PID-like IT2FLC-Based Autonomous Vehicle Control in Urban Areas

Reliable autonomous navigation requires a seamless framework that blends perception, localization, planning, and control. Therefore, this research sought to optimize the accuracy of the steering angle, braking, and throttle control as well as the precise localization of self-driving cars within the complicated urban environment. To provide reliable AV control, cutting-edge technology was used: a proportional–integral–derivative-like interval type 2 fuzzy logic controller (PID-like IT2FLC). This advanced controller improved the AV motion control stability, precision, and efficiency. Multiple technologies worked simultaneously to build perception, path planning, and localization. A minimal convolutional neural network (CNN) trained on red–green–blue (RGB) images precisely localized the vehicle’s position. The A* algorithm, essential for AV path-planning software, determined the optimal trajectories to navigate complex urban areas by avoiding obstructions and obeying traffic laws. Control performance improved by reducing errors using the sophisticated Car Learning to Act (CARLA) simulator for validation. You Only Look Once version 3 (YOLOv3) was 98.87% accurate for object perception in empirical tests. The simulation results confirmed the effectiveness of the suggested approach with mean squared error (MSE) values of 0.039, 0.0099, and 0.0047 to predict the position (x, y) and the orientation, respectively, based on the CNN. With an MSE of 0.0244 and 0.077 for the steering angle and speed, respectively, the simulation results showed that the suggested technique performed well under various weather conditions and when compared to prior research. Specifically, there was a 15.28% enhancement in the MSE for the steering angle and an impressive 88.15% enhancement for speed.