JornayaTrusted Form

Upgrading Urban Mobility: How Self Driving Vehicles Will Change Traffic

Deliver Your Vehicle Stress Free

Compare our rates to or from: any location in the world

  • Deliver Your Car Stress Free and Have Peace of Mind
  • Licensed, Bonded and Insured Carriers Near You
  • Award Winning Service, Always on Time

Get instant quotes using our form, or call:

Auto Shipping Quote Calculator
Upgrading Urban Mobility: How Self Driving Vehicles Will Change Traffic
Upgrading Urban Mobility: How Self Driving Vehicles Will Change Traffic

The following article analyzes what changes can occur if a self driving and shared vehicle fleet is used on a large scale in a medium sized city. We will explore 2 concepts for self driving automobiles which we will name “AutoVot” and “TaxiBot”. The purpose of AutoVots will be to transport individual passengers one at a time. On the other hand, TaxiBots can transport multiple passengers at a time. Both types of cars drive themselves.

For this to work, the resulting mobility system created with AutoVots and TaxiBots should provide as many trips as today’s means with the same or better timing, destination, and origin. Also, the system must replace bus and car trips. We will study the impact this system will have on fleet size, travel volume, and requirements for parking during peak hours and across 24 hours.


  1. Using just 10% of the vehicles we use today, we can provide about the same level of mobility. A city’s cars can be 90% removed if we introduce public transport of high capacity and TaxiBots. Even in the worst case scenario, almost 80% of cars can be reduced (if AutoVots don’t have a high enough capacity for transport).
  2. Auto travel volume will increase overall. We can expect a 6% increase in miles travelled when introducing a public transport system with a large capacity as well as TaxiBots. Not only will the system replace taxis and private cars, but also buses. If a public transport system with a high capacity is absent, then AutoVots can increase traveled miles by up to 89%. This can be achieved by servicing and repositioning trips that the current public transport system is handling.
  3. Congestion will be affected depending on how the system is configured. At peak hours, a transport system using public transport with a high capacity combined with TaxiBots can reduce the need for automobiles by up to 65%. Even AutoVots in the absence of public transport can still reduce the amount of cars used during peak hours by up to 23%. 

But, the amount of miles travelled in these periods would be higher by around 9%, in a scenario where public transport of high capacity and TaxiBots are used, and by 103% if AutoVot vehicles are used without a public system with a high capacity. The second option would likely not be feasible.

  1. A lot of private and public space will be obtained due to needing fewer parking spaces. Fleets that drive themselves would not require on street parking. The amount of space this scenario would free up can exceed 200 football fields in some cities. Furthermore, off street parking can be reduced by almost 80%.
  2. More automobiles can be replaced by sharing rides with TaxiBots than with AutoVots. To deliver the same mobility service, a system based on TaxiBots would need fewer vehicles than one based on AutoVots.
  3. Public transport’s availability will influence the fleet’s size. If public transport of high capacity is not provided, more AutoVots and TaxiBots will be needed (by 25% and 18% respectively). Miles traveled would also increase by 24% and 13% respectively.
  4. It will be difficult to manage this transition. In a scenario where half of all auto transport is done with self driving shared automobiles while the other half is handled with combustion vehicles, the amount of auto travel can increase by up to 90%. Whether public transport with a high capacity will become available or not will not matter. Considering peak hour traffic, we’ll need more cars in most scenarios, with the exception of one where public transport with a high capacity as well as TaxiBots are used.


We can change the current public transport system with self driving automobiles. Traditional means of public transport will no longer be necessary. There will be a significant impact on mobility in urban spaces caused by shared self driving auto fleets. The fleet’s size and type, the amount of emissions, congestions, car travel, as well as how shared vehicles and public transport will mix, will be influenced by transport policies.

Cities will get back a lot of space when shared automobile fleets are introduced. To fully reap the benefits of this extra space, active management will be needed. Among the management strategies used, allocating these spaces to specific recreational and commercial uses can be implemented. Enlarged footpaths, bicycle tracks, delivery bays, and even distribution centers can be constructed.

Though the amount of auto travel will likely increase, the severity and number of crashes should be reduced significantly. The impact on the environment will still be tied to emissions released per mile, so less polluting and fuel efficient technologies will be necessary.

AutoVots and TaxiBots will function for 12 hours each day and travel almost 124 miles. Currently, private vehicles travel for around 50 minutes per day and travel about 19 miles. However, this much more intense use can lead to shorter lifecycles for vehicles and the faster adoption of cleaner technologies.

The business models of car manufacturers will have to change drastically due to the reduction in car sales. Due to less cars being needed, car manufacturers will have to develop new services. However, how to monetize and manage them is currently unclear. 

To maintain market barriers or guide developments, fiscal and regulatory authorities will have to take charge. A service which manufacturers could provide are new maintenance programs specifically designed for the novel vehicles.

The current public transport and taxis will face great competition from the fleets of self driving automobiles. These fleets may replace them as a high quality, low capacity form of public transport. Labor issues will likely result from this situation. 

