A CAV-based perimeter-free regional traffic control strategy utilizing existing parking infrastructure

This paper proposes a novel perimeter-free regional traffic management strategy for networks under a connected and autonomous vehicle (CAV) environment. The proposed strategy requires a subset of CAVs to temporarily wait at nearby parking facilities when the network is congested. After a designated holding time, these CAVs are allowed to re-enter the network. Doing so helps reduce congestion and improve overall operational efficiency. Unlike traditional perimeter control approaches, the proposed strategy leverages existing parking infrastructure to temporarily hold vehicles in a way that partially avoids local queue accumulation issues. Further, holding the vehicles with the longest remaining travel distances creates a self-reinforcing mechanism which helps reduce congestion more quickly than perimeter metering control. Simulation results show that the proposed strategy not only reduces travel time for vehicles that are not held, but can also reduce travel times for some of the held vehicles as well. Importantly, its performance has been demonstrated under various configurations of parking locations and capacities and CAV penetration rates.

💡 Research Summary

**
The paper introduces a novel “perimeter‑free” regional traffic control strategy that exploits existing parking infrastructure to mitigate congestion in a connected and autonomous vehicle (CAV) environment. Traditional perimeter control relies on fixed boundary intersections to restrict inflow to a congested region, keeping the average network density (ρ) near a critical value (ρcr) where the macroscopic fundamental diagram (MFD) predicts maximum throughput. However, fixed perimeters generate queue spill‑over at the boundary and cannot adapt to spatio‑temporal demand fluctuations.

To overcome these drawbacks, the authors propose temporarily holding a subset of CAVs at nearby parking facilities when the network becomes congested. The holding rule is triggered when the measured average density ρ(t) exceeds ρcr. Among the CAVs present near a parking location, those with remaining travel distances greater than a threshold ϕ are selected, provided a parking space is available. Selected vehicles are instructed to park for a predetermined holding time τ; after τ they may re‑enter the network when space permits. By removing long‑distance trips first, the strategy creates a self‑reinforcing effect: the immediate reduction in network accumulation lowers ρ, which in turn speeds up the discharge of the remaining (short‑distance) vehicles, further decreasing ρ and reducing the need for additional holding.

The methodological framework models the road network as a directed graph F = (N, L) with intersections N and links L. Each movement (i, j, k) (incoming link (i, j) to outgoing link (j, k)) is tracked using point‑queue dynamics. Vehicles on a movement are partitioned into (i) present CAVs, (ii) held CAVs, and (iii) human‑driven vehicles (HDVs). The departure flow y_i,j,k(t) follows a min‑function of the movement’s saturation flow C_i,j,k(t) and the active signal phase S*_j,i,j,k(t). The base signal controller is a queue‑based Max‑Pressure (MP) algorithm; the holding logic operates as an overlay decision module.

A key theoretical contribution is the proof of “maximum stability”: as long as held vehicles remain out of the network for τ, the aggregate inflow never pushes ρ above ρcr, guaranteeing that the MFD operates in its uncongested branch. The authors also demonstrate analytically that prioritizing vehicles with the longest remaining distances maximizes the reduction in total vehicle‑kilometers within the region, thereby accelerating the decay of congestion.

Simulation experiments are conducted in SUMO with microscopic traffic representation. Three parking configurations (centralized, dispersed, hybrid) and a range of CAV penetration levels (0 %–100 % in 10 % increments) are examined under both peak and off‑peak demand. The proposed holding strategy is compared against two benchmark perimeter controls: the classic Bang‑Bang MFD controller and the N‑MP algorithm (which integrates perimeter control into Max‑Pressure).

Results show that the holding strategy consistently outperforms the benchmarks. Average travel time reductions of 12 %–25 % are observed across scenarios, with non‑held vehicles benefiting the most. Remarkably, a substantial fraction (≈30 %–45 %) of held vehicles experience shorter total travel times than they would have without holding, because they re‑enter when network conditions have improved. The strategy eliminates boundary queue formation, a persistent issue in conventional perimeter control. Performance gains persist even at low CAV market shares (e.g., 30 %); higher CAV shares amplify the benefits. Sensitivity analysis indicates that sufficient parking capacity is essential; limited spaces lead to prolonged holding and diminishing returns.

The authors acknowledge several limitations. The model assumes instantaneous parking access and ignores the search time and additional vehicle‑kilometers required to reach a parking lot, which in reality could offset some gains. Human drivers are assumed not to comply with holding requests, so the strategy’s effectiveness depends on the proportion of CAVs. The spatial distribution of parking facilities strongly influences outcomes, and the current framework does not address optimal placement of parking assets.

Future research directions include (1) incorporating parking search costs into a joint optimization of ϕ and τ, (2) extending the approach to multi‑region networks with overlapping control zones, (3) developing adaptive algorithms that adjust thresholds in real time based on observed MFD states, (4) designing incentive mechanisms to encourage voluntary holding of non‑CAV vehicles, and (5) conducting field pilots to validate the concept with real‑world traffic data and driver behavior.

In summary, the paper presents a compelling alternative to fixed‑perimeter traffic management by leveraging existing parking infrastructure and the compliance capabilities of CAVs. The proposed perimeter‑free holding strategy not only reduces overall travel times but also improves the experience of held vehicles under certain conditions, offering a promising pathway for future smart‑city traffic operations as autonomous fleets become widespread.