Modelling and Control of Ankle Foot Orthosis (AFO) for Children Utilising Soft Computing Towards Intelligent Approach
Ankle foot orthosis (AFO) is frequently utilised to offer practical help for patients with lower limb injuries or defects. However, ground response force modelling has historically proven difficult, especially when the robot is modelled with a high degree of freedom and has numerous points of contac...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2024
|
| Subjects: | |
| Online Access: | http://ir.unimas.my/id/eprint/46385/1/19020156_Thesis_3.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Ankle foot orthosis (AFO) is frequently utilised to offer practical help for patients with lower limb injuries or defects. However, ground response force modelling has historically proven difficult, especially when the robot is modelled with a high degree of freedom and has numerous points of contact with the ground. Therefore, this research presents the development of a dynamic system and controller of AFO. In order to accurately reflect the system dynamics, a prototype AFO model for kids was created and constructed. Using data immediately obtained from the experimental setup, the system's dynamic behaviour was simulated. The research uses both non-parametric and parametric techniques, namely Multilayer Perceptron Neural Network (MLP NN) and Particle Swarm Optimisation (PSO), for system modelling. The modelling stage comprised breaking down the dynamic AFO models into single-input single-output models. Validation methods such as the meansquared error (MSE) and correlation tests were utilized to verify the performance of the models. The results suggested that MLP NN excelled over PSO in recognizing the AFO dynamic system by utilizing the model structure of delay 8, with unbiased results whereby the confidence level fell within the 95% range and showed the lowest mean squared prediction error is 1.1034 x 10-5 . In comparison, it can be concluded that MLP NN provided a more accurate estimation of the AFO dynamic model. In controllers’ development, the research employed both conventional and intelligent Proportional–Integral–Derivative (PID) based controllers. For the conventional controller, Proportional–Integral–Derivative Ziegler-Nichols (PID-ZN) was tuned by using the mathematical calculation based on the identified models obtained earlier. Meanwhile, for the intelligent controller, Proportional– Integral–Derivative Particle Swarm Optimisation (PID-PSO) was utilized to find the optimum value of PID controller parameters to track the desired position during the ankle training. The PID controllers were tuned automatically by the PSO algorithm based on the same identified models. The performance of both developed controllers was compared and analyzed. The system's conduct in tracking the position of the AFO along its trajectory was observed. Both control approaches' comparative evaluation was provided and discussed. Based on the results, it is evident that PID-PSO performs superiorly due to its lower steadystate error and shorter settling time when compared to PID-ZN. This comparison indicates that the intelligent controller, PID-PSO showed a significant improvement as compared to the conventional controller, PID-ZN. It was found that the percentage of improvement achieved by the intelligent controller over the conventional controller for settling time is 62.6 % and the steady state error is improved by 75.56 %. |
|---|