However, these new services can be administered by the current taxi companies and public transport operators. We will need to adapt the transport services’ governance, including arrangements and concession rules.

If fleets are mixed together, the amount of travel done with vehicles will increase. Furthermore, in ¾ scenarios for peak hours, the amount of cars will increase as well. We can mitigate the congestion with the traffic flow improvement automated cars will bring.

But, to make a case just for self driving vehicle fleets, in the absence of public transport with a high capacity, can be difficult because travel volumes will increase despite the congestion and space benefits. 

Still, even when scenarios are mixed, self driving shared auto fleets can replace current public transport means in a cheaper fashion if we can mitigate the additional travel’s impacts. It will probably be easier to fully deploy such fleets in circumscribed spaces like islands, campuses, business parks, or cities where the rates of motorization are low.


In this article, we’ll examine what changes and potential outcomes in terms of urban mobility would come from the use of a fully autonomous and shared fleet of vehicles.

For this purpose, we’ve created an agent based model which we’ll use to simulate the system players’ behavior. We will analyze travelers which will likely use the new mobility system. We will then analyze the cars that will be routed dynamically on a road network in order to transport clients, and/or move between stations. Finally, we will discuss an efficient dispatcher system which will send automobiles to clients while maintaining high quality standards for the service in terms of detour and waiting time.

Self driving and shared car fleets

As far as assets come, vehicles are underused. Their active period is most often around 10% of a day, during peak hours. Most vehicles are used for an hour or less each day. Because few people travel in them during trips, their capacity is underused as well. 

Yet, despite these problems, their values as assets are very high, to the point where people accept these inefficiencies just so they can benefit from schedule-less, door to door, comfortable travel. However, can these benefits be maintained while reducing inefficiencies?

We will study where shared auto transportation services converge, such as self driving automobiles and car/ride sharing. The later used to involve ad-hoc and informal sharing (car pooling, household vehicle sharing, etc.). However, during the 1980s, commercial cars based on cooperation appeared. 

Through these types of vehicle sharing, people could reserve automobiles that were part of shared fleets, and then use them only when necessary. These services were calculated based on mileage or hours. Being something between a taxi and a car rental operation, they’ve become popular in cities, since people who didn’t own cars could use them. 

Through the internet and services based on apps, sharing vehicles is now more sophisticated and popular. Many such services have appeared across the globe. However, as far as technological sophistication is concerned, car sharing services have gone through an analogous development, especially on demand and app based services. Billions of dollars have been earned by companies which pioneered car sharing based on apps as they proved to be extremely popular.

A driver is currently needed for all these services, so what would happen when self driving cars are introduced? This exercise doesn’t have to be theoretical. 

Uber and Google have implicitly and explicitly signaled that they’re very interested in autonomous and shared automobile fleets both in terms of ride sharing and car sharing. Furthermore, the impacts self driving and shared automobile fleets will have in various situations have been examined by several researchers.  

Mixing traditional cars with shared autonomous fleets

Kockelman and Fagnant developed a scenario for an SAV (shared autonomous vehicle) fleet in a city such as Austin, Texas. The characteristics of their model are:

  • The SAVs must move without the intervention of humans (autonomously),
  • They must transport a minimum of 1 passenger.
  • Between the destination and origin points no stops are made to pick up other passengers, so the trip maintains its course without deviations.

After a trip is completed, the shared autonomous vehicle will go to another client or move towards an area where future passengers can be picked up faster and the cost of parking is low. Because the SAV will travel to customers, there will not be a need for fixed stands from which transport can be started. 

SAV petrol fuelled sedans will comprise the fleet, as no alternative fuel, electric, or hybrid vehicles were considered for the model. Kockelman and Fagnant concluded that just 3.5% of transports will be done using SAVs, while the rest will be carried out by automobiles with human drivers.

Up to 41 people would be served each day by an SAV, and the waiting time will be around 20 seconds. An SAV can replace up to 12 normal automobiles and eliminate the need for more than 10 parking spaces.

The distance people will travel will increase by 11% as opposed to a fleet owned and driven by humans. The cause for this travel distance increase is because of the distance needed to pick up another passenger and the SAVs relocation. 

A SAVs impact on the environment will be positive. Greenhouse gas emissions will be reduced by 5.6%, carbon monoxide by 34%, and organic volatile compounds by 49%. This is when compared with the light duty vehicle fleet currently in use in the United States. 

If SAVs are used more intensively, then we can reduce emissions even further, though at the cost of a shorter usage period (up to 2 years per vehicle). But, this too can help with environmental protection, since old cars will be replaced faster by less polluting alternatives. If we use electric vehicles, emissions will be reduced even further.

Such modeling exercises have a few limitations, such as no context in the real world. Models of the future must take into consideration the real geographical characteristics of specific urban areas, so the results will be more precise. They must include in their considerations weekends, seasonality, travel patterns, and the use of heterogeneous land. 

Among other changes we can include are car pooling, reducing the distances being travels, and the related impacts on the environment. In this article, we will not measure the congestion impact of SAVs.

