TL;DR
This paper introduces a novel "Dynamic Bus Priority Lane" (DBPL) strategy that intelligently manages mixed traffic of human-driven and autonomous vehicles. By allowing certain autonomous vehicles (CAVs) to use bus lanes while leveraging their "leading effect" in general lanes, the strategy aims to reduce overall traffic congestion and travel times without compromising bus priority.
What Problem Does This Paper Address?
Traditional bus lanes, while improving bus efficiency, can exacerbate congestion in general lanes. Dynamic bus lanes (DBLs) offer a solution by allowing other vehicles to use bus lanes when buses are not present. However, existing DBL strategies often rely on simple rules and don't fully leverage the advanced capabilities of connected and autonomous vehicles (CAVs), particularly their trajectory planning abilities. This paper addresses the challenge of integrating CAVs into DBL strategies in a way that maximizes system-wide efficiency. Specifically, it tackles how to selectively grant CAVs access to bus lanes, how to utilize their "leading effect" in general lanes to guide traffic, and how to accommodate complex scenarios like right-turning vehicles at intersections, all while ensuring bus priority is maintained.
Methods
The core of the proposed solution is a Dynamic Bus Priority Lane (DBPL) strategy. This strategy utilizes a Mixed-Integer Linear Programming (MILP) model to optimize the Right-of-Way (ROW) allocation for CAVs. Key methodological components include:
- Selective CAV Access: Instead of a blanket permission, the DBPL strategy selectively grants CAVs access to bus lanes based on optimization.
- Leveraging "Leading Effect": CAVs that remain in general lanes are used to guide platoons through intersections, indirectly influencing other vehicles' trajectories.
- Virtual Vehicles: A novel modeling approach using "virtual vehicles" is employed to estimate vehicle travel times and passing states, accounting for lane changes and their impact on following vehicles.
- Dynamic Control Framework with Rolling Horizon: To adapt to real-time traffic conditions and uncertainties in human-driven vehicle (HDV) behavior, a rolling horizon procedure is used to dynamically update the ROW allocation.
- Linearization Techniques: Nonlinear constraints within the optimization model are linearized to ensure it can be solved efficiently as a MILP problem.
- ROW Pre-allocation Heuristic Algorithm (ROWPH): To reduce computational complexity, a heuristic algorithm is introduced to pre-allocate ROW opportunities.
Datasets
The paper does not explicitly mention training datasets. The evaluation is performed using simulations conducted with SUMO (Simulation of Urban Mobility). The simulations are set up with specific traffic parameters and scenarios (Scenario A: without right-turn pocket, Scenario B: with right-turn pocket) to evaluate the proposed strategy.
Key Insights and Ideas
- The "Leading Effect" is Key: The strategy's effectiveness hinges on strategically using CAVs in general lanes to influence traffic flow, not just on allowing them into bus lanes. This dual approach is crucial for system-wide optimization.
- Selective Access Outperforms Blanket Access: Simply opening bus lanes to all CAVs can be detrimental. The DBPL's selective approach, guided by optimization, is more effective.
- Adaptability is Crucial: The dynamic control framework with a rolling horizon is essential for handling the inherent unpredictability of HDVs and real-time traffic fluctuations.
- Right-Turners Need Special Consideration: The strategy explicitly accounts for right-turning vehicles, which can temporarily use bus lanes, demonstrating its applicability in complex intersection environments.
Key Results
- Reduced Travel Times: The DBPL strategy significantly reduces private car travel times. At a 40% CAV market penetration rate (MPR), travel times were reduced by approximately 20% in Scenario A and 8% in Scenario B compared to an Exclusive Bus Lane (EBL) strategy.
- Maintained Bus Priority: CAB (Connected Autonomous Bus) travel times remained nearly identical between DBPL and EBL strategies, with a maximum difference of only 0.3 seconds, ensuring bus priority is maintained.
- Effectiveness Across Scenarios: The strategy demonstrated effectiveness in both scenarios (with and without right-turn pockets) and across various CAV MPRs.
- Sensitivity to Demand: Under high traffic demand, the DBPL strategy showed a more pronounced reduction in travel times, peaking at ~31% when MPR reached 40%.
- Robustness to Parameters: The strategy showed good performance across varying bus arrival intervals, bus stop positions, and right-turn ratios, highlighting its adaptability.
Limitations and Future Work
- No Signal Timing Optimization: The study does not consider the potential benefits of optimizing traffic signal timing alongside ROW allocation.
- Corridor/Network Level: The current study focuses on intersection-level optimization. Future research could explore ROW optimization at the corridor or network level.
- Partially Connected Environments: The study could be extended to investigate ROW optimization strategies in partially connected vehicle environments.
Summary updated