An Intelligent Connected Vehicle Path Tracking Control Based on Pure Pursuit Control with Particle Swarm Optimization Algorithm

In recent years, autonomous driving developed to cope with expanded domain to improve perception, decision and control over path tracking. The sensor fusion relies on real-time data in various sensors such as cameras, lidar, Global Positioning System (GPS) and radar to make steering adjustments and detect obstacles. However, due to complex driving environment factors such as road curvature, unpredictable obstacles, varying speed and weather conditions which constantly influence vehicle behavior. Due to nonlinear vehicle dynamics impacting tracking errors and lead to errors in localization for multi-sensor data. Therefore, this paper proposes Particle Swarm Optimization- Pure Pursuit Control (PSO-PPC) for vehicle path tracking control. This involves image augmentation techniques to increase sensor data, bounding boxes for spatial tracking through Kalman filter. After that, for feature extraction, Scale-Invariant Feature Transform (SIFT) technique is used to local image operator which takes image and collected local feature vector in object recognition. Finally, path tracking average distance between vehicle position and target path is defined by path deviation. Therefore, particle represents potential solution for selection choices or for specific parameters in tracking. The proposed PSO-PPC for vehicle path tracking achieved better results in terms of accuracy ${9 5. 7 2 \%}$, precision ${9 0. 8 7 \%}$ and F1-Score of ${9 8. 7 6 \%}$ when compared with existing Faster-RCNN.