Curvature-Continuous Trajectory Refinement for Autonomous Vehicles Using a Pure Pursuit Based Path Smoothing Framework

Path planning is a key element of autonomous vehicle navigation, allowing vehicles to calculate feasible paths in challenging environments. Algorithms such as Hybrid A*, Reeds-Shepp, and Dubins paths effectively generate collision-free paths but tend to create curvature discontinuities. These discontinuities result in sudden steering transitions, which create control instabilities, higher mechanical stress, and lower passenger comfort. To overcome these issues, this paper suggests a path-smoothing technique based on the pure-pursuit algorithm to produce curvature-continuous, distortion-free paths appropriate for real-world driving. This method utilizes the practical approach of the original path but removes sudden transitions that destabilize control. By ensuring smooth curvature, the vehicle undergoes fewer jerky steering actions, enhanced energy efficiency, less actuator wear, and improved high-speed tracking. Experimental verifications, such as real-time tests on autonomous vehicles, show that the proposed approach greatly enhances steering smoothness and control performance over conventional path planning results. This paper provides a valuable approach to usual limitations of discrete path planning, on the contribution of control algorithms such as pure-pursuit to bridging the gap between planning and execution towards more adaptable autonomous driving.