Resolving the Mobility-Equity-Emissions Trilemma: A Causal Machine Learning Framework for Dynamic Tolling Policy

TL;DR

The study introduces a causal machine learning approach to optimize dynamic tolling policies that simultaneously improve traffic flow, ensure equitable access, and reduce emissions. It identifies a stable toll range that maintains performance across varying traffic conditions and demonstrates significant emissions reductions while expanding coverage for residential zones.

Why It Matters

Dynamic tolling policies often face trade-offs between efficiency, fairness, and environmental impact. This work provides a practical, robust framework that enables real-world deployment by identifying a policy-invariant toll range, simplifying implementation and enhancing public acceptance.

Key Findings

• Commercial zones experience nearly double the delay reduction from dynamic tolling compared to residential zones.
• An adaptive quota mechanism increases residential zone coverage by 34 times while retaining 93% of delay savings and maintaining high return on investment.
• Congestion-sensitive pricing reduces emissions by 743.5 metric tons of CO2 under moderate traffic, with sensitivity bounds between 557.6 and 929.3 tons, and a policy-invariant toll range ensures consistent performance across demand scenarios.