Guidelines and recommendations for future policy of cooperative and automated passenger cars

Deliverable D6.5 of the H2020 project LEVITATE
Author(s)
Gebhard, S.; Nabavi Niaki, M.; Schermers, G.; Goldenbeld, C.; Chaudhry, A.
Year

The aim of the LEVITATE project is to prepare a new impact assessment framework to enable policymakers to manage the introduction of connected and automated transport systems, maximise the benefits and utilise the technologies to achieve societal objectives. As part of this work, LEVITATE seeks to forecast societal level impacts of connected and automated transport systems (CATS) or, as these systems are more recently referred to, cooperative, connected, and automated mobility (CCAM). These impacts include effects on mobility, safety, environment and economy. Work Package 6 (WP6) considers the societal impacts of automated passenger car use in urban environments. Within WP6, the impacts of six policy measures related to particular developments in automated passenger cars are considered in what are termed sub-use cases (SUCs).

The goal of this Deliverable is to summarize the more detailed results presented in deliverables D6.2-D6.4 (Haouari, et al., 2021; Sha, et al., 2021; Chaudhry et al., 2021), to provide an overview of the main expected trends for each of the quantified impacts (e.g. emissions, congestion). Four methods were employed to quantify the future impacts of connected and automated vehicles: microsimulation, mesosimulation, system dynamics, and Delphi. Estimates derived from the microsimulation are based on models from the road networks in three European cities: Manchester, Santander and Leicester. The mesosimulation and system dynamics models are based on road networks from the city of Vienna, and the Delphi survey is not location-specific and relies on consensus-based estimates of a group of experts in the topic field. The methods were applied based on their strengths and capabilities to derive different impacts of an increasing penetration rate of CAVs within the total vehicle fleet, together with the deployment of each of the following six passenger car sub-use cases (SUCs)

  • Provision of dedicated lanes for CAVs (microsimulation, Delphi)
  • Replace on-street parking with other facilities (microsimulation, system dynamics, Delphi)
  • Road use pricing (mesosimulation, system dynamics, Delphi)
  • Parking price regulation (microsimulation, system dynamics, Delphi)
  • Green light optimal speed advisory (GLOSA) (microsimulation, Delphi)
  • Automated ride sharing (microsimulation, system dynamics, Delphi)

For each of these six SUCs, several implementation scenarios were estimated, varying for example the type of road use pricing (static or dynamic) or the various alternatives which may replace on-street parking (e.g. a driving lane or public space). For each impact, a baseline scenario is estimated; the baseline scenario refers to a “no intervention” scenario, representing the expected transition (autonomous development) of vehicle operations from human-dependence to fully automated vehicles. Any additional effect of the SUC interventions can be determined by comparing the baseline situation for a given penetration rate with the specific SUC results; the difference between the baseline and the SUC is the added effect produced by implementing the specific SUC intervention in the simulated network.

Approach to summarizing LEVITATE results

To summarize the many results from the three deliverables of Work Package 6 (Haouari, et al., 2021; Sha, et al., 2021; Chaudhry et al., 2021), a selection of sub-use case scenarios are presented in overview tables with their predicted trend for each of the considered impacts. Depending on the sub-use case, this meant excluding some scenarios from the overview which provided similar insights to other scenarios (e.g. lower demand shares within automated ride sharing) or represented less realistic extremes (e.g. cars not parking and remaining on the roads due to parking price regulation). For some sub-use cases, an average is taken of a few scenarios which consist of small variations of the same intervention and show similar results (e.g. different placements of a dedicated lane for the UK context). An average was chosen in order to 1) provide a more generalised trend for the SUC less specific to the simulation parameters and 2) due to the largely similar results between scenarios suggesting an overarching trend regardless of scenario. In addition, for some impacts where estimates are calculated by both a simulation/modelling method (microsimulation, mesosimulation, system dynamics) as well as the Delphi survey, Delphi results are excluded from the overview due to the other methods being considered the more rigorous methods within LEVITATE.