Singapore automated mobility

Spieser, in a study from 2014, asked what would happen if Singapore’s entire fleet of private vehicles were removed and replaced by self driving cars. According to this study, around 66% of Singapore automobiles could be removed while maintaining the same number of trips. According to the author, autonomous driving would provide numerous benefits like less space needed for parking, costs lowering, less congestion, optimized and more convenient trips, higher safety, etc.

Although the study covers self driving shared automobiles, it’s believed by the authors that their findings can be applicable to general situations like human driven shared vehicles. But, an AMOD (automated mobility on demand) system is believed to be the most time and cost effective alternative. 

The model that uses self driving shared vehicles is believed to be almost half as expensive as the alternative. The only downside is that due to increased travelled distances, congestion and travel time may not decrease.

New Jersey autonomous taxi systems

Implementing ATaxis (autonomous taxi fleets) was modeled for New Jersey by Zachariah et al. according to trips based on travel surveys from origin to destination. They served as approximations of trips people made in the city on a daily basis. 

According to the model, people would get on an ATaxi at a station, and then get off at the nearest station to their destination. The trip can be joined by other people if they want to go to places near the first user’s destination.

High potential for ride sharing was suggested by the results. The more time is spent at the station, the more the vehicle will be occupied. If the passengers and the destinations are close, then the ATaxi’s occupancy will increase as well.

There are spatial and temporary variables to passenger demand. During peak hours, there should be a higher demand for the service. The system can also be used to reduce pollution and congestion in areas with heavy traffic.

New York City taxi pooling

Santi et al.’ modeling work analyses what impact would taxi ride sharing have on New York City’s taxi fleet. To do this, the destination, origin, and timing for each taxi trip are noted. The study covers a full year and checks which trips can be shared, since customers are going between similar locations at similar times.

The construction of a shared auto fleet allows the model’s trips to have delays of at most 5 minutes. Using this type of taxi system, we may notice a 40% reduction in the miles driven by New York taxis. As a result, service costs would be reduced considerably, along with emissions, congestion, and fares people have to pay. In conclusion, a New York City taxi service which can be shared with others is an efficient and possible solution.

According to the study, although the model took into consideration 13.000 taxis and 150 million trips taken in New York, cities up to 75% smaller can use it as well. 

One thing of note is that passengers’ behavior was not considered. They may use the system more due to the lower fares. The market’s potential segmentation was also not addressed fully, specifically the high end for single party or passenger trips and the low end with shared rides.

3 regional cases for changing personal mobility

At Columbia University, Burns et al carried out this model by examining a centrally dispatched, self driving, shared vehicle fleet in 3 environments: a U.S. city of a medium size (Michigan, Ann Arbor), a suburban development of low density (Florida, Babcock Ranch), and a densely populated and large urban area (New York, Manhattan). 

The model covers average distances of trips using data based on surveys, as well as travel speeds and rates to provide characterizations for travel in the studied regions. To calculate automobile requirements and travel patterns, simulation, network, and queuing models were used.

Using the modeling system, trips are generated that will then be serviced using self driving shared automobiles through centralized dispatch centers that track each vehicle’s locations. The destinations and origins of trips are randomly generated across the entire area. An average rate is used for requesting trips, the time between all requests is distributed, and the model’s vehicle class operates at the same speed of travel.

Ann Arbor’s 120.000 citizens travel 70 miles each day or less. According to the study, a shared fleet can provide almost instant access to transportation with just 15% of the automobiles that are currently necessary for these trips.

But, travel time will likely increase because cars will need to be repositioned. The case study at Babcock Ranch revealed similar findings. Up to 4000 automobiles would be needed for 50.000 people. When it comes to Manhattan, according to the study, the current 13.000 taxis can be replaced by just 9.000 shared self driving cars leaving the waiting time at below 1 minute. This is a vast improvement to today’s situation.

A study in Columbia analyzed the perspective of consumers on costs from all 3 contexts. Along with trip making and travel rates, they increase. But, they do differ according to the type of car used, either self driving shared cars built for this purpose or traditional mimic cars. 

When it comes to Ann Arbor, the combined time value, parking, operating, and ownership costs associated with finding parking and driving are estimated at $1.60/mile for a personally owned conventional car driven across 10.000 miles each year. 

As an alternative, a self driving and shared service with conventional cars costs $0.41/mile. With a 2 occupant, small, and purpose built vehicle, costs are further reduced to $0.15/mile. These reductions are significant.

Self driving shared fleets may be more expensive in the case study conducted at the Babcock Ranch ($0.46/mile), as well as in the study for Manhattan ($0.50/mile). However, these instances are still less than what a conventional automobile fleet would cost.

The relevance and interest in autonomous and shared mobility services is accentuated by these publications. It’s also noticeable that this area’s findings are properly aligned suggesting great potential for lowering the number of parking spaces and automobiles required without limiting services and mobility. 

