Karaağaç, Abdullah2025-04-222025-04-222025Karaağaç, A. (2025). A novel dynamic path planning method TD learning supported modified spatiotemporal GNN-LSTM model on large urban networks. Transportation : Planning - Policy - Research - Practice, 1–34. https://doi.org/10.1007/s11116-025-10600-100494488https://doi.org/10.1007/s11116-025-10600-1In this study, a new approach will be discussed in which routing is done by predicting future traffic and the learning algorithm is optimized during navigation. Traffic has a complex structure that is constantly changing. Especially for long-term travel, it is not an optimum approach to suggest a route only by considering the traffic situation at the time the navigation request is made. For this reason, the proposed algorithm recommends a route by taking into account future saturation conditions on the vehicle’s route. Singapore was chosen as the study area. The tests were carried out in a simulation environment. The four selected algorithms were tested spatially and temporally. Especially in long-term travels, the superior success of the proposed method compared to other selected methods has been demonstrated. © The Author(s) 2025.enGraph convolution networksLong-short term memoryModified Bellman–Ford algorithmNavigationPath planningSpatiotemporal graph neural networksTemporal difference learninginfo:eu-repo/semantics/openAccessWOS:00145618990000110.1007/s11116-025-10600-1Q1Q1