Robotic Traffic Cone System for Lane Closure using Computer Vision and Decentralized PID Control

The manual placement of traffic cones increases the risk of accidents for road workers due to vehicle collisions. This paper proposes a system of cone-shaped robots that automatically position themselves to gradually close a lane. Image segmentation and Hough transform, with a neural networks model to correct position distortion, are proposed to estimate the position of each robot. Additionally, to implement decentralized PID control, the current position and orientation is determined by a Kalman filter in conjunction with an embedded gyroscope and the motor's incremental encoders. Validation was conducted using three robots within a controlled, scaled-down environment.