Lisbon city case study

To give a reference frame for the model, we need a view of the mobility patterns in the test city as well as its various characteristics. Of course, because the exercise’s results will be embedded firmly in this unique context, the results’ transferability will be constrained. Still, the exercise’s derived insights will indicate the direction and scale of the impacts we can expect from deploying the proposed upgrade in a city’s mobility system.

We will base the simulation on data obtained from Lisbon. It is Portugal’s capital and the country’s largest city, holding more than 550.000 citizens in an 84 square km area. The city is located in the middle of the LMA (Lisbon Metropolitan area), which is home to almost 3 million people or around 25% of the country’s population. 

There are around 5 million trips in the LMA every day. Around 55% of them are commutes to or from work/study. From the LMA’s total activity, around 1.2 million trips occur within the municipality of Lisbon. In this model, we will discuss these trips specifically. 

Lisbon municipality’s inhabitants show a somewhat low rate of car ownership. For 1.000 citizens, there are 217 cars. The number of trips each inhabitant takes every day is 1.9. In the city’s heritage areas, there is little space for parking, which makes people not want to own a car. 

The population of the city centre is related to the low number of trips each day. The demographic in city boroughs consists of 55% elderly people that are older than 65 years of age. Still, these numbers are somewhat similar with other cities in Europe when considering the per capita GDP. 

The capital of Portugal uses an important underground network. Workers and inhabitants use it in Lisbon’s municipality and in the whole metropolitan area. In 1959, it was inaugurated and it’s been growing since then. Currently, a large part of the city’s area is covered by it, and there are suburban area links as well. 

The Metro is a network of medium size made out of 4 lines and 52 stations that covers around 43 km and transports more than 175 million people each year. Furthermore, the greater Lisbon area is connected with the city’s centre through 4 rail lines meant for commuters. Because in the city centre there are just 13 stations, the lines’ relevance for our study is small.

According to newer data, around 2.000 taxis, 400 buses, and 60.000 cars travel at peak hours at the same time in Lisbon. As a result, there are around 60 automobiles on each km. of road. It’s a relatively high density especially if we take into account the numerous narrow streets that comprise a large portion of the network.

The city’s recent value shows around 160.000 automobiles parked at the same time with a 78% utilization rate. Because of the constraint on parking, the rate of car ownership in the city centre is lower, so the distribution is more balanced than in the entire metropolitan area. Although the newest data shows private cars are used for around 60% of trips in the metropolitan area, the percentage is 40% in the heart of the city. There, around 20% of trips are done through walking.

Describing the model

A model based on the agent was developed to simulate each day’s operations in a hypothetical Lisbon shared mobility system. The basis for the model is real trips, and the actual road networks of Lisbon are used for the simulation. 

The model’s setup allows self driving and shared vehicle fleet to provide the necessary trips for a synthetic full scale Lisbon population based on our travel survey. These trips will be generalized in a 200 by 200 m. cell grid. 

Serving mobility requests will be managed by a dispatcher using the automobiles’ locations, the clients’ locations, and the cars’ occupancy level. The trip routing is estimated based on an algorithm which provides the path with the lowest price between node pairs within the network.

Vehicle and client interactions will be addressed in the model. We will not add a dynamic model of traffic to simulate automobile level interactions between them and the environment. 

The approach in this article uses a static traffic environment. The flows between origin and destination will be assigned to a topologically correct, simple representation of a road network that takes occupancy per link into consideration.

Generating demand

Using the Travel Survey from Lisbon, we’ve developed a trip population for the city collected into grids. The simulation model we used for synthetic travel was created and calibrated in past studies for the LMA. Contained in its output are the destinations and origins of trips allocated spatially at the level of the census block. We’ve also included their timing during a weekday.

Characterizing each trip is done not just according to occurrence time, destination, and origin, but according to the person’s age and the trip’s purpose as well. The considered modes are suburban or underground train, cycling or walking, and shared vehicles (car or ride sharing). For all of them, the trip’s characteristics are travel time, waiting time, access time, and when applicable transfers between various grids.

An approach based on rules was adopted to mention a restricted and simplified choice process. Trips under 1 km. were taken by bicycle or foot. The remaining trips were assigned either to the self driving and shared mode of transport or to the underground.

When the destination and origin points were near underground stations and just 1 transfer was needed for the completed trip inside the underground network, the trips were placed under that category.

Trips that couldn’t be served properly by walking or by the underground network were left to the self driving and shared fleet. If underground services weren’t available, any trip longer than a single km. was given to the self driving and shared mobility option.

When the self driving and shared option is chosen, we generate a new agent/user in the environment that has its own starting time, as well as departure and arrival nodes. At the trip’s beginning, users don’t form parties, but they can share automobiles afterwards.

User trip generation

When someone requests transportation between 2 points in the environment we’ve simulated, a trip is generated. Simulation parameters (car and ride sharing) are accounted for in the model, as well as arrival time, detour time, and waiting time. Afterwards, the dispatcher discovers the best route and assigns an available automobile type in ride or car sharing mode.

