A Smart Traffic Controlling System Using Auxiliary Sensors with Deep Learning and Its Simulation

The traffic jam is a severe problem in modern cities, and many researchers tried to reduce the pain by optimizing the traffic system and traffic participants. This paper introduces a hardware system to optimize the time management of the crossings to improve the overall efficiency of the city's transportation. After installing this system, redundant seconds will be eliminated, and more time will be given to pathways that are in congestion. We separate our research into two parts. The aim of conducting the auxiliary sensors of light gates is to measure the accurate speed of vehicles passing by and counting the cars passing by for one lane for the calibration of the camera identification, in the form of model simulation. The detector algorithm uses Faster-RCNN as the base with real-time image processing to detect the existence of cars. The algorithm provides the numerical results of the amounts of vehicles. And the auxiliary light gate measures accurately the speed of the vehicles. The method is sufficient for a traffic lighting system with also elementary principles that reduces the cost of massively producing the system integrated chips. With the implementation of this system, the efficiency of a city will be promoted. Finally, we use the simulation method to get the total time of the traffic flow with and without modifying the time interval and periods of green lights, and the result shows the modification has a positive effect of reducing the total passing time of the cars.