|
چکیده
|
This paper presents a structured hybrid control framework for mobile robot navigation and path tracking in uncertain and dynamic environments. The proposed architecture integrates a fuzzy logic controller as the primary decision-making layer with a multilayer perception (MLP) neural network used for nonlinear compensation. The fuzzy module processes real-time sensory inputs from front, left, and right distance measurements and generates the initial steering command using a Mamdani inference system. To enhance tracking accuracy and robustness, the neural network is trained in a supervised manner using the mean squared error (MSE) loss function to learn adaptive correction signals. The final steering command is obtained by combining the fuzzy output with the neural compensation term, forming a hybrid control strategy that preserves interpretability while improving adaptability. The system is implemented in a simulation environment and evaluated under obstacle-rich scenarios. Quantitative performance metrics, including MSE and root mean square error (RMSE), are used to assess tracking accuracy. Experimental results demonstrate improved trajectory precision, reduced tracking error, and enhanced stability compared to standalone fuzzy and neural approaches. The proposed framework maintains clear functional separation between decision-making and compensation layers, improving modularity, transparency, and reproducibility. The results confirm that the integration of fuzzy reasoning with neural compensation provides an effective and computationally efficient solution for autonomous mo-bile robot navigation in nonlinear and uncertain environments.
|