The Traffic Overflow Dissipation Prediction and Optimization Control Method for Short-Linked Intersections Based on LSTM

In view of the fact that short-linked intersections are highly prone to traffic overflow, and the overflow is immediately controlled and diverted, while ignoring the issue that overflow can dissipate on its own, this paper proposes a traffic overflow dissipation prediction and optimization control method for short-linked intersections based on LSTM. The method analyzes traffic overflow and dissipation at short-linked intersections based on traffic wave theory, and constructs a traffic overflow dissipation prediction model using LSTM. When the model predicts that an irreversible traffic overflow will occur in a future time period, the proposed traffic overflow optimization control algorithm is used to adjust the signal timing plan. Finally, VISSIM traffic simulation is employed to verify the effectiveness of the proposed method. The results show that the traffic overflow dissipation prediction model proposed in this paper achieves an accuracy of 95% for traffic overflow identification and 92.68% for dissipation prediction, demonstrating good prediction accuracy. Based on this, the proposed traffic overflow optimization control algorithm can effectively facilitate traffic overflow dissipation. Compared to the current plan and real-time overflow control, the maximum queue length is reduced by 19.41% and 7.02%, respectively, while the average queue length at upstream intersections is also smaller, indicating that the proposed method can optimize and control traffic overflow while ensuring the traffic capacity of short-linked intersections.