چكيده به لاتين
The non-stop growth of urban traffic, especially on highways, has led to traffic jams and, as a result, long queues, increased travel time, increased chances of accidents, and environmental pollution. Since the extensive development of infrastructures cannot be a permanent solution to solve such problems, the existing facilities should be used optimally and appropriately. Several methods, such as ramp metering control, variable speed limit, lane change control, tolls, etc., have been studied and investigated to control highways. One of the approaches to improve the performance of these controls is their simultaneous and coordinated integration and implementation. In this thesis, we will investigate the effect of the combination of three concurrent controls of ramp metering, variable speed limit, and highway tolls on the traffic conditions of the highway network. For this purpose, the reinforcement learning method has been used to manage the control values as variables. Traffic data is first predicted by predictive models based on neural networks and then given to the simulation of Sumo software. The controller model then selects the ramp rate, speed limit, and toll values based on the simulation outputs and applies them to the network. The results of implementing the proposed model show the significant impact of the model in improving traffic parameters such as average travel time, the average speed of cars, waiting time, queue length, average network density, and the number of vehicles arriving at their destination. The model reduced the average travel time by 11.5% compared to the condition without controls, which is more than the effect of implementing single controls or a double combination of the mentioned methods. In addition, in the event of an accident in the network, the proposed management can significantly improve the network conditions compared to the situation without control.
