Report of the joint project with Bell Textron Inc. on the feasibility of VTOL as a new form of transportation in urban areas.

In this research project with Bell Textron Inc., we conducted complex simulations and modeling of air taxi transportation. We calculated the number of travelers who can benefit from this opportunity and the placement of (dis)embarking stations for an improved traveler experience. Read more about our findings.

Improving the mobility experience

Tech companies including Bell, Uber, NASA, Volcopter, German start-up Lilium, or the Czech Zuri have lately been looking into the possibility of air taxis as a means of transportation. Their deployment into the existing mobility-on-demand system however comes with a few challenges. The research team of Jiří Vokřínek from the AI Center FEE CTU tackled these issues in a joint project with Bell which was recently finished.

Examples of other VTOL concepts (from top left: Zuri, Lilium, Volocopter Velocity, NASA Air Taxi)

The future air taxis are usually (e)VTOLs as in (electric-powered) Vertical Take-Off and Landing aircraft which indicates that the vehicle can fly from the ground straight up (as opposed to airplanes that need a landing strip to gradually take off in a diagonal manner). Places from which these aircraft take off are called vertiports and their placement is a critical part of the VTOL deployment strategy. Why? The positioning of such stations directly affects the efficiency of this mobility system and the passenger’s decision to use this network to travel across the town. 

However, the customer’s decision-making is much more complex. Factors like travel time or cost will play an important role that can only be properly evaluated through a simulation of the travel selection scenarios. And that’s where AI research comes in. But first, let’s take a look at the bigger picture of transportation in the 21st century. 

Efficient and environmentally friendly solutions

For commuting or other short trips, travelers usually choose from three transport options: private transport like private vehicles, cycling, or walking, public transport, or vehicles for hire (FHV), typically taxis or transportation network companies like Uber. The third option is currently on the rise. In New York City, the metropolitan area we studied in our project, the number of FHV trips rose by 13 % in the last two years before the COVID pandemic (from 870 000 trips per day in 2017 to 1 000 000 trips in 2019). And we can see the same trend worldwide.

Daily trip counts in TLC (the NYC Taxi and Limousine Commission) -licensed vehicles. The black color marks the High Volume For-Hire Vehicle, yellow marks the Medallion Taxi, green the Street Hail Livery, and grey the Traditional For-Hire Vehicle. Source: TLC Factbook (2020)

With the rising share of Uber, Lyft, and other taxi companies in total road traffic, some controversies have emerged. One of them was the impact of these services on congestion. This not only has negative environmental impacts but also negatively affects the speed of travel. The extra time spent in traffic jams can create a new opportunity for a service that offers a shorter travel time for an additional price – the air taxis as the new FHV transportation mode operated by aircraft. So… what stands in the way you might ask. 

Air taxi obstacles

Unlike other vehicles for hire, VTOLs cannot use the dense road network for (dis)embarkation. Instead, the proposed systems use a set of vertiports that also serves the purpose of recharging, refueling, and maintenance. They have to be built prior to the system operation and are expected to be pretty expensive.

Deloitte identified the ground infrastructure as the biggest challenge for air taxis. Thus, careful vertiport positioning has to occur to achieve a good trade-off between the capital cost and the system coverage. In addition, we cannot expect the vertiport network to be dense, therefore, another mode of transport has to be used as a first/last-mile solution

Example of a VTOL plan that is clearly worse than the taxi plan due to the specific vertiport placement. Because VTOL needs to use the vertiports, some transport demand can not be efficiently satisfied by it. 

Furthermore, a stable regulation for VTOL traffic is not established yet in many countries. And finally, there are some technological barriers that prevent the fast adoption of air taxis, the most serious one being the electric power source. The problem lies in the low energy density in the batteries, a factor that is important for electrical cars and crucial for VTOLs. The required range of lower tens of kilometers corresponds to the vehicles that are currently in the experimental stage, rather than on the market. Therefore, some additional progress is required for the air taxis to be a feasible transport option. Nevertheless, the simulations for the time when the technology is ready can be done now – as we did.

