隨著臺灣汽機車持有比率持續提升，都市幹道於每日的上下班尖峰時段總是充斥著車多壅塞的狀況，造成大量的時間成本與燃油消耗，並使得交通壅塞成為都市交通管理中的一個難題，因此如何有效地解決交通壅塞之問題，是交通工程師必須研究的課題。為了有效的緩解壅塞問題，優化號誌設計至關重要，目前多數路口的交通控制系統大多以定時號誌為主，而幹道連鎖號誌之時制計畫相關產製軟體多朝「最大化綠燈帶寬」或「最小化延滯」等兩大方向思考，以求最佳週期、時比、時差及時相順序等重要交通號誌控制之參數值，且應用之範圍可為獨立路口以及幹道系統等，國內之都市交通管理系統，大都皆已採用此類管理策略，以規範各路口號誌系統運作狀況。
近年來隨著人工智慧的發展，如何藉由人工智慧深度學習之類神經網路模式解決交通問題為當前之重要課題。因此，本研究將利用深度強化學習應用於幹道連鎖適應性號誌策略，並以高雄市與台南市交通最繁忙之幹道為例，透過車流模擬軟體SUMO評斷深度強化學習所產出之最佳獎勵動作與Synchro之延滯最小化之時制計畫針對不同車流情境下績效優劣比較，根據研究結果在順暢、車多、壅塞的情境下，深度強化學習能因應車流變化進行號誌最佳化的演算，所得到的交通績效皆優於Synchro時制，希望研究結果可以提供給未來發展人工智慧應用於都市幹道號誌連鎖設計之參考依據。

With growing ownership ratio of automobile and motorcycle in Taiwan, all cities, especially at peak periods, are facing one of the challenges in urban transportation management - heavy traffic congestion. Heavy traffic is time and fuel consuming, which is why most transportation engineers are committed to finding the solution to this problem. To relieve the traffic congestion, optimizing traffic-signal management is the most essential approach. Most traffic control systems are fixed-time, and time plans applied at isolated intersections and arterial roads, are proposing the best cycle, time split, offset and phase sequence by maximizing green band and minimizing delay to manipulate traffic system at every intersections in Taiwan.
As artificial intelligence developing, using artificial neuro network structure on smoothing traffic flow becomes a fundamental issue nowadays. This research would conduct deep reinforcement learning algorithm on adaptive arterial signal coordination strategy, using the highest-capacity roads in Kaohsiung and Tainan city for instance. This research would use the traffic flow simulation software, SUMO, to evaluate the performances of deep reinforcement learning with different time plans under different traffic flow, which are simulated by the signal simulation software, Synchro. The result of this research is expected to be the references for future AI applications on arterial signal coordination.

1、Arsava, T., Xie, Y., Gartner, N. H., & Mwakalonge, J. (2014). Arterial traffic signal coordination utilizlar traffic origin-destination information. 17th International IEEE Conference on Intelligent Transportation Systems (ITSC),
2、Christofa, E., Ampountolas, K., & Skabardonis, A. (2016). Arterial traffic signal optimization: A person-based approach. Transportation Research Part C: Emerging Technologies, 66, 27-47.
3、Genders, W., & Razavi, S. (2016). Using a deep reinforcement learning agent for traffic signal control. arXiv preprint arXiv:1611.01142.
4、Hao, W., Lin, Y., Cheng, Y., & Yang, X. (2018). Signal progression model for long arterial: intersection grouping and coordination. IEEE Access, 6, 30128-30136.
5、Jing, B., Lin, Y., Shou, Y., Lu, K., & Xu, J. (2020). Pband: A General Signal Progression Model With Phase Optimization Along Urban Arterial. IEEE Transactions on Intelligent Transportation Systems.
6、Lan, C.-L., & Chang, G.-L. (2016). Optimizing signals for arterials experiencing heavy mixed scooter-vehicle flows. Transportation Research Part C: Emerging Technologies, 72, 182-201.
7、Li, L., Lv, Y., & Wang, F.-Y. (2016). Traffic signal timing via deep reinforcement learning. IEEE/CAA Journal of Automatica Sinica, 3(3), 247-254.