Then, users must access a location to board the automobile or wait for his car. When they arrive at the destination point, the user leaves which generates an indicator for the trip log that will be used in the system evaluation.

Vehicle configurations

Idle automobiles are placed in 60 stations within the model city. If an automobile isn’t carrying passengers or is dispatched towards one, it will park itself or relocate in a station. 

We assume that 3 types of automobiles are included in the fleet according to their passenger capacity. They will either transport 2, 5, or 8 passengers. Propulsion technology was not specified since it’s not significant for the study. Energy use and emissions were not modeled. However, we did check the system’s sensitivity to the additional recharging time as well as what electric vehicles would require.

Mobility dispatchers and their roles

The primary role of dispatchers is to set the rules with which users and cars will be matched and to centralize the necessary information to monitor and produce trips. To match a user with a car, he/she must consider the principle of minimizing time which applies to the current users and to the ones underway.

Several parametric constraints were defined regarding what must be done for every route the dispatcher proposes. These runs include requirements like the delay for starting trips should be at most 5 minutes. 

The number of people which can use one car is also limited to 8. When adding additional passengers, the time increase must not be higher than 20% of the original trip or 10 minutes maximum. The distance must be at most 20% higher or a maximum of 2 km.

Therefore, a 10 minute trip would at most take 17 (5 for waiting and 2 for travel time).

However, in reality the travel times for self driving and shared automobiles are most often lower and rarely surpass trips from the base scenario. Also, the time needed to find parking spaces and to enter the automobile is not accounted for in the base scenario. 

Shared mobility tests

Several scenarios were devised in order to deploy self driving and shared vehicles in Lisbon. We generated these scenarios through 4 main parameters:

  • The self driving and shared operation mode, ride or car sharing.
  • Whether public transport with a high capacity is available or not.
  • The self driving and shared fleet’s penetration rate, 50% or 100% of trips.
  • What period of time we took into account for the model: a weekday’s entire 24 hours or just peak hours.

For our fleet, we checked 2 system configurations formed of self driving automobiles:

  • The ride sharing system through which space and time resources are shared by using the same vehicle to its capacity. The automobiles can be owned privately by a rider or be part of a company’s fleet. This will be known as a Taxi Robot or TaxiBot.
  • The car sharing system through which people use cars sequentially and therefore only share time with each other. Here, fleet managers would normally own vehicles, though experiments from peers may be present. This system was named Automated Vehicle Robot or AutoVot.

We decided to include the option of not having a transport system with a high capacity because we wanted to know to what degree can the self driving and shared options take on these trips. In our scenarios, we used the underground system as the public transport with a high capacity. But, other solutions can be used too, like BRT, commuter rail, or LRT if their station density level is similar to Lisbon.

Scenarios where the self driving and shared fleet’s penetration rate was 50% were also modeled. Here, we assigned 50% of trips that were not handled by through public transport with a high capacity to conventional cars operated traditionally. The other 50% belonged to the self driving and shared fleet. Scenarios were executed using generic distance, detour time, and waiting time, as well as the automobile capacity constraints we’ve mentioned above.

We performed a simulation of a normal day to check the outcomes in terms of mobility for different scenarios. While results may show variability from one trial to the next, there is high stability for our main indicators.

Different scenarios and modal shares

Regardless of scenario, the shares we’ve obtained from the modal choice model were stable. According to results, there is an increase in the use of public transport with a high capacity in comparison to Lisbon’s current travel pattern. 

The cause of this deviation is bus users that opt for the underground even if they prefer buses. The threshold of 1 km we’ve set aside for walking is pretty low, since the walking share in this city is much higher than in others.

If every motorized trip is done using public transport without a high capacity, AutoVot and Taxibot fleets will have a 70% share, when public transport with a high capacity complements them, and a 92% share in their absence.

Fleet size impact

There is a high potential for self driving and shared fleets to reduce the needed vehicles in today’s fleet. When using a TaxiBot configuration with public transport of a high capacity and a 24 hour weekday model, 90% of automobiles may be removed while maintaining almost the same mobility level when it comes to trip length, destinations, and origins.

More striking is the fact that we would need just 5.000 self driving and shared TaxiBots to complete the trips we’ve modeled currently for the Metro in Lisbon. The potential to reduce the fleet is lower for AutoVots because more cars are needed as well as repositioning. 

Without public transport of high capacity, more cars will be needed compared to the base case in the 50% AutoVot and Taxibot penetration scenarios. Even when the underground is absent, private car numbers will only be reduced slightly. Therefore, the fleet reduction levels we expect may not be achieved in transition scenarios, at least not in the beginning stages.

Travel volume impact

There is high potential for self driving and shared fleets to lower automobile numbers in cities. However, when it comes to car travel, we cannot say the same. 

In every studied scenario, there’s an increase in distance travelled. When using TaxiBots, the number of km travelled by car every day rose by 6%. When using no public transport and only AutoVots, there was an 89% increase in km. travelled.

