arxiv: 2604.21941 · v1 · submitted 2026-04-12 · 📡 eess.SY · cs.GT· cs.SY

Recognition: unknown

When Altruism Meets Autonomy: Managing Bottleneck Congestion with Strategic Autonomous Vehicles

Kexin Wang , Haohui He , Ruolin Li

Authors on Pith no claims yet

Pith reviewed 2026-05-10 15:17 UTC · model grok-4.3

classification 📡 eess.SY cs.GTcs.SY

keywords mixed autonomy trafficweaving rampsWardrop equilibriumStackelberg gameautonomous vehiclesbottleneck congestionlane choicesocial value orientation

0 comments

The pith

Under selfish human lane choices, autonomous vehicle penetration improves weaving-ramp performance only at critical thresholds with flat plateaus in between.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper builds a unified equilibrium model for weaving ramps that mixes selfish human-driven vehicles with centrally controllable autonomous vehicles. It proves that system performance as a function of AV share is non-increasing and stays constant across ranges of penetration until a threshold is crossed, at which point strategic AV control can shift the equilibrium to better outcomes. The model first uses Wardrop equilibrium for the humans and then treats AVs as Stackelberg leaders; it further allows heterogeneous social-value preferences in both populations. A sympathetic reader would care because the result shows how even a modest fleet of controllable vehicles can produce system-wide efficiency gains without requiring every driver to become altruistic or perfectly coordinated. The analysis supplies concrete guidance on where to focus AV deployment and incentive design rather than assuming benefits scale linearly with adoption.

Core claim

In a mixed-autonomy weaving ramp, when human-driven vehicles follow a Wardrop equilibrium based on selfish lane choices, the system performance as a function of AV penetration rate is non-increasing and exhibits plateau regions; performance improves only when AV penetration reaches certain critical thresholds where the strategic AVs can induce better equilibria.

What carries the argument

The Stackelberg-Wardrop formulation in which autonomous vehicles act as leaders optimizing aggregate performance while human-driven vehicles adapt through Wardrop equilibrium.

Load-bearing premise

Human drivers always choose lanes to minimize only their own travel cost according to Wardrop equilibrium, while AVs can be instructed to prioritize overall system performance.

What would settle it

Traffic-flow measurements at a weaving ramp that show continuous improvement in throughput or delay for every incremental increase in AV penetration, without any flat regions.

Figures

Figures reproduced from arXiv: 2604.21941 by Haohui He, Kexin Wang, Ruolin Li.

**Figure 2.** Figure 2: This figure illustrates the structure of lane-changing decisions within a highway weaving [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

**Figure 3.** Figure 3: The highway weaving ramp consists of three lanes. Lane 0 accommodates the entering flow [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

**Figure 4.** Figure 4: Two types of traffic interactions contribute to the costs of bypassing and steadfast be [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗

**Figure 5.** Figure 5: Validation results demonstrate strong agreement between the proposed lane-choice model [PITH_FULL_IMAGE:figures/full_fig_p015_5.png] view at source ↗

**Figure 6.** Figure 6: These figures show the system-level and individual-level impact of CAV penetration rate [PITH_FULL_IMAGE:figures/full_fig_p019_6.png] view at source ↗

**Figure 7.** Figure 7: This figure illustrates the SVO circle ( [PITH_FULL_IMAGE:figures/full_fig_p021_7.png] view at source ↗

**Figure 8.** Figure 8: Figure (a) presents the social cost phase comparison between homogeneous and hetero [PITH_FULL_IMAGE:figures/full_fig_p024_8.png] view at source ↗

**Figure 9.** Figure 9: Figure (a) presents the change of the social cost [PITH_FULL_IMAGE:figures/full_fig_p025_9.png] view at source ↗

**Figure 10.** Figure 10: These figures show three equilibrium cases for HDVs. For Case(a), The bypassing [PITH_FULL_IMAGE:figures/full_fig_p027_10.png] view at source ↗

**Figure 11.** Figure 11: This figure shows three equilibrium cases for mixed autonomy scenario. [PITH_FULL_IMAGE:figures/full_fig_p030_11.png] view at source ↗

