By Michael Brenneis
The anticipated shift to autonomous vehicles raises several concerns, among them are whether AVs will increase total vehicle miles traveled, exacerbate congestion, or replace the use of transit and active modes. A new study focused on Austin, TX, models the effects that two adoption scenarios may have, comparing the results to current conditions. The experiment shows a jump in VMT, increased congestion, and a shift away from public transportation. The authors then mitigate the scenarios by applying four tolling schemes, and examine the results.
If AV adoption rates are high, road capacity may initially increase since AV connectivity would permit them to operate closer together, at least until the ease of traveling in an AV encourages more driving. AV users may be willing to travel farther, and spend more time in vehicles, because they can use time in the vehicle to do other things. For those conventional drivers remaining on the road as this transition occurs, travel times could increase, making conventional driving even more unattractive.
The modelers examined three scenarios: widespread adoption of private AVs; widespread use of shared autonomous vehicles (SAVs); and a baseline scenario with no replacement of cars by AVs, reflecting current conditions. According to the researchers it is possible that AVs would replace privately owned vehicles or be adopted in a shared capacity.
The study employed MATSim software, a multi-agent transportation simulation capable of modeling reactions to demand management strategies, including modified start times, trip duration, and destination shuffling, in an effort to calculate results more realistic than traditional single-trip models. Model parameters have generally been derived from previous work, and further calibration of model parameters is left to future work, although the authors argue that their parameters are justifiable and have produced reasonable results. The SAV model minimized the size of the fleet needed to serve all users, and utilized dynamic routing to more realistically represent conditions.
The simulated mode split for the three scenarios (Fig. 1) shows increases to both VMT and travel delay (Table 1.) A reduction in the proportion of trips made by walking/biking or public transit is also evident, down from 16 percent to 6 and 12 percent respectively. Some of the mode shift observed in this study may be attributable to a lack of transit or active mode options positioned to replace auto travel.
Conventional congestion pricing tends to be facility-based, including tolls on bridges, tunnels, and highways. Four tolling schemes were modeled including a link-based approach that applies tolls to the most congested road segments during peak travel time, and a distance-based toll applied ($0.10 per mile) between the hours of 7AM and 8PM to all roads. These are both considered traditional forms of congestion mitigation.
The connectivity of AVs allows tolls to be dynamically applied to each road segment according to traffic conditions. This is referred to as the marginal cost pricing scheme (MCP). The other advanced tolling strategy applied in the study is the travel-time congestion scheme, which charges drivers for the network level delay that they cause. It costs more to travel at times of peak demand on congested roads, similar to ridesharing companies’ surge pricing.
As one might expect, all of the tolling scenarios result in reduced VMT and delay (Fig. 2). The AV scenario sees the most VMT reduction under the link-based scheme, a traditional method, and the most congestion relief under either the link-based scheme or the MCP scheme. For the SAV scenario MCP appears to reduce VMT the most, while also significantly reducing delay. In order to benefit social welfare, revenue from any of the tolling schemes would need to be reinvested, particularly in transit or in improvements to pedestrian and bicycling facilities.
The technology required to implement some of the tolling schemes that connected vehicles make possible can be much cheaper than traditional tolling infrastructure, allowing a more widespread implementation.
Michael Brenneis is an Associate Researcher at SSTI.
By Michael Brenneis