Auction-based autonomous intersection management

Carlino, Dustin; Boyles, Stephen D.; Stone, Peter · 2013 · OpenAlex-citations

DOI: 10.1109/itsc.2013.6728285

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper proposes a decentralized, auction-based mechanism for autonomous intersection management (AIM) to optimize traffic flow and address congestion. The authors argue that while autonomous vehicles (AVs) offer significant potential for reducing delay, existing reservation systems often rely on inefficient first-come, first-served protocols that ignore travelers’ varying values of time. By implementing auctions at intersections, the system allows drivers to bid for priority, enabling higher-valued trips (e.g., emergency vehicles or time-sensitive commuters) to proceed sooner. The study addresses the feasibility of this approach for AVs, which can handle the computational burden of bidding, unlike human drivers. Crucially, the paper also tackles social equity concerns by introducing a "benevolent system agent" that subsidizes bids to prevent wealthier drivers from indefinitely stalling poorer ones. The proposed framework utilizes second-price sealed-bid auctions. Drivers are represented by automated "wallet agents" that bid on their behalf based on trip characteristics, remaining budget, and distance to destination. Three wallet types are defined: a "free-rider" (never bids), a "static" wallet (fixed bid, modeling emergency vehicles), and a "fair" wallet (distributes remaining budget across remaining intersections). To ensure fairness and throughput, the system agent applies multiplicative rates to bids, effectively setting reserve prices that reward drivers for reducing queue spillback, maintaining capacity, and ensuring throughput. This mechanism is compatible with stop signs, traffic signals, and autonomous reservation protocols. Experiments were conducted using AORTA, an open-source microscopic traffic simulator, across four city networks: Austin, Baton Rouge, San Francisco, and Seattle. Each simulation involved 30,000 drivers with uniformly distributed budgets. The auction-based schemes were compared against baseline first-come, first-served (FIFO) and equal-priority orderings, with and without system bid regulation. Results indicated that auction-based management generally reduced total trip times compared to FIFO. For instance, in Austin, weighted trip times decreased from 4,853 seconds (FIFO) to 4,240 seconds (Auction with system bids). In Baton Rouge, the inclusion of system bids significantly improved performance, reducing weighted trip times from 1,758 seconds (Equal) to 1,235 seconds (Auction with system bids). However, in Seattle, system bids slightly increased trip times compared to auctions without them, suggesting that parameter tuning is context-dependent. The study concludes that auction-based AIM can effectively prioritize trips based on value of time, leading to substantial time savings in congested networks. The integration of system bids is critical for maintaining social equity, preventing the exclusion of low-budget drivers, and ensuring overall network throughput. The authors note that while the current implementation uses manually tuned parameters for system bids, future work should focus on automated optimization and more diverse bidding strategies to better reflect real-world market dynamics. The open-source nature of the implementation allows for further reproducibility and extension of these findings.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• traffic density