Approximating Steering Control for Level 3 Autonomous Driving on Highways Using Fuzzy-PID

Level 3 autonomous vehicles (AV) manage critical driving functions with minimal driver intervention, relying on advanced decision-making strategies for safe and efficient control. Among these, steering control is a key driving function for active lane keeping. Both inexact reasoning and learning-based approaches have been used to develop lane-keeping systems. While the rule-based systems are well suited for dynamic driving conditions, they often struggle in sharp curves or poor road visibility, leading to under- or over-steering in such scenarios. Existing fuzzy-PID lane-keeping approaches typically only consider lane offset and error rate [4], which limits their performance in curved highway segments. This work adds real-time curvature as a third input in the fuzzy inference system (FIS), leading to a 23.7 % reduction in lane deviation and enhanced damping response. The fuzzy-PID active lane-keeping system is implemented with the vehicle's real-time lane offset, its rate of change, and the current road curvature which is validated using the CARLA Simulator. The system relies on an adaptive cruise control module that manages the longitudinal control and a lane detection module that identifies lanes using computer vision. Overall, the proposed approach demonstrates the potential of inexact reasoning techniques in approximating expert-like decisions for steering control and the effectiveness of extending fuzzy lane-keeping systems with curvature awareness.