Application of Artificial Neural Network for Successful Prediction of Lower Limb Dynamics and Improvement in the Mathematical Representation of Knee Dynamics in Human Locomotion

The motivating factor behind this research is the significance and demand for developing relatively affordable yet effective assistive technologies for disabled people. In the presented work, an intelligent model that predicts the lower limb joint angles for an entire gait cycle and a model that modifies the constant terms of the average value-based modeling representation of the knee dynamics are both developed through the application of artificial neural networks (ANN). Hip and knee joint angles were predicted using a model with ground reaction force (GRF) and joint angle being input parameters. The coefficients of correlation and determination values of the model were found to be close to the ideal value, while the mean square error value was determined to be within the tolerance limit. In another developed model, the linear displacement of a human gait cycle was predicted based on the inputs like angular displacement, velocity, and acceleration of the hip and knee joints. The average value-based modeling representation of the knee dynamics was more accurately represented after obtaining the modified values of the constant terms (C₀, C₁, C₂ ). The resulting models can be used to design and develop assistive technologies for physically disabled people, thereby enabling their reintegration into society and helping them to lead normal lives.