Methodology for computing the viability of the VTOL system and the optimal allocation of vertiports

Our job was to explore the feasibility of VTOL as a new form of city transportation. In the project, we had two goals: simulate the decision-making process of customers to determine who and where could benefit from VTOLs and improve the vertiport placement (where should the stations be located) so the benefits to the commuters could be maximized. Now let's dig deeper into each step.

Firstly, we designed the urban mobility model for the New York metropolitan area using various data sources and a traveler choice model. Three sets of real-world data were used for the modeling: Street-Light (travel demand data), HERE API (travel options data), and OpenStreetMap (travel times approximation data). To give you an idea of the size of the data samples, the travel demands dataset consists of 2 405 736 travelers with a unique origin, destination, departure time, and traveler income.

Area covered by the StreetLight dataset (blue), showing filtered zones used in the simulations (yellow).

We came up with a set of metrics to evaluate how likely the passengers are to use the existing travel options (walk, bike, private car, public transport, or taxi)  and how they could benefit from using a  VTOL taxi. Then we were ready to run the simulations.

The conceptual structure of the city mobility model

 In the simulation, we simulate each traveler using either a private car, taxi, public transport, or VTOL. For each of these options, we measure how long it takes them to get to their destination, how much it cost as well as other metrics. Next, we compare these options through the Mode Choice Function, which in our simulations is either preference for faster trips, preference for cheaper trips, and a complex combination of different traveler features and plan properties that try to capture the relative suitability of each plan to the traveler in terms of the traveler budget and comfort. 

Looking at the results of these simulations, we can accurately estimate how many people VTOL would be either a faster, cheaper, or overall preferable (according to our complex function) option. Importantly, we simulated the VTOL transport with a number of different parameters that are currently unknown. For example, nobody knows for sure the actual range, speed, boarding time, and number of people in the VTOL. Especially the latter two depend on the design of not just the vehicle but the whole VTOL transportation system and will be critical to its future competitiveness. 

Road network used for vertiport positioning.

As for the second part of the project, we focused on the vertiport positioning problem for which we propose an optimized solution and evaluated the sensitivity of the transport system to the placement of air taxi stations. Our team analyzed several approaches and scenarios of vertiport positioning from random to clustered and optimized according to their accessibility by the road network. 

Example of the simulation. Here you can see the maps of faster VTOL trips with naive vertiport placement. Shown VTOL trips are faster than the Taxi baseline.

Optimization turned out to be the best option for the positioning as it decreased the travel time over the regular taxi mode almost five times more likely than the baseline clustering approach. And the results?

Results: Faster, greener, and sensitive to vertiport placement

We have shown that the eVTOL transportation system has the potential to decrease travel time for a small section of existing travelers, especially those who currently have to travel around otherwise impassable barriers such as large rivers or bays. 

While the eVTOL is expected to be fast on its own, the first- and last-mile sections of the trips seriously degrade its usefulness for most travelers. As it turns out, most people currently travel to places they can get to quite effectively with at least a taxi. 

However, this is true for the people who travel today. Introducing eVTOL into the cities could generate its own new travel demand between places that are today difficult to travel between. This would be similar to how new rail or metro lines can lead to people moving to towns that would previously be too difficult to get to from city centers.

In the air taxi parameters sensitivity analysis we also show that the boarding/disembarking time can make or break the case for eVTOL as an air taxi. Here, if the airport check-in would be the guiding example, the future might not be too rosy.  The experiments also confirm the strong influence of vertiport positioning on the overall performance of the transportation system. 

The possible directions for future work are immense. To name a few, we can run simulations in different or larger areas, we can incorporate real demand from the NYC dataset into the analysis or some external datasets for pricing too, and we could also consider a ride-sharing model as part of the simulations. The possibilities are endless and we are excited to see where the future (or the air taxis) will take us.