Source-linked AI summary
Congestion Pricing in a World of Self-driving vehicles: an Analysis of Different Strategies in Alternative Future Scenarios
Michele D. Simoni, Kara M. Kockelman, Krishna M. Gurumurthy, Joschka Bischoff
TL;DR
AVs and SAVs may reduce some traffic inefficiencies but also increase travel demand and VMT, making congestion pricing important. The paper uses MATSim to compare pricing strategies across Austin scenarios with high private-AV or SAV adoption. All strategies reduce congestion, while welfare outcomes vary by strategy and scenario, with revenue reinvestment affecting efficiency and acceptability.
Problem
AVs and SAVs may increase car-trip frequency, travel distances, VMT, and congestion despite potential capacity improvements, motivating analysis of demand-management strategies.
Method
The study uses MATSim to compare travel-time, link-based, distance-based, and flat facility-based tolls in alternative Austin AV/SAV adoption scenarios.
Results
All mobility schemes considerably reduce congestion, while the Travel Time-Congestion scheme produces the largest social welfare improvements across scenarios.
Takeaways & Limitations
More advanced pricing schemes can bring higher economic gains, but their effectiveness depends on the autonomous-mobility scenario and toll-revenue reinvestment.
Takeaways & Limitations
Scenario parameters received limited calibration, and predicting Austin’s future mobility would require land-use, detailed AV-ownership, and gas-price information.
Abstract
from arXiv · showhide
The introduction of autonomous (self-driving) and shared autonomous vehicles (AVs and SAVs) will affect travel destinations and distances, mode choice, and congestion. From a traffic perspective, although some congestion reduction may be achieved (thanks to fewer crashes and tighter headways), car-trip frequencies and vehicle miles traveled (VMT) are likely to rise significantly, reducing the benefits of driverless vehicles. Congestion pricing (CP) and road tolls are key tools for moderating demand and incentivizing more socially and environmentally optimal travel choices. This work develops multiple CP and tolling strategies in alternative future scenarios, and investigates their effects on the Austin, Texas network conditions and traveler welfare, using the agent-based simulation model MATSim. Results suggest that, while all pricing strategies reduce congestion, their social welfare impacts differ in meaningful ways. More complex and advanced strategies perform better in terms of traffic conditions and traveler welfare, depending on the development of the mobility landscape of autonomous driving. The possibility to refund users by reinvesting toll revenues as traveler budgets plays a salient role in the overall efficiency of each CP strategy as well as in the public acceptability.
INTRODUCTION
Autonomous and shared autonomous vehicles may improve accessibility, safety, and traffic flow, but could also increase motorized travel and congestion. The study compares congestion-pricing strategies across alternative AV/SAV futures using MATSim to assess traffic and social welfare effects.
- AVs and SAVs may replace trips from conventional cars, public transit, and bicycles, while improving accessibility, road safety, and energy consumption.
- Automation can improve throughput through fewer crashes, tighter headways, and better intersection use, but may increase trip frequency and distance by reducing driving burdens.
- Existing congestion-pricing policies are often simple cordon- or area-based tolls that do not vary with congestion, despite connected technologies enabling more efficient strategies.
- The study compares travel-time, link-based, distance-based, and flat facility-based tolls in high-private-AV and high-SAV adoption scenarios.
- MATSim represents traveler responses through changes in departure times, routes, activities, and modes, while keeping destinations fixed.
MOBILITY IMPACTS OF SELF-DRIVING VEHICLES AND CONGESTION PRICING
AVs and SAVs may improve road capacity while increasing car use, travel distances, and empty-vehicle travel. These changes strengthen the case for congestion pricing, and connected vehicles make more dynamic strategies feasible.
- Mobility impacts of self-driving vehicles: Reduced reaction times and following distances may increase road and intersection capacity, while cooperative communication can improve network performance.
- Mobility impacts of self-driving vehicles: Lower driving burdens, travel costs, and improved access will probably increase car use and travel distances before autonomous technologies resolve capacity constraints.
- Mobility impacts of self-driving vehicles: Empty SAV trips between passenger journeys may further increase VMT, with the magnitude depending on regional policies, mode options, parking costs, and trip patterns.
- Congestion pricing: Congestion pricing seeks to charge travelers for marginal external congestion costs, a longstanding concept in transportation economics and engineering.
- Congestion pricing: Real-time information exchange among connected vehicles could support tolls that vary dynamically across time, space, and charge level.