Several factors can explain this increase. We’ve assumed in every scenario that self driving and shared automobile fleets will replace travel by bus in Lisbon. Our base scenario shows 20% bus occupancy per day. These citizens will likely be served better with self driving and shared automobiles. 

Around 30% of auto km. are accounted by diverting bus passengers in the scenario where rides are shared. 50% are diverted in the scenario where cars are shared. Regardless of choice, repositioning empty automobiles accounts for the remaining increment. In scenarios where rides are shared, passenger pickup detours also play a factor.

The car travel increase is significant when using mixed fleets with conventional and self driving shared cars. When using AutoVots and no public transport, the increase is by around 90%. 

In every scenario we’ve considered, peak hour car km. increase as well. The increase is of 9% when using public transport and TaxiBots. However, there’s a 103% increase when AutoVots are used in the absence of public transport. As a result, this second scenario would simply not be manageable.

There isn’t a uniform increase in car km. at peak hours on various road classes or in the city. We studied the situation according to day time. The base case, as well as AutoVot, and TaxiBot scenarios were taken into account. Furthermore, the base case covered trips done by taxis, motorcycles, and private cars. 

But, travel was no accounted for, because we did not have car km spatial allocation available. Because 13% of car km. in the city is travelled with buses, this makes a difference.

According to our research, when using public transport and TaxiBots, there is a small rise in the volume of travel which does not add too much during peak hours except on Lisbon’s local road network. 

Here, the increase is caused by drop off and pick up movements which are more common during morning peak hours. We’ve also noted that on every road network there was some degree of peak travel as a result of relocating vehicles to stations when the periods of peak demand ended. 

From an operational point of view, this result is relevant because it suggests that we’ll need strategies to deal with this relocation to prepare for the following peak period. Unlike when using TaxiBots, AutoVots in the absence of public transport brings a travel volume increase combined with a shift and spreading on every road network of peak periods.

Since AutoVot and TaxiBot scenarios increase the amount of trips on local roads, the road network’s characteristics and performance may change. When it comes to performance, the peak hour road occupancy is no higher than 40%, not even a half of the one for the road network. 

In the scenario where public transport and TaxiBots are used, road occupancy at peak hours doesn’t change except for local roads. But, even here the increase is just 23%.

The situation is different when just AutoVots are used.  Here, the increase of road occupancy is up to 50% regardless of road class. 

The biggest growth is noticed on local networks. Due to this, the performance will likely be poorer, and we may have congestions.

Street traffic’s nature may also change due to additional traffic being added to former less used, quiet roads. There can be a bad impact caused by this traffic increase on the livability and attractiveness of local areas, since local streets will not be as available for activities aside from transport.

Certain city areas and some roads may have lower traffic at peak hours when using TaxiBots, but other areas will have a small increase. There are 2 main findings on TaxiBots’ impact:

  • At some bottlenecks, the impact on congestion will be slightly negative.
  • Overall, the fluidity of traffic will be maintained. 

In local networks, traffic will increase when it had previously been absent. The traffic increase may clash with cycling and walking. We should also mention that places with the highest traffic volume increase are those who are least connected with public transport of high capacity. As such, these 2 systems seem to complement each other.

Peak hour car fleet requirements

To verify how AutoVot and TaxiBot fleets function during peak hours, we assessed how many cars travelled within Lisbon at the morning peak. We did not include the link level capacity limitations in parts of our study, but it did show us travel volumes and requirements at peak hours and allowed us to compare them with the base scenario.

It’s important to note the amount car reduction thanks to self driving shared fleets during these periods. If such fleets are implemented, the amount of cars on the road during peak hours would be far less while maintaining service at a similar level to the baseline. 

There were 65% less cars when using the public transport and TaxiBot scenario. Even with AutoVots in the absence of public transport, we needed 23% less automobiles.

Our tests of scenarios where mixed fleets were used indicate that systems based on ride sharing can reduce peak hour traffic if combined with public transport. There may be a difficult to absorb traffic increase if relocation operations of automobile sharing systems increase.

There may be transition issues for self driving shared fleets because of legacy fleets. If we do not manage peak hour travel increases, we may find it hard to create a policy case for shared self driving automobiles solely based on freed congestion and space while maintaining conventional cars. However, even in such situations, self driving shared cars can be a cheap alternative to our current public transport system if we can mitigate the additional travel’s impact.

The traveling ratio at peak hours is another important output of our scenarios. It’s the ratio of cars making trips at 3 p.m. and 8 a.m. According to this ratio, sharing automobiles with an AutoVot system may increase the number of automobiles currently running during peak hours as opposed to a ride sharing system using TaxiBots. When taking into account scenarios with mixed fleets, the traveling ratio increases leading to a larger concentration of automobiles at peak hours and high congestion during these periods.

Impact on street space and parking

When parked and when moving, a lot of city space is taken up by automobiles. Therefore, we checked the effect self driving shared cars would bring on parking. According to our research, the potential for reducing off street and on street parking spaces is huge for all scenarios where a self driving shared fleet is used.

