Characterizing, modeling, and predicting the external ground reaction forces of legged movement

In this thesis, I explore legged agility with a primary focus on controlling external ground reaction forces produced by our legs. I conduct empirical studies and develop models to understand how we control external forces and develop innovative means to measure them. My research consists of four aims, each contributing to our understanding of human performance and driving potential advancements in sports, robotics, and rehabilitative technologies. In Aim 1, I characterize the control of leg external forces (n=14). To achieve this, I construct a mechanical system and a real-time visual feedback system to capture force magnitudes and positions exerted by my leg. Using system identification, I gain insights into the control of leg external forces across different magnitudes and positions. In Aim 2, I examine the effects of neuromuscular fatigue on our nervous system's capacity to control leg external forces (n=18). I hypothesize that heightened fatigue results in a decrease in both the responsiveness and accuracy of leg force control. My results reveal a significant reduction in mean maximum force production, leading to a substantial decline in my leg force control responsiveness. These findings enhance our understanding of how fatigue influences agility and may guide strategies to sustain performance in the presence of fatigue. In Aim 3, I set out to understand the limit of vertical jumping by studying the external forces generated during jumping (n=10). I develop physics-based models of varying complexity to predict external forces during vertical jumps and identify the simplest model that accurately predicts human-like forces. This model, capable of simulating jumps from different depths, highlights the significance of force-velocity properties and maximal force as limiting factors for jump height. In Aim 4, I develop a novel approach to estimate the external forces generated by each leg during vertical jumps. Using a transformer-based neural network and video data (n=30), I demonstrate that the model accurately predicts each leg's external forces, offering a new tool for measuring jump height and forces from video. My work aims to make biomechanical analysis accessible, a task typically confined to laboratory settings. In summary, this thesis investigates the control of leg external forces, the effects of fatigue, and the development of predictive models. It underscores the potential of machine learning in biomechanical analysis, contributing to a broader understanding of human performance and paving the way for new technological advancements