Decision-focused learning (DFL) offers an end-to-end approach to the predict-then-optimize (PO) framework by training predictive models directly on decision loss (DL), enhancing decision-making performance within PO contexts. However, the implementation of DFL poses distinct challenges. Primarily, DL can result in deviation from the physical significance of the predictions under limited data. Additionally, some predictive models are non-differentiable or black-box, which cannot be adjusted using gradient-based methods. To tackle the above challenges, we propose a novel framework, Decision-Focused Fine-tuning (DFF), which embeds the DFL module into the PO pipeline via a novel bias correction module. DFF is formulated as a constrained optimization problem that maintains the proximity of the DL-enhanced model to the original predictive model within a defined trust region. We theoretically prove that DFF strictly confines prediction bias within a predetermined upper bound, even with limited datasets, thereby substantially reducing prediction shifts caused by DL under limited data. Furthermore, the bias correction module can be integrated into diverse predictive models, enhancing adaptability to a broad range of PO tasks. Extensive evaluations on synthetic and real-world datasets, including network flow, portfolio optimization, and resource allocation problems with different predictive models, demonstrate that DFF not only improves decision performance but also adheres to fine-tuning constraints, showcasing robust adaptability across various scenarios.

Arterial traffic signal coordination makes traffic flow more efficient and safer. This paper presents a partition-enabled multi-mode band (PM-BAND) model that is designed to solve the signal coordination problem for arterials with multiple modes, i.e., passenger cars and transit vehicles. The proposed method permits the progression bands to be broken if necessary and optimizes system partition and signal coordination in one unified framework. The impacts of traffic demand of passenger cars and transit vehicles as well as the geometry characteristics of the arterials are taken into account. Signal timings and waiting time of transit vehicles at stations are optimized simultaneously. The PM-BAND model is formulated as a mixed-integer linear program, which can be solved by the standard branch-and-bound technique. Numerical example results have demonstrated that the PM-BAND model can significantly reduce the average number of stops and delay compared with the other models, i.e., MAXBAND and MULTIBAND. Moreover, the progression bands generated by the PM-BAND model have a higher reliability and effectiveness.

This paper presents an advanced transit signal priority (ATSP) control model that considers bus progression at downstream intersections when giving priority at upstream intersections and stochastic bus arrival times. The ATSP control model is applicable to arterials with bus lanes. At the center of the ATSP control model is a stochastic programming model formulated to find the optimum priority strategies at each intersection of interest, which minimize bus delays while having the least negative impact on general traffic. The uncertainty in bus arrivals is taken into account by considering stochastic bus dwell times. The ATSP control is implemented in a traffic micro-simulation environment and compared with conventional transit signal priority (CTSP) control. Extensive simulation experiments are conducted with different traffic congestion levels, bus headway levels, and dwell-time distributions. Results show that the ATSP control generates an additional reduction in bus delay of around 10% when compared to the CTSP control.

Conflicting transit signal priority (TSP) requests can significantly influence the efficiency of TSP. Determining the best serve sequence for TSP requests is important for both the transit and other vehicles. The majority of previous studies focused on resolving conflicting TSP requests at the intersection level. For urban arterials, few studies take coordination into consideration, so the benefits of TSP may be lost due to the lack of progression. In view of this, this paper proposes a bi-level optimization model to resolve the problem of conflicting TSP requests at arterial corridors. The upper level of the model is progression control and the lower level is intersection control. At the progression control level, by considering the characteristics of bus operation, this paper proposes a signal coordination method that generates a green band between the two adjacent bus stops (stop-to-stop). At the intersection control level, a linear programming model is developed to resolve conflicting TSP requests. Then, this paper uses a hybrid genetic algorithm to solve the proposed model. Finally, an arterial with seven intersections in Nanjing, China, is selected to evaluate the performance of the proposed model. Three other models (the baseline model without TSP, Model 1 with the regular TSP and the classical coordination, and Model 2 with the lower model and the classical coordination) are also utilized to conduct the comparison analysis. The results show the proposed model can significantly reduce the bus passenger delay and stops compared to the three other models and meanwhile the operation of other vehicles will not turn worse.

Aimed at similarity between neural network ensemble and simple neural network, a novel gradient learning algorithm for ensemble (GLAENN) is presented based on the gradient descent method. The new algorithm can improve the generalization error by modifying subnet weights after the ensemble subnets are trained individually. The simulation results indicate GLAENN is of similar function to GASEN but with a different idea; further, it is of better generalization performance than GASEN, bagging and single network