Optimization of Four-Way Controlled Intersections with Autonomous and Human-Driven Vehicles

Many car crashes occur at four-way stop sign intersections due to error-prone humans who misinterpret their surroundings. A solution is needed to prevent possible accidents and streamline traffic at four-way stop sign intersections. This paper proposes a software-based monitoring solution to eliminate this miscommunication that leads to accidents in a roadway setting where human-driven and autonomous cars are integrated. Relying on computer vision through an overhead camera and a convolutional neural network, the monitoring system is able to detect stopped vehicles at the intersection and direct traffic. Additionally, a self-driving car equipped with line-tracking was built along with a testing track to evaluate the efficacy of the monitoring software. Due to the project's limited resources and insufficient processing power, the software was unable to function as a real-time system. However, the system works with still images of various intersection scenarios with high accuracy.