- Congestion pricing: The paper presents two advanced AV/SAV-enabled pricing schemes and analyzes them alongside alternative future scenarios with different AV and SAV market penetrations.
MODELING AVs AND SAVs WITH AN AGENT-BASED MODEL
The study uses MATSim as a dynamic agent-based model of daily activity plans, travel choices, congestion, and learning. AV and SAV behavior is represented through explicit mode costs, activity scheduling, vehicle dispatch, and increased autonomous-flow capacity.
- General Framework of MATSim: MATSim models endogenous mode, departure-time, and route choices across linked daily activities rather than isolated trips.
- General Framework of MATSim: Agents generate activity-and-trip plans, simulate them in a congested physical system, score alternatives, replan, and iterate toward a stable equilibrium.
- Choice dimensions and parameters: The daily plans include Home, Education, Work, Shopping, and Leisure activities connected by travel choices derived from real travel-demand data.
- Choice dimensions and parameters: Plan utility combines activity utility with trip disutility, including mode-specific constants, travel time, and monetary cost.
- Choice dimensions and parameters: Mode alternatives include car, public transit, bike and walk jointly, AV, and SAV, with SAVs additionally modeled using waiting times and fixed, distance, and time fares.
- Simulation of shared mobility services: SAV operations use dynamic vehicle routing with online dispatch and a demand-supply balancing strategy for large fleets.
SIMULATION SCENARIOS
The study compares a realistic Austin base scenario with AV-oriented and SAV-oriented futures, examining how new autonomous travel options alter modal choices and traffic conditions.
- Base Scenario: The Base Scenario represents Greater Austin, including Round Rock, Cedar Park, and Pflugerville, with a population exceeding 2 million.The region includes Interstate Highway 35 and lacks high-quality public transit options.
- Future Scenarios: AV-oriented and SAV-oriented scenarios add AVs and SAVs as travel options, reflecting uncertainty over private ownership versus shared-taxi adoption.These scenarios represent possible futures with strong autonomous-vehicle market penetration.
- Modal Split: Public transit falls to 4% of mode share in the AV-oriented scenario but rises to 8% in the SAV-oriented scenario.The differing shifts accompany the introduction of autonomous and shared autonomous travel options.
- Modal Split: Active trips decrease to 2% in the AV-oriented scenario and 4% in the SAV-oriented scenario.Lower private-vehicle ownership shifts some commuters toward public transit or SAVs.
- Traffic Conditions: Daily total VMT and travel delay increase in both autonomous scenarios, with empty SAV trips contributing 6.2% of total VMT in the SAV-oriented scenario.Autonomous-driving capacity gains are offset by increased trip-making, particularly empty SAV travel.
CONGESTION PRICING STRATEGIES
The study compares traditional and advanced congestion-pricing strategies for AV/SAV scenarios, including fixed, distance-based, link-level, and network travel-time-dependent tolls. Advanced schemes use traffic dynamics and iterative simulation feedback to determine tolls, but their implementation faces practical constraints.
- Strategy types: Four strategies comprise two traditional schemes—facility-based and distance-based—and two advanced schemes—link-based marginal-cost pricing and travel-time-congestion pricing.The advanced schemes require newer connected-vehicle technologies and greater computational or operational complexity.
- Traditional schemes: The Link-based Scheme applies a flat toll to the most congested links during morning and evening peak periods.Selected links are identified using congestion-related criteria, while the toll rate is uniform across those links.
- Traditional schemes: The Distance-based Scheme charges $0.10 per mile between 7 AM and 8 PM, without directly reflecting traffic dynamics.The strategy could become more time- or location-dependent with vehicle-position tracking technologies.
- Advanced schemes: The MCP-based scheme derives dynamic link charges from traffic volume, congestion costs, and Fundamental Diagram relationships between density and throughput.The approach estimates delay and tolls intended to eliminate queues and adjust throughput toward capacity.
- Limitations: MCP faces theoretical and practical difficulties because congestion is dynamic and operationally and socially optimal link tolls are difficult to set across large networks.The implementation also analyzed only a subset of 15,020 centrally located links and capped tolls at $0.30.
- Advanced schemes: The Travel Time-Congestion-based scheme varies network-level tolls with travel time and congestion, increasing penalties during more congested periods.The congestion component uses traffic measurements aggregated over 30-minute windows, with α = 0.1 in both AV- and SAV-oriented scenarios.
