Modelling and AI based Control of upper body Ball and Socket joints for Swimming in Sports Biomechanics

Abstract

In the swimming sports biomechanics, the vulnerability of the shoulder joint has been introduced as a critical area of concern, where understanding and mitigating injury factors play a crucial part in swimming. The intricate movements involved in swimming, particularly the front crawl stroke, place considerable stress on the upper limb, making the shoulder susceptible to injuries. This study deals with the biomechanics of ball and socket joint and intricacies of shoulder vulnerability during front crawl swimming to reduce chance of injury. The focus of this study lies in the synthesis of sports biomechanics for examination of arm movement in swimming by employing a 3-degree-of-freedom (3-DOF) biomechanical model containing the shoulder, elbow, and wrist joints using D’Albert’s principle, the study delves into the mechanical intricacies that make the shoulder particularly prone to stress and potential injury in the aquatic environment. The research extends the traditional biomechanical analyses by incorporating artifcial intelligence techniques such as Decision Trees (DT) and Random Forests (RF) to predict to predict shoulder angles from elbow angles, offering insights into the dynamic relationships between joints during swimming. The dataset is generated synthetically from mathamically equation which showed resemblance with the real experimental data as shown in previous study. Furthermore, the application of advanced control theory, with Linear Quadratic Regulator (LQR) and Proportional-Integral-Derivative (PID) controllers are used to track the shoulder joint, adds a practical dimension to the study, exploring optimal joint trajectories and their implications for injury prevention. The AI techniques showed goods results in predicting the shoulder joint angle however DT showed slightly with RMSE value of 0.27 than RF which is 0.33 for 10 trees better results for this particular application due to its nature to deal with simple dataset than RF which is switable to deal with complex datset. In tracking the the shoulder angle trajectory both LQR and PID showed better results but LQR controller showed better results than PID controller due to its ability to optimize control actions based on a mathematical model, providing superior performance in handling complex and dynamic systems with precise state feedback and optimal control strategies, resulting in improved stability and responsiveness in tracking tasks compared to the PID controllers. The shoulder joint showed more torque in front crawl swimming than other joints showing that shoulder joint generate more propulsive force than elbow and wrist joint and if not managed correctly is at risk of injury