8、Li, Z., Yu, H., Zhang, G., Dong, S., & Xu, C.-Z. (2021). Network-wide traffic signal control optimization using a multi-agent deep reinforcement learning. Transportation Research Part C: Emerging Technologies, 125, 103059.
9、Liang, X., Du, X., Wang, G., & Han, Z. (2019). A deep reinforcement learning network for traffic light cycle control. IEEE Transactions on Vehicular Technology, 68(2), 1243-1253.
10、 Liu, M., Deng, J., Xu, M., Zhang, X., & Wang, W. (2017). Cooperative deep reinforcement learning for tra ic signal control. Proc. 23rd ACM SIGKDD Conf. Knowl. Discovery Data Mining (KDD),
11、Rasheed, F., Yau, K.-L. A., Noor, R. M., Wu, C., & Low, Y.-C. (2020). Deep Reinforcement Learning for Traffic Signal Control: A Review. IEEE Access.
12、Stevanovic, A., Martin, P. T., & Stevanovic, J. (2007). VisSim-based genetic algorithm optimization of signal timings. Transportation Research Record, 2035(1), 59-68.
13、Sutton, R. S., & Barto, A. G. (2018). Reinforcement learning: An introduction. MIT press.
14、Van der Pol, E., & Oliehoek, F. A. (2016). Coordinated deep reinforcement learners for traffic light control. Proceedings of Learning, Inference and Control of Multi-Agent Systems (at NIPS 2016).
15、Watkins, C. J., & Dayan, P. (1992). Q-learning. Machine learning, 8(3-4), 279-292.
16、Wei, H., Chen, C., Zheng, G., Wu, K., Gayah, V., Xu, K., & Li, Z. (2019). Presslight: Learning max pressure control to coordinate traffic signals in arterial network. Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining,
17、Wu, X., Tian, Z., Hu, P., & Yuan, Z. (2012). Bandwidth optimization of coordinated arterials based on group partition method. Procedia-Social and Behavioral Sciences, 43, 232-244.
18、Zhang, L., Song, Z., Tang, X., & Wang, D. (2016). Signal coordination models for long arterials and grid networks. Transportation Research Part C: Emerging Technologies, 71, 215-230.
19、石家豪. (2001). 幹道號誌續進式時制設計模式之開發研究，碩士論文，國立成功大學.
20、何志宏. (2003). 交通工程人才培訓計畫 (2). 交通部運輸研究所.
21、吳沛儒. (2020). 應用AI技術進行交通數據蒐集暨號誌控制之研究. 交通部運輸研究所.
22、吳悅慈. (2011). 幹道群組適應性號誌控制模式之開發研究，碩士論文，國立成功大學.
23、卓訓榮、曾明德、周幼琳. (2018). 台灣號誌控制軟體（PaSO）示範驗證與推廣計畫. 國立交通大學.
24、林良泰、黃華宇、黃啟倡. (2012). 幹道系統延滯最小下續進路口數最大化模式之研究. 運輸學刊, 24(4), 529-554.
25、林良泰、謝長明、古新全. (2010). 高飽和下幹道號誌系統續進路口數最大化模式. 中國土木水利工程學刊, 22(3), 319-331.
26、胡大瀛、林婉婷、梁力元. (2014). 幹道號誌連鎖之減碳效益比較與分析：以小東路為例. 中華民國運輸學會103年,P.1587-1604.
27、陳一昌、張開國、張仲杰、黃惠隆、黃文鑑、張景平、翁忠川. (2007). 交通號誌時制重整計畫(1)－標準作業程序建立. 交通部運輸研究所.
28、陳麗雯, & 胡大瀛. (2020). 都市幹道連鎖時制設計之研究與實例分析. 土木水利, 47(4), 40-50.
29、黃秀雲. (2007). 都市路網號誌連鎖最佳化之研究，碩士論文，國立成功大學.
30、李卓育(2022). 深度強化學習法於交通號誌連鎖之應用，碩士論文，國立成功大學.