How Will the Presence of Autonomous Vehicles Affect the Equilibrium State of Traffic Networks?

Negar Mehr, Roberto Horowitz

TL;DR

While autonomous vehicles (AVs) can increase road capacity through tighter "platooning," their introduction into a network of selfish drivers doesn't always improve traffic. This paper proves that unless capacity gains are distributed uniformly across the network, increasing the fraction of AVs can actually lead to higher total delays—a phenomenon similar to the famous Braess’s Paradox.

What Problem Does This Paper Address?

The common wisdom is that AVs will be a boon for traffic. Because they can communicate and react faster than humans, they can maintain shorter headways (the distance between vehicles), effectively "platooning" to increase the capacity of existing roads.

However, most traffic networks operate under selfish routing: drivers choose the path that is fastest for them individually, often leading to a "Wardrop Equilibrium" where no one can reduce their own travel time by switching routes. The authors ask a critical question: If we replace a fraction of human-driven vehicles with AVs that also route themselves selfishly, is the network guaranteed to get faster? Or does the increased capacity on certain links tempt drivers into routes that ultimately clog the system for everyone?

Methods

The authors use a nonatomic routing game framework to model traffic. Key components include:

  • Modified BPR Models: They adapt the standard U.S. Bureau of Public Roads (BPR) delay functions. In their model, a road's capacity is no longer a fixed constant but a function of the "autonomy ratio" on that specific link.
  • Capacity Asymmetry ($\mu$): They define a ratio between the capacity of a road used only by regular vehicles ($m_l$) versus one used only by AVs ($M_l$).
  • Equilibrium Analysis: They investigate the uniqueness and monotonicity of the Wardrop Equilibrium. Specifically, they look at two scenarios: Homogeneous networks (where AVs provide the same relative capacity boost on every road) and Heterogeneous networks (where some roads benefit more from autonomy than others).

Datasets

This is a theoretical paper and does not use real-world traffic datasets. Instead, the authors evaluate their mathematical proofs using:

  • Synthetic Network Topologies: Including the classic Wheatstone bridge network (often used to demonstrate Braess's Paradox).
  • Numerical Simulations: Five specific examples are constructed to demonstrate non-uniqueness of equilibrium and counter-intuitive delay increases.

Key Insights and Ideas

  • The Loss of Uniqueness: In standard traffic models, the total network delay at equilibrium is usually unique. The authors show that in mixed-autonomy networks, this is no longer true. Depending on how traffic is distributed, there could be multiple equilibrium states with different total delays for the same number of AVs.
  • Braess’s Paradox via Autonomy: The paper demonstrates that increasing AV penetration is mathematically equivalent to adding capacity to specific links. Just as building a new road can sometimes slow down a city (Braess's Paradox), "upgrading" a road's capacity via AV platooning can shift traffic patterns in a way that increases the total social delay.
  • The Homogeneity Requirement: The authors prove that if every road in a single O/D (Origin/Destination) network benefits equally from AV technology (homogeneous asymmetry), the total delay is guaranteed to be a non-increasing function of the AV fraction. In this specific case, more AVs always mean better traffic.
  • The Price of Vehicle Autonomy: They introduce a new metric, $\eta$, which measures the worst-case increase in delay caused by the introduction of AVs.

Key Results

  • Performance Degradation: In heterogeneous networks, the authors show cases where a network with 100% AVs actually has a higher total delay than a network with 0% AVs.
  • Bounding the Damage: For networks with homogeneous capacity asymmetry, the authors derive an upper bound on how much the "selfishness" of mixed traffic can hurt performance. For standard polynomial delay functions (degree $\le 4$), the total delay can be at most ~2.15 times the delay of the non-autonomous case.
  • Multiple O/D Pairs: Even with homogeneous roads, having multiple origins and destinations can reintroduce paradoxes. An increase in AVs for one group of commuters might significantly worsen the delay for another group.

Limitations and Future Work

  • Selfish AVs: The study assumes AVs are just as "selfish" as humans. In reality, AVs could be programmed to follow "socially optimal" routes, though this raises questions about user acceptance and competition between AV manufacturers.
  • Equilibrium Stability: While the paper proves that multiple equilibria can exist, it doesn't explore which one a real-world system is likely to settle into or how "stable" those states are.
  • Tightness of Bounds: The authors note that while they have established a bound for the "Price of Vehicle Autonomy," further research is needed to determine if this bound is "tight" (i.e., the absolute worst-case scenario).
  • Policy Implications: The results suggest that simply deploying AVs isn't enough. To actually improve mobility, cities may need to implement active traffic management, such as dynamic tolling or centralized routing strategies, to prevent these paradoxes from occurring.

Summary updated

Abstract

It is known that connected and autonomous vehicles are capable of maintaining shorter headway and distances when they form platoons of vehicles. Thus, such technologies can potentially increase the road capacities of traffic networks. Consequently, it is envisioned that their deployment will also increase the overall network mobility. In this paper, we examine the validity of this expected impact, assuming that drivers select their routes selfishly, in traffic networks with mixed vehicle autonomy, that is, traffic networks with both regular and autonomous vehicles. We consider a nonatomic routing game on a network with inelastic (fixed) demands for a set of network origin destination (O/D) pairs, and study how replacing a fraction of regular vehicles by autonomous vehicles will affect the mobility of the network. Using well-known U.S. Bureau of Public Roads traffic delay models, we show that the resulting Wardrop equilibrium is not necessarily unique for networks with mixed autonomy. Then, we state the conditions under which the total network delay at equilibrium is guaranteed to not increase as the fraction of autonomous vehicles increases. However, we show that when these conditions do not hold, counterintuitive behaviors may occur-the total network delay can grow as the fraction of autonomous cars increases. In particular, we prove that for networks with a single O/D pair, if the road degrees of capacity asymmetry (i.e., the ratio between the road capacity when all vehicles are regular and the road capacity when all vehicles are autonomous) are homogeneous, the total delay is: 1) unique and 2) a nonincreasing continuous function of the fraction of autonomous vehicles in the network. We show that for heterogeneous degrees of capacity asymmetry, the total delay is not unique, and it can further grow as the fraction of autonomous vehicles increases. We demonstrate that similar behaviors may be observed in networks with multiple O/D pairs. We further bound such performance degradations due to the introduction of autonomy in general homogeneous networks.
BibTeX
@article{mehr2020how,
  author = {Negar Mehr and Roberto Horowitz},
  title = {{How Will the Presence of Autonomous Vehicles Affect the Equilibrium State of Traffic Networks?}},
  journal = {IEEE Trans. Control. Netw. Syst.},
  year = {2020},
  url = {https://openreview.net/forum?id=368BbQNYXOe},
}

Loading related content...

Figure