Evaluating a system of exclusive lanes forautonomous vehicle platoons

Gustavo Covas, Eduardo F. Z. Santana, Fabio KonDepartment of Computer Science - IME / University of São Paulo

Introduction

The adoption of Autonomous Vehicle (AV) technolo-gies has the potential to reduce traffic jams and in-crease traffic safety. However, despite recent advance-ments on automation technology, full autonomous ve-hicles are not expected to be available in the short-term [1].In this work, we analyzed a proposal to allow AVs toshare the roads with regular vehicles: Digital Rails(DR). The concept of DR was first elaborated by de-signers at Questtonó, a design consultancy firm.The impacts of implementing DR in São Paulo wereanalyzed with traffic simulations modeled after realdata and conducted using InterSCSimulator, a smartcity simulator developed at IME/USP.

Figure 1: Concept art for Digital Rails, elaborated by Questtonó

Digital RailsThe authors of the Digital Rails proposal present threemain pillars for the system:

1 Open data network: A network that vehicles canaccess to obtain data about traffic signals.

2 Exclusive lanes: On selected arterial roads,exclusive lanes would be assigned to the system.

3 Vehicle platoons: Vehicles using Digital Railswould organize themselves in platoons and travelthrough traffic signals progressions.

However, they left open the required level of vehicleautomation, communication protocols, and platoon-ing techniques. We assumed that these problems aresolved and abstracted them away in our simulationmodel.

InterSCSimulatorInterSCSimulator is a smart-city simulator based ondiscrete events. It is written in Erlang and relies onSimDiasca as a discrete event simulation engine. Theroad network is represented as a graph where eachroad segment is a link between two vertices.In traffic simulations, vehicles travel each link in aconstant speed that depends on the link vehicle den-sity on the moment the vehicle enters it. The speedv when entering a link with vehicle density k is givenby:

v = vfree ∗ (1 − ( k

kjam)β)α (1)

Where vfree is the maximum speed that a vehiclecould develop in the link, kjam is the maximum ve-hicle density supported by the link, and α and β areconfigurable parameters for the model.

Simulation ScenariosWe began simulating DR on a segment of AvenidaPaulista, first developing a benchmark scenario basedon reports published by CET and the current traf-fic signal timing plan. Using a mixed integer linearprogramming formulation [2], we calculated a traf-fic signal timing plan to allow DR platoons to travelwithout stops.For a constant traffic flow equivalent to peak hourson Avenida Paulista, we simulated different ratios ofvehicles travelling on DR platoons. We also devel-oped scenarios to study the travel time for vehiclestravelling on crossing roads.

Figure 2: Location of traffic signals on the selected segment ofAvenida Paulista.

Based on CET reports, we selected a region of SãoPaulo and some major routes with significant trafficvolume to expand the DR system. The selected routeswere:

1 Avenida 23 de Maio2 Avenida Pedro Álvares Cabral / Avenida Brasil3 Marginal Pinheiros (between Jaguaré bridge andEng. Ari Torres bridge)

4 Ligação Leste-Oeste5 Avenida Cidade Jardim / Avenida Europa6 Avenida Paulista7 Avenida Eusébio Matoso / Avenida Rebouças

The resulting network can be seen in the followingfigure:

Figure 3: Proposed Digital Rails network in São Paulo, highlightedin green

Using a simulated mobility trace [3], we studied howdifferent ratios of vehicles able to use DR would af-fect the average travel time on a regular day in theselected region. Since we did not simulate individualtraffic signals, we re-calibrated the simulation modelparameters to take the delay introduced by them intoaccount.

ResultsOn Avenida Paulista, the average travel time waslower than the benchmark when more than 25% of thevehicles used DR. We also found that the travel timefor vehicles on the crossing roads did not increase.

0 20 40 60 80 100ratio

0

200

400

600

800

1000

Travel time (s) vs. DR ratio (%)Mean, no DR95% CI, no DRMean, DR

Figure 4: Travel time vs. ratio of vehicles using DR on AvenidaPaulista, compared to the average travel time on the benchmarkscenario.

With the proposed DR network, we compared the av-erage travel time with the benchmark scenario. Sincethe range of travel distances was large, we dividedthe analysis in quartiles. The reduction of travel timewas less expressive. Even with 100% of the vehiclesable to use DR, the travel time was about 70% of thebenchmark.

0

1

2 (29.999, 3677.0] (3677.0, 6545.0]

0.00 0.25 0.50 0.75 1.00

ratio

0

1

2 (6545.0, 11162.0]

0.00 0.25 0.50 0.75 1.00

ratio

(11162.0, 33201.0]

Fraction of benchmark travel time vs DR ratio

Figure 5: Evolution of travel time on the selected region of SãoPaulo with DR ratio, by distance quartiles. The legend on eachplot indicates the travel distance range on each quartile.

ConclusionImplementing Digital Rails on São Paulo could leadto reduced travel times, even for vehicles that will notuse the system. However, our simulation model wassimple and did not capture individual vehicle interac-tions. The DR proposal should be further evaluatedby professionals with experience in traffic operations.

References

[1] T. Litman, Autonomous vehicle implementationpredictions.Victoria Transport Policy Institute, 2017.

[2] J. D. Little, M. D. Kelson, and N. M. Gartner, “Maxband:A program for setting signals on arteries and triangularnetworks,” Transportation Research Record, no. 795, 1981.

[3] E. F. Z. Santana, L. Kanashiro, and F. Kon, “Geração derastros de mobilidade para experimentos em redesveiculares,” 2018.