Skip to main content
conference lunch move company map contacts lindholmen lindholmen 2 travel info


BCG’s Autonomous Vehicle Deep Dive

Over the last year Boston Consulting Group and the University of St. Gallen carried out a detailed simulation of transportation conditions using a tool that can model 1.7 billion trips. [1] We’ve summarised their findings for you. 

The one-year study combined qualitative and quantitative approaches with current industry insights. For the qualitative part, they asked more than 30 leading executives from universities, cities, and transportation-related industries for their perspectives on the key enablers, success factors, and roadblocks facing AVs.

The quantitative part used a simulation tool to assess the technology’s effects over time on six key performance indicators (KPIs): traffic volume, road fatalities, transportation costs, total parking space, energy consumption, and journey times. This was so that the authors could examine how AVs might improve or worsen the urban environment and quality of life in different types of cities – and across a range of different future scenarios. The city types simulated using this tool were:

  • High compact middleweight (e.g. Berlin, Seattle)
  • Car-centric grid (e.g. Los Angeles, Toronto)
  • Prosperous innovation center (e.g. London, San Francisco)
  • Developing urban powerhouse (e.g. Bangkok, Buenos Aires)
  • High-density megacity (e.g. New York, Shanghai)

These city types, or ‘virtual cities’ as referred to in the study, emerged from an analysis of more than 40 cities from around the world. The virtual cities were identified based on similarities such as the cities’ age, population density, congestion, urban street pattern, journey times, and topography. 

Behind the scenes of the simulation: 

Before the simulation was run, the authors had to first model the current and future modal split for each of the city types. This was done using publicly available data from the city analyses to model the current modal split and to establish an initial set of KPIs. For the current modal split, five transportation modes were included: private cars and motorbikes, public transportation, taxis/ride-hailing, micromobility (including bicycles), and walking. 

To create a projection of what the future modal split might be in each virtual city if inhabitants had the option of using AVs, the authors used the findings of a study conducted jointly by BCG and the World Economic Forum (WEF)—Reshaping Urban Mobility with Autono­mous Vehicles, published in June 2018—that asked 5,500 inhabitants across the globe what mode of transportation they would choose for making different types of journeys. Hence, three on-demand AV modes were added to the 5 existing modal split options: robo-pods (which seat up to 2 passengers), robo-taxis (up to 5 passengers), and robo-shuttles (up to 15 passengers). The resulting future modal split then served as the base-case scenario for the simulation. 

Note that the modal splits are inputs into the simulation, rather than outputs. The simulation only provided outputs in terms of the KPIs mentioned before (traffic volume, road fatalities, transportation costs, total parking space, energy consumption, and journey times).

As well as the modal splits, the simulation also took into account the distribution of mobility ­options across short-, medium-, and long-distance trips, using current data along with the results of the BCG/WEF survey. The authors also factored average vehicle speed, parking spaces for private cars, curb space for AVs, and average vehicle occupancy into the model.

Simulation results for the base-case:

The key finding was that, on average, AVs delivered improvements across all KPIs: 

  • The land area needed for parking spaces shrinks by 35%.
  • Traffic volume declines by 4% as new shared transportation modes cause the number of vehicles on city streets to drop.
  • Fatalities from road accidents decline by 37% per year because AVs reduce the incidence of driver errors.
  • Transportation costs as a percentage of household incomes dip by 13%.
  • Average journey times fall by 3% as the lower traffic volume resulting from the use of shared AVs reduces congestion.
  • Energy consumption declines by 12% because of the switch from private cars to more-efficient electric-powered AVs.

Four future scenarios:

The authors then sought to explore how four possible scenarios (that differ from the base-case in terms of modal split) would play out across our five city archetypes. To accomplish this, they raised or lowered the share of a given transportation option in the future modal split and then assessed the impact of that change on the KPIs. 

Scenario 1: Shift from Private Cars to Non-AV Transportation Modes 

This is a scenario where city authorities introduce policies that curb private car trips and encourage other forms of transportation. In this scenario, it is assumed that self-driving vehicles are still in their infancy and therefore had little impact on the modal split. In lieu of a reduction in the modal share of private cars and AVs compared to the base-case scenario, the shares for public transit and micromobility were increased. 

Scenario 2: Dominance of Micromobility.