The predicted trends for each sub-use case scenario are quantified in terms of percentage changes reported across an increasing market penetration rate of connected and automated vehicles (CAVs) in the network vehicle fleet (see Figure 1 below). For each impact, the overview tables distinguish between the:

  • Baseline development (no intervention): the expected development as the proportion of CAVs in the traffic network increases to 100%
  • Intervention-based development (SUC): the expected development of the same impact when both the sub-use case intervention and increasing penetration levels of CAVs are at work

The percentage change takes the first stage of the baseline scenario as the reference point (zero percentage change). At this starting point of the baseline, no intervention has been implemented and no automated vehicles are present in the network (0% penetration of CAVs). The development of impacts (expressed as percentages indicating a decrease or increase from the initial value) under the baseline indicates the sole expected effect of increasing CAV penetration in total traffic. The development of impacts under the intervention-based condition indicates the expected effect of the combination of the SUC intervention and the growing automation level. The specific effect or impact of the SUCs can be determined by comparing the baseline situation for any given penetration rate with the specific SUC results; the difference between the baseline and the SUC is the added effect produced by implementing the specific SUC intervention in the simulated network.

Because of differences between the four studied cities’ road networks and traffic conditions, as well as differences in each methodology’s assumptions and scale, impact estimates across methods and cities can vary. This also results in variations in the baseline estimations for a given impact, depending on the city/method applied. Generalising the specific results is therefore not advised unless these are applied to similar conditions as adopted in Levitate. However, the estimates within and across methods and cities are indicative of expected trends for the SUCs tested.\

Findings

The main findings from the work done in WP6 on expected developments from the six major SUCs and the various scenarios are presented below. It should be noted that the impacts of some of the SUCs have been estimated using one or more of the different methods: microsimulation, mesosimulation, system dynamics and Delphi method. The microsimulation method was applied to 5 sub-use cases, the mesosimulation was appropriate to evaluate 1 sub-use case, the system dynamics method was utilized to evaluate 4 of the sub-use cases, while the Delphi method was applied to all 6 sub-use cases. The four different methodologies (microsimulation, mesosimulation, system dynamics, Delphi) were also each used to calculate a different set of mobility, environmental and societal/safety/economic impacts depending on the method’s characteristics (e.g. microsimulation is more suitable to small-scale traffic analysis while mesosimulation and system dynamics take systemic/wider impacts into account). Some impacts were calculated by multiple methods (but as explained earlier, not all are presented in this synthesis).

Baseline developments

The increasing penetration levels of connected and automated vehicles (CAVs) in the urban city area is estimated (for most baselines) to have a positive impact on the environment (less emissions, higher energy efficiency), on society, safety & economy (improved road safety, public health, and lower vehicle operating costs) and on most mobility indicators (more access to travel and less congestion). In the absence of policy interventions, some potentially negative effects could be realised if private automated vehicle transport leads to a decline in walking, cycling, and/or public transport trips.

The impacts for the various SUC are measured relative to the baseline starting point: the situation with no intervention or presence of automated passenger cars. Important to note is that baseline estimated vary across methods and the city networks to which CAVs and the SUCs were applied. In the microsimulation, results for the baseline estimates differ between SUCs due to different networks being studied for each SUC. For the mesosimulation and system dynamics impacts, one baseline was calculated for the entire city of Vienna, which may also show different effects from the networks used in the microsimulation. Also the baseline estimates in the Delphi method differ across SUCs because different expert groups evaluated different SUCs. The results therefore reflect the implementation of CAVs under a wide range of conditions, networks, and methodologies. The results serve as indicative of impact ranges rather than definitive estimates, which would have required a much larger study as well as more observational data which is unavailable due to the early stages of automated technology. Care must be taken in generalising the results to situations to those which are comparable to those modelled in Levitate. The results are transferable in as far as they are applied to networks that are comparable to those used in Levitate (Haouari et al., 2021; Sha et al., 2021; Chaudhry et al., 2021).

The following points summarise the primary baseline results:

