Optimization of dynamic simulations to identify movement patterns that simultaneously reduce the risk of injury and enhance human performance

Every movement, whether routine or sporting, achieves certain goals. Routine movements like walking takes us from one place to the other and sporting movements like hitting a volleyball help win the game. But each motion puts strain on certain joints of the body putting them at risk of injury. Walking can lead to chronic disorders like knee osteoarthritis over the years. Hitting a volleyball can put the shoulder at risk of a rotator cuff injury. The purpose of this work is to find optimal movement patterns that enhance human ability to achieve the goals of the movement, but at the same time, reduce the risk of injury while performing the movement.

Musculoskeletal modeling and simulation are powerful tools that allow researchers to generate dynamic simulations of human movement and answer questions like “what is the effect on certain biomechanical parameters if the movement changes in certain ways?” Additionally, the simulation environment can be combined with optimization methods to essentially check thousands of variations of the same overall movement and find the optimal one that both enhances human ability and reduces the risk of injury, something that cannot be achieved in experimental studies.

In this work, we used optimization methods combined with musculoskeletal modeling and simulations to find optimal whole-body, participant-specific movement patterns for landing, volleyball hitting, and walking that reduce the risk of anterior cruciate ligament injury, enhance hitting performance while reducing the risk of shoulder injury, and reduce the risk of progression of knee osteoarthritis, respectively. While doing so, a) guidelines were established that help the research community distinguish accurate simulations of dynamic simulations from errored ones, b) concepts from other disciplines like robotics were integrated to human simulation platforms to enhance dynamic consistency of human simulations, and c) the computational speed of the optimization process was reduced from days to minutes, making the process clinically relevant. Additional care was taken to ensure that the optimal whole-body movement patterns are participant-specific and not too different from the experimental data, ensuring that these new patterns can potentially be learned, enhancing the practical applicability of this work.