A possible scenario where the share of e-bikes and e-scooters increases substantially. Rather than being primarily a last-mile transportation solution, micromobility would become a citywide phenomenon and encourage more multimodal journeys. In this scenario, modal share for micromobility increases while private vehicles and AVs decrease. 

Scenario 3: Strong Push for Robo-Shuttles.

Shared robo-taxis and robo-shuttles become the central component of urban transportation systems. The new vehicles are a hybrid form, with far less passenger occupancy than mass transit offers but far more than private cars. In this scenario, robo-shuttles carrying up to 15 passengers on semi-flexible routes win out over robo-pods (carrying only 2 passengers) and the dominant form of AV in the modal mix compared to the base-case. 

Scenario 4: Strong Uptake of Robo-Pods. 

Small robo-pods dominate the modal split, as their greater flexibility and privacy in comparison with other options make them a popular choice. This is the inverse of Scenario 3, where smaller robo-pods (up to 2 passengers) become the dominant form of AVs over larger robo-taxis and robo-shuttles. 

Simulation results for the 4 scenarios:

Just like the base-case, simulations for the 4 scenarios were done for each of the 5 city types. The results of this analysis are clearly summed up in a matrix diagram showing the impacts on the KPIs (compared to the base-case) across the 20 different combinations of scenarios and city types. To allow for easy general comparisons, a 100-point scale index based on the percentage change in the simulated KPIs for each city type was used - with each KPI contributing equally to the overall score.


What does it all mean? 

How AVs perform in practice will depend on each city’s particular characteristics and policies. In all cities, introducing robo-pods (up to 2 passengers) into the transportation mix would result in greater congestion. Although self-driving vehicles will emerge to some extent in every archetype, other forms of transportation, such as micromobility, could deliver greater benefits for city dwellers in some circumstances. In addressing their urban mobility challenges, cities should take a holistic approach that considers levers such as promoting micromobility, further regulating private cars, and introducing AVs.

The article goes into a bit more depth on what the simulation results mean for each of the city types modeled. But a broad summary is that for three of the virtual cities (car-centric giants, prosperous innovation centers, and high-density megacities), promoting robo-shuttle use would deliver the greatest advantage as measured by improvements in the model’s KPIs. For developing urban powerhouses, however, micromobility would deliver more benefits; and for highly compact middleweights, a shift from private cars to other non-AV modes of transportation would be the smartest choice. Still, in every case, choosing the best-choice scenario for each archetype’s characteristics would yield significant real-world benefits - see Exhibit 6 in the study article. Furthermore, restricting private car use is critical to all cities’ success.

The article also contains some suggested action steps for the different types of cities – based on the qualitative part of the study, where they asked more than 30 leading executives from universities, cities, and transportation-related industries for their perspectives on the key enablers, success factors, and roadblocks facing AVs. The key areas the authors suggest for cities to focus on should be AV-specific infrastructure(such as separate lanes and sensors), regulations (especially around safety and accident liability), public acceptance (particular for ride-sharing AVs like shuttles), and collaboration (mobility industry need to help cities better understand digital components, software, and data sharing). 


Personal comments

This year-long study by BCG and the University of St. Gallen is surprisingly in-depth and sheds light on some interesting aspects of the AV future based on simulated data. There has been a lot of discussion in the AV literature about whether AVs will make congestion better or worse – so it’s important that this simulation took into account 3 different types of AVs. The results clearly show that the different AV types can lead to dramatically different results – for example, in all cities robo-pods will worsen congestion, while much larger robo-shuttles carrying up to 15 people is likely to be the optimum way to go for 3 of the 5 city types. 

If anything, this study’s most important message to cities and mobility practitioners should be to stop thinking about the impact of AVs as a broad, undifferentiated vehicle class. When they arrive, AVs will likely come in a range of vehicle types and the modal mix between these AV types could have just as large an impact on mobility outcomes as the modal mix more generally between AVs, microbility, public transit, and private vehicles. 

Lastly, the study provides a good starting point for different cities to think about how their particular traits such as age, population density, congestion, urban street pattern, journey times, and topography may affect what the optimal modal mix may be for them – and that some cities would actually function better with public transit and micromobility rather than AVs of any type.    


Written by Bobby Chen, RISE Mobility & Systems.



1. 2020-07-08. Can Self-Driving Cars Stop the Urban Mobility Meltdown?