In our baseline case, there are around 50.000 spots for parking off street. However, even in the scenario which required the most parking spaces (no public transport and just car sharing), only 25.620 spots would be needed. Furthermore, regardless of scenario, on-street parking would not be necessary.

This means that more than 1.500.000 square meters could be allocated to public use, more than 200 football fields, or 20% of Lisbon’s street area. We can then dedicate this space to commercial or recreational uses, parklets, delivery bays, cycling, walking, etc.

In the best parking scenario, we would need at most 8.900 spots. But, the results aren’t as good if we use mixed fleets of traditional cars and AutoVots/TaxiBots. There would be a 25% reduction in needed parking spaces if we use TaxiBots, but the result would be negative when using AutoVots. According to these findings, self driving and shared auto fleets that work alongside conventional cars will require bus services to keep parking requirements low.

Vehicle use impact

Self driving and shared fleets’ impact on making vehicle operations more efficient is another aspect we should consider. Vehicle down time can be reduced significantly with their use. At the moment, Lisbon cars operate for around 50 minutes each day, meaning they sit in a parking sport for 95% of each day. In every scenario we’ve considered, car idle time was reduced to below 40%, while in the best cases it was at 27%.

Vehicle type distribution

We followed the car type’s distribution to verify outcomes. According to results, the self driving shared fleet will be composed in large part of 3-5 passenger automobiles and of small cars that transport up to 2 passengers.

According to 2009 Lisbon car sales, many people living within the city own cars. Because of this, it’s likely that models or makes on the market, as far as car size is concerned, will not differ dramatically when self driving and shared cars will be implemented.

Electric car fleet impact

We mentioned previously that self driving shared cars will increase travel volume. To mitigate the pollutants and greenhouse gas emissions, we would need to deploy energy efficient low emission drivetrain technologies. 

Electric automobiles may even replace tank to wheel cars in these scenarios. However, current technology limits the range of electric cars compared to those running on fossil fuel. They would also need extra time to recharge.

To check what impact smaller travel range and re-charging times would have, we added more car requirements for self driving shared fleets made out of electric cars. A speedy recharge time for a battery was put at 30 minutes and car autonomy was considered 175 km. However, the fleet size impact when deploying self driving and shared electric cars is small (just 2% higher).

Travel and waiting time changes

After calculating the average travel and waiting time caused by shared mobility services with AutoVots and TaxiBots, we compared them to the base case for private cars and public transport. Average travel and waiting times were reduced significantly. The cause for these reductions are the personalized services AutoVots and TaxiBots offer, especially when compared to bus trips, and thanks to their superior travel times even during peak hours.

If we increase the waiting time to 10 minutes, the car fleet can be reduced by 11%. In these situations, we can reduce travel volume by just 3%, because the extra time would be needed for dropping/picking up people from locations further away. 

However, the party size of ride sharing vehicles would increase by 27% with the higher time constraint. Therefore, this configuration can reduce the price of the service if it were created according to the number of people using each vehicle. In most cases, when using public transport and TaxiBots, the waiting time would be less than 4 minutes. 

The dispatching algorithm has 2 branches: 

  • One related to the initial assignment of people to a TaxiBot which takes around 3.3 minutes.
  • One related to connecting a client with a partially filled car which would take around 4.2 minutes.

In our base scenario, travel times for door to door services for underground and bus users had to include the time it takes to get to the station or stop and also time spent waiting for transport. We did not add any access or waiting time to car travel although parking spaces may not be present at the arrival or departure points. We also did not include time for finding parking spaces.

Due to this, the base case we’ve used underestimates travel times. Public transport users will likely see an improvement in their travel time when using AutoVots and TaxiBots, while a portion of car drivers will notice an increase in travel times. Therefore, travel time impact will depend on how many car drivers there are compared to users of public transport.

Except for AutoVot trips at peak hour, travel time is reduced when using car and ride sharing systems. When using a TaxiBot system alongside public transport, we can reduce the time needed for trips by 30%. When using AutoVots in the absence of public transport, the reduction is 18%. 

While on average, both AutoVots and TaxiBots save time, this may not apply to all travelers. For example, traveling with AutoVots at peak hours would require more time. 

However, if car drivers switch to AutoVots, their door to door transport time will be reduced by around 2%. This is assuming it would take them around 3 minutes to find parking spaces for their vehicles. However, when switching to TaxiBots and public transport, car drivers would spend around 8% more time traveling door to door.

Distributing time values and detour distance

As explained above, when using TaxiBots, passengers will have to make small deviations from the shortest path to their destinations. However, this additional time might not be a problem even at peak time.

Matching efficiency and occupancy levels

Ride matching efficiency was investigated, as well as the average occupants’ number, and how ride sharing using TaxiBots was influenced according to demand intensity. A TaxiBot’s average party during the day showed demand peaks in the early evening and morning that improve the vehicle’s ability to gather large groups of clients.

During times of low demand, like in the early morning and late evening, in popular areas the demand was concentrated and offered satisfactory occupancy averages. These areas can be neighborhoods with many clubs and bars.

