[Paper Review] Estimating and Mitigating the Congestion Effect of Curbside Pick-ups and Drop-offs: A Causal Inference Approach

Xiaohui Liu, Sean Qian+2 more|arXiv (Cornell University)|Jun 5, 2022

Transportation and Mobility Innovations4 citations

TL;DR

This paper proposes a causal inference framework using a double and separated machine learning (DSML) method to estimate the congestion effects of curbside pick-ups and drop-offs (PUDOs) on urban traffic speed, validated with Manhattan data. It finds that 100 additional PUDOs reduce traffic speed by 3.70–4.54 mph, and re-routing strategies can reduce system-wide travel time by up to 2.44%.

ABSTRACT

Curb space is one of the busiest areas in urban road networks. Especially in recent years, the rapid increase of ride-hailing trips and commercial deliveries has induced massive pick-ups/drop-offs (PUDOs), which occupy the limited curb space that was designed and built decades ago. These PUDOs could jam curbside utilization and disturb the mainline traffic flow, evidently leading to significant negative societal externalities. However, there is a lack of an analytical framework that rigorously quantifies and mitigates the congestion effect of PUDOs in the system view, particularly with little data support and involvement of confounding effects. To bridge this research gap, this paper develops a rigorous causal inference approach to estimate the congestion effect of PUDOs on general regional networks. A causal graph is set to represent the spatio-temporal relationship between PUDOs and traffic speed, and a double and separated machine learning (DSML) method is proposed to quantify how PUDOs affect traffic congestion. Additionally, a re-routing formulation is developed and solved to encourage passenger walking and traffic flow re-routing to achieve system optimization. Numerical experiments are conducted using real-world data in the Manhattan area. On average, 100 additional units of PUDOs in a region could reduce the traffic speed by 3.70 and 4.54 mph on weekdays and weekends, respectively. Re-routing trips with PUDOs on curb space could respectively reduce the system-wide total travel time by 2.44% and 2.12% in Midtown and Central Park on weekdays. Sensitivity analysis is also conducted to demonstrate the effectiveness and robustness of the proposed framework.

Motivation & Objective

• To address the lack of rigorous analytical frameworks for quantifying PUDO-induced congestion in urban networks with limited curb space.
• To model the causal relationship between PUDOs and traffic speed using a spatio-temporal causal graph to account for confounding factors.
• To develop a double and separated machine learning (DSML) method that enables unbiased estimation of PUDO congestion effects on traffic flow.
• To design and implement a re-routing formulation that optimizes system-wide travel time by encouraging walking and rerouting trips affected by PUDOs.
• To validate the framework using real-world traffic and PUDO data from Manhattan, demonstrating robustness and practical applicability.

Proposed method

• Constructs a causal graph to represent the spatio-temporal dependencies between PUDOs and traffic speed, identifying potential confounders.
• Develops a double and separated machine learning (DSML) method that separately models the outcome (traffic speed) and the treatment (PUDO count) to reduce bias.
• Applies machine learning models (e.g., random forests or gradient boosting) in two stages: first to predict potential outcomes under no-PUDO conditions, then to estimate the average treatment effect.
• Uses the estimated causal effect to inform a re-routing formulation that minimizes total system travel time by shifting trips away from high-PUDO zones.
• Solves the re-routing optimization problem using a customized algorithm that balances travel time, walking distance, and curb space constraints.
• Employs sensitivity analysis to validate the robustness of the DSML estimator and re-routing strategy under varying data and model assumptions.

Experimental results

Research questions

• RQ1 What is the causal impact of PUDOs on regional traffic speed, accounting for spatio-temporal confounders?
• RQ2 How can a double and separated machine learning (DSML) method accurately estimate the congestion effect of PUDOs without strong parametric assumptions?
• RQ3 To what extent can re-routing strategies mitigate the congestion caused by PUDOs in urban networks?
• RQ4 How do the congestion effects of PUDOs vary across different days of the week and urban zones (e.g., Midtown vs. Central Park)?
• RQ5 How does the model perform under varying data granularities and assumptions about error distributions in the estimation process?

Key findings

• On weekdays, 100 additional PUDOs in a region reduce average traffic speed by 3.70 mph, while on weekends the reduction reaches 4.54 mph.
• The re-routing strategy reduces the system-wide total travel time by 2.44% in Midtown and 2.12% in Central Park during weekdays.
• The DSML method provides theoretically grounded, unbiased estimates of PUDO congestion effects, with strong alignment to observed traffic patterns in Manhattan.
• Sensitivity analysis confirms the robustness of the DSML estimator and re-routing formulation under varying model assumptions and data conditions.
• The framework demonstrates potential for fine-grained application if higher-resolution data (e.g., street-level PUDO and speed data) becomes available.
• Future extensions, such as modeling pick-ups and drop-offs separately or tailoring re-routing for different vehicle types, could further improve congestion mitigation.