Abstract
Intensive interactions among vehicles at freeway on-ramp merging areas frequently cause congestion and accidents. The collaboration of connected automated vehicles (CAVs) is promising to effectively coordinate these conflicts. However, CAV-based control encounters significant challenges in real-time optimization of vehicle scheduling and trajectory planning, especially in scenarios involving multiple lanes and a large number of vehicles. To tackle these challenges, this dissertationmathematically models the freeway merging problem and develops three algorithms to solve the problem.
The first work proposes a mixed integer nonlinear programming (MINLP) model for the cooperative merging of two traffic streams at a single-mainline freeway on-ramp merging section. The proposed model simultaneously optimizes multiple vehicles’ trajectories and their merging sequence to improve traffic efficiency and ensure safety. Unlike conventional treatments, which match one mainline facilitating vehicle with one merging vehicle, the proposed model determines the optimal number of facilitating vehicles and which mainline vehicles should serve as the facilitating vehicles to cooperatively minimize disruption from ramps. The safety and feasibility of the planned vehicle trajectories are guaranteed at any time. To solve the model rapidly, we propose a solution algorithm that incorporates an iterative linear programming method into a novel search process based on a necessary
condition for optimality that we identify and prove. The algorithm is highly efficient because it enjoys a significantly reduced search space. The proposed approach, consisting of the MINLP model and the solution algorithm, is evaluated under different traffic demands and mainline-ramp demand ratios and real vehicle arrival patterns from the NGSIM dataset. The performance of the proposed method outperforms benchmark CAV control algorithms, and the computational efficiency is promising for realtime automated merging tasks.
The second work considers a multi-lane freeway on-ramp merging section, focusing on the simultaneous decision-making of lane changes, vehicle sequences, and trajectories. To this end, an integrated MINLP optimization model is proposed to jointly optimize lane change decisions, vehicle sequences, and vehicle trajectories, with the objective of maximizing traffic efficiency and driving comfort. However, such a complicated model cannot be directly solved by existing optimization software. To rapidly obtain solutions, this study develops a Generalized Benders decomposition (GBD)- based solution algorithm to tackle the challenges of multi-vehicle combinatorial optimization and nonlinear trajectory optimization problems. Meanwhile, the property of finite convergence is proved. Numerical experimental results turn out that the traffic performance of the proposed model outperforms benchmark CAV control methods under different traffic demands and mainline-ramp demand ratios, demonstrating significant traffic benefits from jointly regulating lane changes, driving sequences, and utilizing microscopic vehicle information. Also, this study analyses traffic delays and the number of lane changes by the proposed model under varying road lengths, i.e., the lengths of lane-changing and merging areas.
The third work introduces a bi-level approach that nests optimization modelling within deep reinforcement learning to jointly optimize vehicle sequences, lane selection, and trajectories, aiming to provide a rapid, safe, and high-quality solution for the problem of multi-lane freeway merging. In the upper level, we develop an attention-based sequential policy network to sequentially construct driving sequences and lane selections for multiple vehicles. Specifically, we employ an attention mechanism to learn dynamic inter-dependencies with other vehicles, thus facilitating more informed and adaptive decision-making. In the lower level, we utilize a nonlinear model predictive controller to generate safe trajectories and use total travel delay to guide upper-level learning for optimizing long-term traffic efficiency. Additionally, we introduce a leader-and-lane specific credit assignment mechanism to address global credit assignment for the multi-vehicle merging problem. Computational results demonstrate that our method outperforms rule-based and searching-based methods in terms of solution quality, and the computation efficiency of the proposed approach is promising for real-time automated merging tasks.
| Date of Award | 2025 |
|---|---|
| Original language | English |
| Supervisor | Edward Chin Shin Chung (Chief supervisor) |