It’s important to note that the origin point boarding concentration that helps prevent big waiting times was a decisive component in providing high occupancy as opposed to the locations’ densities. 

The party size’s daily elasticity was 1.07 when it comes to city demand. Therefore, when the average demand increases by 1%, the party size is like to increase by 1.07%.

We did not check the costs of self driving shared mobility in urban areas nor the price travelers pay for each trip in the base case. Therefore, we have an agnostic model when it comes to the welfare impact of our scenarios. 

Real life travel demand is greatly influenced by costs. Therefore, further work is needed to check the extent to which travel demand may increase through reducing costs and travel time. More demand may reduce some benefits of deploying self driving shared mobility systems in urban areas. 

We also did not consider labor problems which should be significant. Equity questions were similarly not addressed.

Insights into policy

The size and type of a vehicle fleet can be influenced by transport policy, as well as mixing shared vehicles and public transport. Emissions, congestion, and car travel in urban areas are also subject to their influence. In medium or small cities, self driving and shared automobiles may eliminate the need for public transportation.

A large amount of space off and on the street is cleared up when using shared automobiles. But, to obtain the most benefits, this space must be managed proactively. 

Crash severity and the number of crashes will likely be reduced when using self driving cars on a large scale despite car travel levels. Per km. emissions affect the environment, so the level of pollution will depend on how non-polluting or fuel efficient the self driving shared fleet will be.

Deploying self driving shared fleets in cities will cause competition with public transport and taxi services. They may become a type of high quality, low capacity public transport. Issues with labor will be large, but taxi companies and operators of public transport can deliver such services actively. 

As such, these issues can be somewhat alleviated. Transport service governance, including arrangements and concession rules, will have to adapt. 

Reducing vehicle numbers due to shifting towards self driving shared automobiles will impact how automobile manufacturers conduct their business. Their introduction will require new services, but who can monetize and manage them is unclear. The authorities will play an important fiscal and regulatory role in maintaining market barriers or in guiding developments.

Throughout the scenarios we tested, the use of automobiles was more intense. There was an increase in daily use from 30 km and 50 minutes to almost 200 km and 12 hours. Different automobile models will be needed due to this use increase.

Car lifecycles will be shorter, so cleaner technologies will be adopted faster. Robust and different interior fittings will be needed for shared use, although there may be a chance to reduce crash risks by saving weight. New maintenance programs may be included in the way these services will be monetized. 

In general, car travel will increase in all our scenarios, and the number of cars will go up in 3/4 peak hour simulations. Improving the flow of traffic may help with congestion. But, in extreme scenarios, public policy supporting self driving fleets may be difficult to support if it’s based just on congestion and space benefits.

Still, even when mixing scenarios, shared self driving fleets have the potential of being a cheap alternative to our current public transport so long as we mitigate the impact on travel time. Finally, deploying self driving shared automobiles will likely be easier to implement in places like islands, campuses, cities with few vehicles, and business parks. 

Joe Webster
Written By:Joe Webster
Connect with Joe Webster on facebookConnect with Joe Webster on linkedin

Joe Webster began his journey in the auto transport field by attending the University of Southern California (USC), where he graduated with a Bachelor of Business Marketing. 

After college, he started his career in the auto transport industry from the bottom up and has done virtually every job there is to do at A-1 Auto Transport, including but not limited to: Truck Driver, Dispatch, Sales, PR, Bookkeeping, Transport Planner, Transport Manager, International Transport Manager, Brokering, Customer Service, and Marketing. Working with his mentor Tony Taylor, Joe Webster has learned the ins and outs of this industry which is largely misunderstood. 

With over 30 years experience in the industry, we've been helping people ship their vehicles, motorcycles, RV's, heavy equipment, household goods and more across the country or overseas without a hitch. Ask us anything.

Ship Your Vehicle With A-1 Auto Today
Get an Instant Quote

Or Call Us Today and Speak to Directly to Our Sales Team:

Our Partners:

Better Business Bureau Accredited
FMCSA Member
INC. 500 Member
VeriSign Secured
AMSA Member

A-1 Auto Transport is a disclosed agent for the following shipping companies:

Trans Global Auto Logistics, Inc.
3401 E Randol Mill Rd
Arlington, TX 76011
NO. 018191NF
CFR Rinkens
15501 Texaco Avenue
Paramount, CA 90723
NO. 013055NF
ABC Worldwide LLC
2840 NW 2nd Ave #105
Boca Raton, FL 33431
NO. 025472F
CSI Logistics
435 Division Street
Elizabeth, NJ 07201
FMC 22206
Intl Cargo
45 Campus drive
Edison, NJ 08837
NO. 17858N
ShipYourCarNow LLC
1160 South Rogers Circle Suite 1
Boca Raton, FL 33487
NO. 025646
Merco Air & Ocean Cargo, Inc.
6 Fir Way
Cooper City, FL 33026
NO. 021869F
Copyright © 1988-2020, A-1 Auto Transport - All Rights Reserved.