DESIGN OF PERIMETER CONTROL IN CONTROLLED URBAN TRAFFIC NETWORK WITH MAX-PRESSURE ALGORITHM

• Main Author: Anugerah Kenan Sergio, Mesa
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/83733
• Institution: Institut Teknologi Bandung

The increase in population in urban areas has resulted in various problems, one of which is traffic congestion. To overcome this problem, it is necessary to create adaptive traffic control. In previous research, traffic control using Reinforcement Learning has been made, namely using Cooperative Double Q-learning. Then traffic control using Deep Q-Network was also made, which then improved its performance with the Particle Swarm Optimization optimization algorithm. However, the control system is still not effective because there is still a relatively high vehicle density value. In this research, a perimeter control algorithm is implemented on the traffic network using the Max-Pressure algorithm. This algorithm is applied to the Eclipse SUMO traffic simulator. The best parameters for the Max-Pressure algorithm have been found and the values of K_p and K_i for perimeter control have been tuned. After evaluated using Macroscopic Fundamental Diagram, the perimeter control algorithm with a value of K_p=20 and a value of K_i=5 significantly reduced the maximum density of the network from 40 vehicles/km to a maximum density of 23 vehicles/km. The decrease in vehicle density resulted in reduced congestion so that the traffic network could be more optimized. Keywords: traffic control, perimeter control, max-pressure algorithm, vehicle density