- Results and implementation: MCP tolls were derived after 10 to 15 simulations, with 5,000–7,000 of 28,484 analyzed links tolled at average charges of $0.02-$0.05 per link.Travel-time tolls showed higher morning-peak charges, and AV-oriented charges exceeded SAV-oriented charges because AV travel costs are lower.
RESULTS AND IMPLICATIONS
Across Austin scenarios, all congestion-pricing strategies reduce private travel and delay, but their mode-shift and welfare effects vary by strategy and autonomous-mobility scenario. Advanced schemes and revenue reinvestment generally improve welfare outcomes, while the most effective pricing approach changes with AV/SAV development.
- Mode choice: All congestion-pricing strategies reduce car, AV, and SAV trips while increasing public-transit and slow-mode shares.Road pricing makes road use more expensive, shifting some travelers toward transit and active modes.
- Mode choice: Distance-based pricing changes mode choice more than link-based pricing in the Base Scenario, whereas link-based pricing reduces AV trips more in the AV-Oriented Scenario.The two strategies differ in simplicity, congestion effectiveness, economic effects, distributional effects, and public acceptability.
- Travel demand: Distance-based pricing yields the greatest private-trip reduction in the Base Scenario, while link-based and MCP schemes outperform it in selected AV- and SAV-Oriented scenarios.Travel Time-Congestion pricing has the lowest demand effect in both autonomous-vehicle scenarios.
- Network performance: Link-based schemes reduce delay more than distance-based pricing in AV- and SAV-Oriented scenarios, while distance-based pricing performs better in the Base Scenario.The results indicate that VMT and delay rankings need not coincide across scenarios.
- Network performance: MCP-based pricing reduces delay by 2 to 5 percentage points more than the corresponding traditional link-based scheme, while Travel Time-Congestion pricing achieves comparable delay reductions to distance-based pricing at lower modal shifts.In the AV-Oriented scenario, lower mode shifts can coexist with similar delay reductions because travelers reroute and reschedule rather than switch modes.
- Social welfare: MCP-based and Travel Time-Congestion pricing produce the largest welfare gains when toll revenues are fully reinvested, especially in the SAV-Oriented Scenario.Without revenue consideration, every pricing strategy reduces social welfare; Travel Time-Congestion pricing then has the smallest consumer-surplus reduction in the autonomous scenarios.
- Policy implications: Only some pricing strategies generate sizable welfare gains after compensation, while Austin’s limited transit service constrains efficiency, especially for trips longer than 5 km.Reinvesting revenues in transit or traveler transfers may create net benefits and more winners than losers.
- Policy implications: Link-level strategies perform better in the Base and AV-Oriented scenarios, whereas distance- and travel-time-based strategies are more effective in the SAV-Oriented Scenario.The findings suggest that travel-demand management may need to evolve as autonomous-vehicle offerings change.
CONCLUSION
The study finds that congestion pricing can reduce congestion across autonomous-vehicle scenarios, but strategy performance and economic effects depend on the mobility landscape. It also identifies calibration, city-specific prediction, and distributional analysis as important boundaries for interpreting the results.
- CONCLUSION: Travel demand and delays rise in both future scenarios because travelers shift away from traditional transit and SAVs make empty trips.
- CONCLUSION: All mobility schemes yield considerable congestion reductions, while advanced strategies provide higher economic gains rather than consistently stronger demand or traffic effects.The relative performance of pricing strategies varies across future scenarios.
- CONCLUSION: The Distance-based scheme performs better in SAV-Oriented and Base Scenarios, whereas the Link-based scheme performs better in the AV-Oriented Scenario.
- CONCLUSION: The Travel Time-Congestion scheme produces the largest social welfare improvements in all scenarios.
- CONCLUSION: The simulations are intended as transparent, generalizable benchmarks rather than precise predictions of Austin’s future mobility.Prediction would require additional information on land use, AV ownership, and gas prices; calibration effort was limited.
- CONCLUSION: Future work should examine distributional effects, compensation measures, and interactions between dynamic SAV ride-sharing and pricing strategies.
APPENDIX I
The appendix tests different charge levels for the Distance-based and Link-based congestion-pricing schemes. The tables identify the best results in bold.
- APPENDIX I: Table 9 tests different charge levels for the Link-based scheme.
- APPENDIX I: Best results are indicated in bold in both charge-level tests.
- APPENDIX I: Table 10 tests different charge levels for the Distance-based scheme.