Comparative analysis of fuzzy logic and ANFIS controllers using a novel robotic ankle-foot orthosis

Background Foot drop patients depend on ankle-foot orthosis (AFOs) for locomotion; however, conventional AFOs often lack adequate ankle range of motion (ROM) due to their rigidity. Objective This study aims to develop a robotic ankle-foot orthosis (RAFO) equipped with a hybrid actuator (HA) and evaluate its ROM through a bench test method using a fuzzy logic controller (FLC) and an adaptive neuro-fuzzy inference system (ANFIS). Methods A commercially available AFO was adopted, and its dimensions were used to develop an RAFO with HA using SolidWorks. The FLC was considered for this analysis due to its ability to handle uncertainty and imprecision models effectively. The FLC’s input and output membership function variables are defined based on preliminary experimental readings of the Plantarflexion/Dorsiflexion (PF/DF) motion and actuator displacements. The Mamdani rule-based FLC model was developed using thirty-six rules to control the RAFO motion. The FLC dataset was then implemented to train the ANFIS controller for 1000 epochs using the Takagi-Sugeno model with a hybrid learning approach. ANFIS, with its neural network-based learning capability, was implemented to achieve higher uniform ROM. Results The FLC generated a non-uniform ROM of 29.7° (13.9° PF, 15.8° DF) owing to rule-based input. In contrast, the ANFIS achieved a uniform and higher ROM of 30.4° (16.8° PF, 13.6° DF) by integrating neural network learning with fuzzy logic reasoning. Conclusion The RAFO with an ANFIS controller provided a 2.3% higher uniform ROM than the FLC and also adapts for varying walking conditions. This HA-based assistive device can improve rehabilitation outcomes and mobility of patients with foot drop.