, and will accelerate the convergence of the underlying optimization methods that drive the predictive dynamic simulations. Quick and reliable predictions of human motions and loads will fuel an explosive growth in biomechanics applications, including assistive and rehabilitation devices, personalized treatments, and wearable technologies. The requested funding will support 7 graduate students and 10 undergraduate researchers who will become the next generation of Canadian innovators in these exciting new areas.