Pricing Traffic Networks with Mixed Vehicle Autonomy

Negar Mehr, Roberto Horowitz

Abstract

In a traffic network, vehicles normally select their routes selfishly. Consequently, traffic networks normally operate at an equilibrium characterized by Wardrop conditions. However, it is well known that equilibria are inefficient in general. In addition to the intrinsic inefficiency of equilibria, the authors recently showed that, in mixed-autonomy networks in which autonomous vehicles maintain a shorter headway than human-driven cars, increasing the fraction of autonomous vehicles in the network may increase the inefficiency of equilibria. In this work, we study the possibility of obviating the inefficiency of equilibria in mixed-autonomy traffic networks via pricing mechanisms. In particular, we study assigning prices to network links such that the overall or social delay of the resulting equilibria is minimum. First, we study the possibility of inducing such optimal equilibria by imposing a set of undifferentiated prices, i.e. a set of prices that treat both human-driven and autonomous vehicles similarly at each link. We provide an example which demonstrates that undifferentiated pricing is not sufficient for achieving minimum social delay. Then, we study differentiated pricing where the price of traversing each link may depend on whether vehicles are human-driven or autonomous. Under differentiated pricing, we prove that link prices obtained from the marginal cost taxation of links will induce equilibria with minimum social delay if the degree of road capacity asymmetry (i.e. the ratio between the road capacity when all vehicles are human-driven and the road capacity when all vehicles are autonomous) is homogeneous among network links.

Introduction

Traffic congestion is a pervasive issue in modern urban environments, largely driven by the "selfish" routing behavior of individual drivers. In transportation theory, this behavior leads to a Wardrop Equilibrium, where no driver can reduce their travel time by unilaterally changing routes. Unfortunately, this equilibrium is rarely socially optimal, often resulting in significantly higher total system delay than a centrally managed flow.

The introduction of autonomous vehicles (AVs) introduces a new dynamic. While AVs can increase road capacity by maintaining shorter headways, recent findings suggest that in mixed-autonomy environments, this increased capacity can paradoxically worsen network inefficiency—a phenomenon akin to the Braess Paradox. This research investigates whether congestion pricing (tolls) can mitigate these inefficiencies. Specifically, the study explores whether "undifferentiated pricing" (charging all vehicles the same toll) is sufficient, or if "differentiated pricing" (charging AVs and human drivers different rates) is necessary to reach the social optimum.

Methodology

The researchers model the traffic network as a nonatomic routing game using a modified version of the US Bureau of Public Roads (BPR) delay function. A key parameter in this model is the degree of capacity asymmetry, $\mu_l$, defined as the ratio of road capacity when occupied solely by human drivers to the capacity when occupied solely by AVs. Since AVs are more efficient, $\mu_l$ is typically less than 1.

To evaluate pricing strategies, the authors utilize two primary approaches:

  1. Counter-example Analysis: A synthetic 4-link network with two origin-destination pairs is used to test the efficacy of undifferentiated pricing.
  2. Marginal Cost Taxation: The authors derive differentiated prices based on the marginal cost each vehicle type imposes on the system. They prove the effectiveness of these prices in "homogeneous networks"—where the capacity benefit of AVs is uniform across all links—by constructing an auxiliary game.
A synthetic 4-link network used to demonstrate the limitations of undifferentiated pricing.

Results and Insights

The study yields several critical insights into the management of mixed-autonomy traffic:

The Necessity of Differentiated Pricing

The most significant finding is that undifferentiated pricing is fundamentally insufficient for reaching the social optimum in mixed-autonomy networks. Because human drivers and AVs consume road capacity at different rates, a single price cannot simultaneously incentivize both groups to choose the socially optimal routes. In the synthetic 4-link network shown above, the minimum social delay was calculated at 193.54, while the best possible result under undifferentiated pricing was 195.597.

The "Price Ratio" Rule

For homogeneous networks, the authors propose an optimal pricing scheme where the price for an AV ($\tau_l^a$) is proportional to the price for a human driver ($\tau_l^h$) based on the capacity asymmetry: $\tau_l^a = \mu \tau_l^h$. Since $\mu < 1$, this implies that AVs should be charged less than human drivers to encourage their use on congested links, thereby maximizing the capacity benefits they provide.

Guaranteed Optimality in Homogeneous Networks

Theorem 1 of the paper provides a formal proof that in any homogeneous network, setting link prices according to the marginal cost of each vehicle class guarantees that every resulting equilibrium will achieve the minimum possible social delay. This also addresses the issue of non-uniqueness in mixed-autonomy equilibria, ensuring that total social delay remains minimal regardless of the specific flow distribution.

Conclusion and Future Work

While differentiated pricing offers a theoretical "silver bullet" for homogeneous networks, several challenges remain. The current proof relies on network homogeneity; in heterogeneous networks where AV benefits vary by road, the proposed pricing may not always reach the absolute social optimum. Furthermore, the model assumes all users value time and money identically, ignoring the complexities of user heterogeneity.

Future research should explore path-based pricing strategies and incorporate varying values of time among drivers. Additionally, the practical implementation of differentiated pricing—charging different rates for humans and AVs on the same road—presents significant infrastructure and policy hurdles that must be addressed to realize these theoretical gains in real-world traffic management.

Summary updated

Abstract

In a traffic network, vehicles normally select their routes selfishly. Consequently, traffic networks normally operate at an equilibrium characterized by Wardrop conditions. However, it is well known that equilibria are inefficient in general. In addition to the intrinsic inefficiency of equilibria, the authors recently showed that, in mixed-autonomy networks in which autonomous vehicles maintain a shorter headway than human-driven cars, increasing the fraction of autonomous vehicles in the network may increase the inefficiency of equilibria. In this work, we study the possibility of obviating the inefficiency of equilibria in mixed-autonomy traffic networks via pricing mechanisms. In particular, we study assigning prices to network links such that the overall or social delay of the resulting equilibria is minimum. First, we study the possibility of inducing such optimal equilibria by imposing a set of undifferentiated prices, i.e. a set of prices that treat both human-driven and autonomous vehicles similarly at each link. We provide an example which demonstrates that undifferentiated pricing is not sufficient for achieving minimum social delay. Then, we study differentiated pricing where the price of traversing each link may depend on whether vehicles are human-driven or autonomous. Under differentiated pricing, we prove that link prices obtained from the marginal cost taxation of links will induce equilibria with minimum social delay if the degree of road capacity asymmetry (i.e. the ratio between the road capacity when all vehicles are human-driven and the road capacity when all vehicles are autonomous) is homogeneous among network links.
BibTeX
@inproceedings{mehr2019pricing,
  author = {Negar Mehr and Roberto Horowitz},
  title = {{Pricing Traffic Networks with Mixed Vehicle Autonomy}},
  booktitle = {ACC},
  year = {2019},
  url = {https://openreview.net/forum?id=sY8QFClyHR},
}
Figure