Traffic Light Scheduling for Pedestrian-Vehicle Mixed-Flow Networks

Yi Zhang, Kaizhou Gao, Rong SuIEEE Trans. ITS 2019 · OpenReview

TL;DR

This paper introduces a novel framework for optimizing traffic light schedules in mixed pedestrian-vehicle networks. By developing a "hopping" model for pedestrian flow and integrating it with a vehicle traffic model, they formulate a problem that balances the needs of both road users, aiming to reduce overall delays and pedestrian unhappiness.

What Problem Does This Paper Address?

Urban traffic management traditionally prioritizes vehicle flow, often at the expense of pedestrian convenience and safety. Existing traffic signal control strategies are largely vehicle-centric, leading to significant delays and dissatisfaction for pedestrians, especially in increasingly pedestrian-friendly urban areas. This paper addresses the challenge of creating a unified traffic signal scheduling system that considers both pedestrian and vehicle dynamics simultaneously, aiming for a more equitable and efficient urban transportation network. The authors highlight the unique challenges of modeling pedestrian flow, which is less continuous and more dependent on junction-specific dynamics compared to vehicle flow.

Methods

The core of the paper lies in its novel modeling approaches and optimization strategies:

  • Pedestrian Hopping Model: A key innovation is the "hopping" model for pedestrian flow. This model captures how pedestrians move between waiting zones (corners) at an intersection, considering their dynamics and the capacity of crosswalks. It ensures that all pedestrians are cleared at the end of each GREEN signal phase.
  • Mixed-Integer Linear Programming (MILP): The pedestrian and integrated traffic signal scheduling problems are formulated as MILP problems. This allows for the use of powerful existing optimization solvers like GUROBI.
  • Discrete Harmony Search (DHS) Algorithm: To overcome the computational complexity of MILP, especially for larger networks and longer prediction horizons, a meta-heuristic algorithm called Discrete Harmony Search (DHS) is employed. This algorithm mimics musical improvisation to find near-optimal solutions efficiently.
  • Vehicle Flow Model: A discrete-time vehicle flow model, based on an extended cell transmission model, is used to represent vehicle dynamics. This model incorporates drivers' psychological responses to traffic light states.
  • Weighted Sum Approach: To integrate pedestrian and vehicle objectives, a weighted sum method is used. This allows for adjusting the relative importance of pedestrian versus vehicle needs by assigning different weights to their respective cost functions.
  • Function Transformation: To ensure that the pedestrian and vehicle cost functions are on a comparable scale, a function transformation technique is applied.

Datasets

Not discussed in the paper. The authors rely on simulated data generated by their models.

Key Insights and Ideas

  • Pedestrian "Hopping" is Crucial: The paper demonstrates that a simple cell transmission model is insufficient for pedestrian flow. The "hopping" model, which accounts for waiting zones and crosswalk clearance, is essential for accurate pedestrian traffic simulation.
  • Balancing Act is Complex: Optimizing for both pedestrians and vehicles is not a simple trade-off. The paper shows that the influence of pedestrian needs on vehicle traffic can be non-linear and dependent on the specific weighting assigned.
  • DHS for Scalability: For real-world applications with large networks and long prediction horizons, MILP becomes computationally prohibitive. The DHS algorithm offers a viable alternative for achieving near-optimal solutions in a timely manner.
  • Unpredictability of Pedestrian Influence: While vehicle traffic is generally more dominant, significant changes in pedestrian signal priorities can lead to noticeable shifts in vehicle traffic signal behavior, especially at initial stages of weight adjustments.
  • Switching Frequency as an Indicator: The switching frequency of traffic signals differs significantly between delay-focused and unhappiness-focused pedestrian models, suggesting that the latter can better mitigate the issue of long-waiting pedestrians.

Key Results

  • MILP Solvability: The MILP formulation for pedestrian delay can be solved efficiently for moderately sized networks (e.g., 10x10 junctions with a 30s horizon) within practical time limits using GUROBI.
  • DHS Efficiency: The DHS algorithm significantly reduces computation time for larger networks and longer prediction horizons compared to MILP, achieving near-optimal solutions with a small gap (around 20% for a 10x10 network with a 90s horizon).
  • Integrated Model Performance: The integrated model demonstrates that adjusting the weight given to pedestrian needs can alter vehicle traffic signal behavior. The impact of pedestrian signals on vehicle delay is often piecewise constant, meaning small changes in pedestrian priority might not immediately affect vehicle traffic.
  • Switching Frequency Analysis: The study shows that the pedestrian unhappiness model leads to a higher switching frequency of traffic signals compared to the delay model, suggesting it better addresses fairness concerns.

Limitations and Future Work

  • Assumptions: The model makes several simplifying assumptions, such as known pedestrian demands, diversion ratios, and the absence of pedestrian non-compliance behavior. These assumptions are acknowledged and noted for future relaxation.
  • Real-time Data: The current model relies on pre-defined demands. Future work could incorporate real-time data-driven model identification to adapt to dynamic conditions.
  • Network Complexity: The current case study focuses on simplified one-way vehicle flows and basic intersection configurations. Future research should consider more complex scenarios, including turning flows and multi-phase traffic signals.
  • Weight Selection: Determining the optimal weighting between pedestrian and vehicle priorities remains a challenge, requiring careful consideration for different urban contexts.
  • Vehicle Network Sensitivity: The paper notes that the vehicle network may not always be sensitive to pedestrian traffic, suggesting further investigation into how to design integrated systems that effectively capture these interactions.

Summary updated

Abstract

This paper presents a macroscopic model for pedestrian-vehicle mixed-flow network and a traffic signal scheduling strategy for both pedestrians and vehicles. We first propose a novel mathematical model of pedestrians crossing a junction. By combining a link-based vehicle network model, a traffic light scheduling problem is formulated with the aim to strike a good balance between pedestrians' needs and vehicle drivers' needs. The problem is first converted into a mixed-integer linear programming (MILP) problem via a novel transformation procedure, which is solvable by several existing solvers, e.g., GUROBI. Then a meta-heuristic method called discrete harmony search (DHS) algorithm is also adopted to reduce the computational complexity in MILP. Numerical simulation results are provided to illustrate the effectiveness of our real-time traffic light scheduling strategy for pedestrians and vehicles, and the potential impact of the pedestrian movement to the vehicle traffic flows.
BibTeX
@article{zhang2019traffic,
  author = {Yi Zhang and Kaizhou Gao and Yicheng Zhang and Rong Su},
  title = {{Traffic Light Scheduling for Pedestrian-Vehicle Mixed-Flow Networks}},
  journal = {IEEE Trans. Intell. Transp. Syst.},
  year = {2019},
  url = {https://openreview.net/forum?id=0WAWin3pLl},
}
Figure