Machine Learning Technique based Optimized Fractional Order PID controller of Brushless DC motor in Electric Delta Type Recumbent Tricycles Applications

This paper presents a machine learning–optimized Fractional Order Proportional–Integral–Derivative (ML-FOPID) controller for speed control of Brushless DC (BLDC) motors employed in electric delta-type recumbent tricycles. Conventional PID and metaheuristic-tuned FOPID controllers rely on fixed or offline parameter tuning, which limits their effectiveness under nonlinear motor dynamics, load variations, and territory disturbances commonly encountered in electric locomotive applications. To overcome these limitations, a hybrid learning model combining supervised neural networks and reinforcement learning is proposed to enable online adaptation of the FOPID controller parameters. A mathematical model of the BLDC motor drive system is developed, and controller tuning is formulated using standard time-domain performance indices. The proposed ML-FOPID controller is evaluated through extensive MATLAB/Simulink 2024b simulations, followed by experimental validation on a BLDC motor–tricycle test setup. Simulation results showed that the proposed controller achieves 20% reduction in overshoot, approximately 15% faster settling time, and a 12–18% reduction in power consumption comparing with conventional PID and optimization-based FOPID controllers. The experimental results confirm the improved dynamic response, robustness to load disturbances, and stable real-time operation of the proposed control strategy. The results indicate that machine learning–based adaptive fractional-order control offers a practical and effective solution for enhancing energy efficiency, ride comfort, and reliability in electric recumbent tricycle applications.