Comparison of DQN and DDPG Learning Algorithm for Intelligent Traffic Signal Controller in Semarang Road Network Simulation

Congestion is one of the most common problems, particularly in large cities. The Adaptive Traffic Signal Controller is supposed to help by reducing intersection wait times. However, deciding which algorithm to use is difficult because there are so many options. In this study, we compare the performance of discrete-based and continuous-based algorithms in small lanes local intersections to better understand how they work. In five scenarios, the Deep Q-Network (DQN) and Deep Deterministic Policy Gradient (DDPG) algorithms are used with vehicle limits of 90, 150, 300, 600, and 900. The waiting time simulations of the two algorithms revealed that DDPG was 7.5% to 9.7% faster and more stable than DQN. © 2023 IEEE