  • Environment: For the CO2-emissions, in all five sub-use case networks a substantial positive baseline development was estimated, even at low levels of CAV penetration (CO2 emissions reduction of 20% at 20% CAV penetration rate). This is due to the assumption that all CAVs are electric vehicles. Out of the six estimated baseline scenarios for the development of energy efficiency, all expert groups predicted that energy efficiency will improve once CAV penetration exceeds 60% (5% to 31% improvement at full penetration). For two of these baseline estimates, a slight initial decrease in energy efficiency is predicted at low CAV penetration rates.
  • Mobility: For the mobility domain, on the positive side most (4 out of 5) baselines predict a reduction of congestion (between 9% and 13% reduction), most predict an increase in shared mobility rates (between 11% and 41% increase at 100% CAV penetration level), and most baselines predict an increase in the vehicle utilisation rate (four out of the six baselines estimate an increase between 43% and 50% increase at 100% CAV penetration level). However, on the negative side, most baselines in both the Delphi study and system dynamics model predict a reduction in the modal split of active transportation (between 13% and 56% reduction at 100% CAV penetration level) and public transport (between 10% and 29% reduction at full automation) in favour of CAVs. For total kilometres travelled, travel time and vehicle occupancy rates, different baselines produce very different results; there seems to be no clear pattern for these impacts.
  • Road safety: Regarding safety, for all the baseline networks, a substantial improvement in road safety is predicted at full penetration of automated vehicles (67% to 92% reduction in crash rates of crashes between passenger vehicles). At lower penetration rates when there is still mixed traffic on the road, the impact on crash rates is more gradual. For three out of the five baselines, an improvement in road safety is already clear with the presence of 40% CAVs on the road (11% to 25% reduction in crash rate). For the other two baseline networks (Santander and the GLOSA network from Manchester), a temporary increase in crash rates is predicted in low penetration rates when many human-driven vehicles are still on the road. This is likely due to interactions between human-driven vehicles and CAVs, whose different driving styles may cause some additional conflicts. Once human-driven vehicles are no longer on the road, all baselines show a large decrease.
  • Society/Economy: Most baselines (5 out of 6) predict an improvement in public health (2% to 12% improvement), and all 6 baselines predict an improvement in equal accessibility of transport (mostly between 4% and 25%). Effects on vehicle operating costs are mixed as CAV penetration increases to 100% (-20% to +13%), but at lower penetrations (20% to 40% CAVs) all baselines predict vehicle operating costs to at least temporarily increase. The results for parking space demand estimated by system dynamics predict parking space demand to increase as CAVs become more widespread. This is due to a predicted mode shift towards private automated vehicle travel from other modes (e.g. public transport), thus increasing the demand for parking if no further policy measures are taken.

The effect of the SUC interventions

Below is a summary of the findings from the six major sub-use cases and their implementation scenarios:

  1. Road use pricing: Road use pricing is expected to lead to a number of additional benefits over the baseline impacts: better energy efficiency (dynamic toll more than static toll or empty km pricing), less reduction in the use of active modes and public transport, higher vehicle occupancy rate, and lower parking space demand. On the negative side, road use pricing is expected to lead to increase in vehicle operating costs, and less equal accessibility of transport. The scenario "empty km pricing” is expected to contribute more positively towards keeping vehicle operating costs within bounds compared to the “static toll” and “dynamic toll” scenarios. The “static toll” scenario is expected to result in the highest shares in active transport modes and public transport. The “dynamic toll” scenario is expected to lead to the highest vehicle occupancy rates.
  2. Dedicated CAV lanes: Compared to the baseline, dedicated lanes for CAVs do not make a clear difference for emissions, travel time, kilometres travelled, and road safety. On the positive side, dedicated lanes are expected to lead to better access to travel when lanes are “dynamic,” slightly reduced congestion in mixed human-driven/CAV traffic, a higher vehicle utilisation rate, higher vehicle occupancy rate, and lower vehicle operating costs. The “dynamic” lanes scenario performs better than the “fixed” lane scenarios in terms of improvements on energy efficiency, access to travel, vehicle occupancy rate, vehicle operating costs, and in terms of a lesser decrease of the active mode share.
  3. Parking price regulation: Parking price regulation does not seem to make a noticeable difference on CO2 emissions or shared mobility rate. On the positive side, parking price regulation is expected to compensate some of the negative impacts of CAVs on the mode share of public transport and active modes, resulting in more walking, cycling and public transport use than in the baseline development. However, in both cases these interventions cannot overcompensate for the negative impact that CAVs are expected to cause to the modal share of public transport and active modes. Less private vehicle use than in the baseline also results in a reduced demand for parking space, according to the system dynamics model. The alternative parking behaviours resulting from parking price regulation are predicted to have some potentially negative (or less positive) effects compared to the baseline development on: energy efficiency, travel time and congestion, and road safety. However, these negative effects predicted by microsimulation and Delphi do not take into account the effects on modal split predicted by system dynamics which may counteract these effects to a certain degree if private vehicle transport is reduced. Reduced benefits are also predicted for access to travel, equal accessibility of transport, vehicle utilisation rate, and vehicle operating costs, due to the increased costs of parking in central locations.
  4. Replacing on-street parking: The interventions aimed at replacing on-street parking show similar results to the baseline (no added effect) in terms of CO2 emissions. Positive effects on mobility in terms of reduced travel time and congestion are predicted due to the reduction in parking manoeuvres. For many of the impacts, the effect of replacing on-street parking was dependent on the scenario: removing half of spaces, replacing with driving lanes, replacing with pick-up/drop-off spaces for shared CAVs, or replacing with public space or cycling lanes. Replacing with “public space” was found to be particularly beneficial for energy efficiency, shared mobility rate, modal splits of active and public transport, vehicle occupancy rate, vehicle operating cost, road safety, and public health, but negative in terms of access to travel. Meanwhile, replacing on-street parking with “driving lanes” is expected to improve access to travel and use of active modes (to a lesser degree than public space), but reduce the mode share of public transport and negatively impact public health and parking space demand. The scenario “pick-up/drop-off” generally performs worse than the other scenarios in terms of energy efficiency, travel time, kilometres travelled, congestion, shares of active transport modes and public transport, shared mobility rate, and road safety. On the positive side, the “pick-up/drop-off” scenario is expected to result in better results for access to travel and equal accessibility of transport than the other scenarios in the replacing on-street parking SUC.
  5. Automated ride sharing: Automated ride sharing does not make a noticeable difference for CO2 emissions or parking space demand. Compared to the baseline, extra benefits are expected in terms of energy efficiency, access to travel, public transport use, shared mobility rate (at lower CAV penetrations), vehicle occupancy rate, and vehicle operating costs. Compared to the baseline, it has a negative impact on congestion (due to empty vehicle kilometres needed to reposition vehicles), travel time, use of active modes, and vehicle utilisation rate. The impact on road safety is mixed: at low CAV penetrations, automated ride sharing improves safety by serving a share of otherwise human-driven trips. However, as all trips become automated the added benefit reduces and the extra congestion caused by ride sharing rather serves to slightly increase crash rates compared to the baseline at high penetration rates. Furthermore, the impacts of automated ride sharing depend on what share of the users are willing to share trips with other users. When willingness to share is low (20% scenario), less positive results are predicted for mobility and road safety due to the larger number of trips and vehicles needed to serve the demand (travel time and congestion increase; kilometres travelled and road safety decrease).
  6. GLOSA: The GLOSA sub-use case is associated with no noticeable additional impacts on CO2 emissions or kilometres travelled. Compared to the baseline it shows positive impacts on travel time, congestion, public transport use, road safety, and vehicle operating costs. A negative (or less positive) impact compared to the baseline is predicted for access to travel, active mode share, shared mobility rate, vehicle utilisation rate, public health, and equal accessibility of transport.

