In accident scenarios on freeways, traffic congestion, sharp declines in capacity, and the limitations of closed systems where vehicles cannot turn around or exit freely often pose serious challenges. To address these issues, this study develops an improved Markov Decision Process (MDP) framework for dynamic emergency lane opening. Compared with traditional MDP-based traffic control models, the proposed method integrates three enhancements: Firstly, an explicit action decision space transition mechanism that couples variable speed limits with emergency lane opening decisions; Secondly, vehicle-type–differentiated actions to support fine-grained and adaptive opening strategies; and a redesigned reward function incorporating congestion cost, action cost, rewards earned and penalty received to ensure decision rationality and operational feasibility. Based on this improved structure, the optimal strategy is derived using the value iteration algorithm. Finally, simulation results using SUMO demonstrate that under the innermost-lane accident scenario, the optimal strategy is to first implement “variable speed limits”, followed by “opening the emergency lane to all vehicles”. This approach increases average speed by 10.81% but enlarges the headway between vehicles at the detection cross-section, leading to an 8.79% decrease in vehicle throughput. Under the outermost-lane accident scenario, the optimal strategy is to first implement “variable speed limits”, followed by “opening the emergency lane to HOVs”, which increases average speed by 5.47% while reducing vehicle throughput by 3.13%. These results validate the effectiveness of the proposed model.
Yao et al. (2026) studied this question.