**Figure 12.** Figure 12: These figures show type k’s behavior regimes. the regime where type k is mixed to the regime where type k + 1 becomes the potentially mixed type. When vehicle type k satisfy: χk − Wk(p) = wk(p), (60) vehicle type k becomes purely steadfast (i.e., x s 1,k = wk(p)). Vehicle types with χj ≥ χk are fully steadfast, while vehicle types with χj < χk are fully bypass. In this case, the system transitions to the … view at source ↗

read the original abstract

Weaving ramps are critical bottlenecks in highway networks due to conflicting traffic flows and complex interactions among heterogeneous vehicle types. In mixed-autonomy settings, the presence of controllable autonomous vehicles (AVs) introduces new opportunities to influence system-level outcomes, yet the structural impact of such control remains poorly understood. This paper develops a unified equilibrium framework to capture, predict, and optimize aggregate lane-choice behavior in weaving ramps with heterogeneous vehicle populations. We first formulate a Wardrop-based model capturing the selfish behavior of human-driven vehicles (HDVs) and establish existence, uniqueness, and validity of the resulting equilibrium. We then introduce a Stackelberg--Wardrop formulation in which AVs act as strategic leaders optimizing system performance, while HDVs respond through equilibrium adaptation. The framework is further generalized to incorporate heterogeneous behavioral preferences of HDVs and AVs via a Social Value Orientation (SVO) model. Our analysis reveals a fundamental structural property of mixed-autonomy traffic systems: under selfish HDV behavior, the impact of AV penetration is inherently non-increasing, exhibiting plateau regions where performance remains unchanged and improves only at critical thresholds. These results provide principled guidance for the design of AV control and incentive mechanisms in the presence of selfish human behavior, and demonstrate how strategically controlled autonomous agents can be deployed to induce system-level efficiency gains in mixed-autonomy transportation networks.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper derives a non-increasing impact of AV penetration on weaving-ramp performance under selfish HDV behavior, with clear plateau regions, using standard Wardrop-Stackelberg plus SVO extensions.

read the letter

The main thing to know is that this work identifies a structural property: when human drivers stick to selfish Wardrop lane choices at weaving ramps, adding autonomous vehicles improves system performance only at certain thresholds, with flat regions in between. The authors build a base Wardrop model for HDVs, prove existence and uniqueness under monotonic cost assumptions, then layer in Stackelberg leadership for AVs that optimize the aggregate outcome. They extend the setup to heterogeneous social value orientations and show the non-increasing property survives the generalization for the selfish case. That structural finding is the concrete contribution; it gives a clean way to think about critical AV penetration levels for bottleneck control without needing full system redesign. The modeling is transparent and sticks to established equilibrium concepts rather than inventing new ones. The proofs appear to rest on the usual monotonicity conditions that are stated up front, and the stress-test found no internal contradictions in how the property is preserved. One limitation is that the result is sensitive to how well those cost functions capture real driver behavior; if lane-choice costs have non-monotonic elements or if HDVs anticipate AV control, the plateaus could shift or disappear. The abstract does not include numerical checks or data comparisons, so the practical size of the thresholds remains open. This is worth a serious referee for anyone working on mixed-autonomy traffic management or Stackelberg-style control policies. The structural insight is specific enough to be useful for deployment planning even if the assumptions need testing. I would send it out for review rather than desk-reject.

Referee Report

1 major / 3 minor

Summary. The paper develops a unified equilibrium framework for aggregate lane-choice behavior at weaving ramps under mixed human-driven vehicles (HDVs) and autonomous vehicles (AVs). It formulates a Wardrop model for selfish HDV lane choice and proves existence, uniqueness, and validity of the equilibrium. It then introduces a Stackelberg-Wardrop formulation in which AVs act as strategic leaders optimizing system performance while HDVs adapt via equilibrium. The model is generalized via a Social Value Orientation (SVO) parameterization for heterogeneous preferences. The central structural result is that, under selfish HDV behavior, the impact of AV penetration on system performance is non-increasing and exhibits plateau regions, with improvements occurring only at critical thresholds.

Significance. If the monotonicity property and equilibrium results hold under the stated assumptions on cost functions, the work provides principled guidance for AV control and incentive design in mixed-autonomy bottleneck settings. The identification of plateau regions and threshold behavior is a useful structural insight that extends standard Wardrop and Stackelberg concepts to heterogeneous vehicle populations. The framework's reliance on explicitly maintained monotonicity assumptions on costs is a strength for reproducibility of the qualitative claims.

major comments (1)

[Abstract and §4] Abstract and §4 (Stackelberg-Wardrop extension): the non-increasing property and plateau behavior are asserted to hold under selfish HDV behavior, but the derivation appears to rely on the specific form of the cost functions and the Stackelberg leadership assumption; a concrete counter-example or boundary case where the property fails when monotonicity is relaxed would strengthen the claim that the result is structural rather than assumption-dependent.

minor comments (3)