Overall conclusions

In summary, the following primary conclusions were drawn:

  • Increasing penetration levels of connected and automated vehicles in the urban city area are estimated (for most baselines) to have positive impacts on the environment (less emissions, higher energy efficiency), on society and economy (improved road safety, public health, and lower vehicle operating costs) and on mobility (more access to travel and less congestion). The predicted decrease in the modal share of public transport, walking and cycling in favour of automated passenger cars, however, may lead to some undesirable effects (such as the predicted increase in demand for parking space) without further policy measures.
  • Road use pricing is expected to lead to a number of benefits above baseline developments, especially regarding mobility and environmental concerns: better energy efficiency, less reduction of active mode share, higher vehicle occupancy rate, less negative impact on public transport mode share, and less parking demand. On the negative side, road use pricing is expected to lead to an increase in vehicle operating costs, and lower accessibility to transport.
  • Dedicated CAV lanes are predicted to have limited additional impacts on most indicators. Slight benefits were estimated for congestion, vehicle operating costs, vehicle utilisation and occupancy rates, as well as public health.
  • Parking price regulations causing CAVs to return to other locations to park or drive around while waiting for passengers showed mixed results. Some negative effects are predicted on mobility (e.g. congestion), as well as the environment (energy efficiency), road safety, public health and accessibility of transport. These negative effects are primarily due to extra empty vehicle kilometres needed to reposition vehicles after passenger drop-off. However, increased parking costs also have the potential to stimulate a moderate mode shift away from private vehicle transport, which may benefit use of active modes of travel and decrease the demand for parking space. Results regarding the effects on public transport use are mixed.
  • Of the six major sub-use cases, replacing on-street parking is associated with a wide range of positive benefits over the baseline, including a large improvement in traffic conditions (reduced travel time and congestion), more positive development in active mode share, more shared mobility, better development of road safety and lesser demand for parking space. The facilities chosen to replace on-street parking also influence the impacts. Replacing on-street parking with public space is particularly associated with societal and environmental benefits (e.g. road safety, public health, energy efficiency) and is beneficial for shared, public, and active forms of mobility. Replacing on-street parking with driving lanes or pick-up/drop-off points is generally associated with lesser benefits, except for improved access to travel. Pick-up/drop-off points or removing only half of spaces also reduce the benefits to congestion due to maintaining some of the parking manoeuvres.
  • Automated ride sharing is expected to benefit vehicle sharing, accessibility, and energy efficiency. While it is predicted to attract a moderate mode shift away from private vehicle transport, automated ride sharing is also predicted to attract trips away from walking and cycling. Furthermore, the additional empty vehicle kilometres necessary to reposition the vehicles to pick up their next passengers may lead to an increase in congestion, counteracting the benefits of the trips which can be shared. The impact of an automated ride sharing system is also dependent on the population’s willingness to share trips with other travellers: a higher willingness to share is associated with less negative effects on congestion and marginally better road safety.
  • GLOSA is not predicted to have large additional impacts on most indicators. Slight benefits to the traffic conditions are predicted (reduced congestion and travel time) as well as less decrease in public transport use and reduced vehicle operating costs. Potential negative effects on shared mobility rate, active travel, vehicle utilisation and occupancy rates, access to travel and public health are predicted. These negative effects may be due to a predicted increase in private vehicle travel with implementation of GLOSA.
  • The policies considered in the SUCs have little additional impact on generated emissions; the large, expected reductions are primarily driven by the transition to CAVs which are assumed to be electric vehicles. The large, expected improvements in road safety with increasing automation are also driven by behavioural differences in CAVs (e.g. quicker reaction times) compared to human-driven vehicles, which is impacted minimally by SUC policies.

Sub-use case related recommendations

  • Automated vehicles may provide benefits such as additional comfort, efficiency, the potential for multitasking, and accessibility to travellers who are not able to drive a vehicle themselves. This may cause a potential modal shift from other modes of travel (e.g. public transport, cycling, walking) towards private vehicle travel, which may be undesirable to cities for a number of reasons (e.g. energy usage, public health, use of public space). In order to limit potential increases in private vehicle transport, road use pricing, replacing on-street parking with public space, and parking price regulation may be useful policy measures.
  • Replacing on-street parking with public space is predicted to be associated with more benefits than replacing the space with driving lanes, provided care is taken that sufficient accessibility is retained.
  • The benefits of an automated ride sharing system are highly dependent on the users’ willingness to combine trips and it has the potential to increase congestion due to empty repositioning trips. Therefore, the suitability of local conditions for an automated ride sharing system should first be studied before implementation.
  • GLOSA is associated with some moderate benefits to traffic conditions, although more efficient traffic flow may also attract more private vehicle use. Therefore, GLOSA may be best paired with other measures to encourage practices such as vehicle sharing and active travel.

Strengths and limitations of Levitate

The followings observations pertain to strengths and limitations of research within WP6 LEVITATE. A potential strength of the LEVITATE project is that both smart city transport policy interventions and the associated impacts have been selected by a diverse group of stakeholders. A wide variety of impacts were studied at the same time and the project tried to capture interdependencies. The best available methods - microsimulation, mesosimulation, Delphi, and other complementary methods such as system dynamics and operations research - were used to study and quantify the expected impacts of mobility interventions intended to support CAV deployment and sustainable city goals. Within LEVITATE project these impacts provide essential input for developing a practical Policy Support Tool for city policy makers.

Concerning limitations, it should be pointed out there are general scientific difficulties in predicting impacts of connected and automated mobility due to uncertainties about propulsion energy, future capacity of power grids, employment, development of costs, and about the behaviour and acceptance with regard automated vehicles. The results of the models in LEVITATE are dependent upon specific assumptions. The simulation models used examined only two CAV profiles (aggressive vs. cautious); future work may extend the number of profiles. The safety results of the microsimulation did not include crashes where vulnerable road users are involved.

Given the many uncertainties in prediction, it is obvious that any predicted values are associated with large uncertainty. For WP6, it was decided not to include estimates of confidence intervals based on the standard error, derived from repeated trail runs of models, as a standard output since these intervals would be broad and non-informative. Also, the estimation of these intervals would tend to be biased in itself since the input variables and assumptions in the models are very likely much stronger determinants of predicted values than the variability in sample runs.

LEVITATE has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 824361.

Pages
130
Publisher
European Commission, Brussels

SWOV publication

This is a publication by SWOV, or that SWOV has contributed to.