[Abstract] Abstract: the phrase 'validity of the resulting equilibrium' is imprecise; clarify whether it refers to stability, consistency with observed flows, or another property.
[SVO generalization] SVO generalization section: the preservation of the non-increasing property under heterogeneous SVO parameters should be stated explicitly as a theorem or corollary rather than left as an extension remark.
[Notation] Notation: ensure consistent use of symbols for penetration rate, equilibrium flows, and system cost across the Wardrop and Stackelberg formulations to avoid reader confusion.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive comment and positive evaluation of our framework. We address the concern point by point below and will revise the manuscript to strengthen the exposition of our structural results.

read point-by-point responses

Referee: [Abstract and §4] Abstract and §4 (Stackelberg-Wardrop extension): the non-increasing property and plateau behavior are asserted to hold under selfish HDV behavior, but the derivation appears to rely on the specific form of the cost functions and the Stackelberg leadership assumption; a concrete counter-example or boundary case where the property fails when monotonicity is relaxed would strengthen the claim that the result is structural rather than assumption-dependent.

Authors: We agree that clarifying the dependence on monotonicity would improve the presentation. The non-increasing property and plateau behavior are proved in §4 under the standard assumption that lane-specific cost functions are strictly increasing in aggregate flows; this condition guarantees uniqueness of the Wardrop equilibrium for HDVs (see Theorem 1 in §3) and thereby makes performance comparisons across AV penetration rates well-defined. The Stackelberg leadership structure is part of the model definition rather than an auxiliary assumption. When monotonicity is relaxed, multiple HDV equilibria can coexist, so the notion of system performance in the Stackelberg-Wardrop game becomes set-valued and the monotonicity claim need not hold. In the revised §4 we will insert a short boundary-case example (two-lane weaving ramp with a non-monotonic cost segment) showing that equilibrium selection can produce non-monotonic performance changes with AV penetration. This addition will explicitly tie the structural result to the maintained monotonicity assumptions without altering the main theorems. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation relies on external equilibrium concepts

full rationale

The paper formulates a Wardrop equilibrium model for selfish HDV lane choice and establishes its existence, uniqueness, and validity from monotonicity assumptions on cost functions (explicitly stated and independent of the target result). It then extends to a Stackelberg-Wardrop game with AVs as leaders and generalizes via SVO, deriving the non-increasing AV penetration impact (with plateaus and thresholds) as a structural consequence of the selfish HDV equilibrium under those assumptions. No load-bearing step reduces by construction to a fitted parameter, self-definition, or self-citation chain; all cited concepts (Wardrop, Stackelberg) are standard external references, and the central property is not renamed or smuggled in but proven from the model equations.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Based solely on the abstract, the model rests on standard traffic equilibrium assumptions and introduces SVO for heterogeneity; no explicit free parameters or invented entities are detailed.

axioms (2)

domain assumption Human-driven vehicles exhibit selfish behavior captured by Wardrop equilibrium
Explicitly stated as the foundation for HDV lane-choice modeling.
domain assumption Autonomous vehicles can be controlled as strategic leaders in a Stackelberg game
Core of the mixed-autonomy formulation.

pith-pipeline@v0.9.0 · 5548 in / 1277 out tokens · 99413 ms · 2026-05-10T15:17:40.880140+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

3 extracted references · 3 canonical work pages

[1]

Adebisi, A., Liu, Y., Schroeder, B., Ma, J., Cesme, B., Jia, A., and Morgan, A. (2020). Developing highway capacity manual capacity adjustment factors for connected and automated traffic on freeway segments. Transportation Research Record, 2674(10):401–415. Ao, D., Lai, Z., and Li, S. (2024). Control of dynamic ride-hailing networks with a mixed fleet of ...

work page arXiv 2020
[2]

Safe, Multi-Agent, Reinforcement Learning for Autonomous Driving

Springer Science & Business Media. 37 Rios-Torres, J. and Malikopoulos, A. A. (2016). A survey on the coordination of connected and automated vehicles at intersections and merging at highway on-ramps.IEEE Transactions on Intelligent Transporta- tion Systems, 18(5):1066–1077. Rodrigues, M., McGordon, A., Gest, G., and Marco, J. (2018). Autonomous navigatio...

work page Pith review arXiv 2016
[3]

Van Lange, P

Springer. Van Lange, P. A. (1999). The pursuit of joint outcomes and equality in outcomes: an integrative model of social value orientation.Journal of personality and social psychology, 77(2):337. Vinitsky, E., Kreidieh, A., Le Flem, L., Kheterpal, N., Jang, K., Wu, C., Wu, F., Liaw, R., Liang, E., and Bayen, A. M. (2018). Benchmarks for reinforcement lea...

work page arXiv 1999