Smart electric vehicles – GreenWheels, RoboScooters, or CityCars – can simply be thought of as appealing consumer products. But they can also be combined with smart fleet management solutions – Dynamic Pricing – to launch new kinds of mobility services – Mobility on Demand Systems – that enable convenient point-to-point travel within urban areas, enable very high vehicle utilization rates, and extend availability to those who cannot or don’t want to own their own vehicles. This category of users includes visitors to a city who generally don’t bring their own vehicles with them, occasional riders and drivers who cannot justify the cost of ownership, those who don’t have anywhere to store a vehicle, and those who don’t want the responsibility and bother of ownership and maintenance.

Large-scale systems employing traditional, non-electric bicycles – for example Vélib in Paris, Vélov in Lyon, Bicing in Barcelona, and Bixi in Montreal – have already demonstrated the feasibility of mobility-on-demand. In these systems, racks of bicycles are spaced around the city such that potential users are rarely more than a short walk away from a rack. In order to make a trip, a user walks to a nearby rack, swipes a card to provide identification and unlock a bicycle, rides to a rack near the trip destination, drops off the bicycle, and walks the rest of the way.

Substitution of lightweight electric vehicles bicycles increases the range and utility of these systems, and makes them useable by more people. Where this begins with GreenWheel bicycles it requires little additional infrastructure, since racks for traditional mobility-on-demand bicycle fleets require power supply and data connection in any case. It is straightforward to upgrade them to provide battery charging as well.

Since acquiring real estate for vehicle pickup and dropoff points, and providing power supply at these points, are key issues in the implementation of mobility-on-demand systems, starting with a relatively simple, low-investment, GreenWheel-based system makes sense. This establishes the foundation for later expanding the system to scooters or automobiles.

Retail location theory suggests that, where pickup and dropoff points are of equal capacity, they should serve equal population catchments. This means that they will be closely spaced in areas of high population density and more sparsely spaced in areas of low population density. Alternatively, pickup and dropoff points might be evenly spaced, at intervals determined by comfortable walking distance, and varied in size according to surrounding population density.

Once pickup and dropoff points have been deployed and stocked, the fundamental management challenge with mobility-on-demand systems is to keep the system balanced. Across the system’s service area, demand for vehicles – as expressed by customers showing up and wanting to pick up vehicles – varies dynamically from location to location and over time. Similarly, the supply of vehicles and parking spaces – as expressed by stocks available at access locations – varies dynamically. The task is to keep supply and demand in

balance, such that customers never have to wait for unacceptable lengths of time for vehicles or parking spaces, and the numbers of vehicles and parking spaces required to achieve this balance are minimized. (This task is closely analogous to the task of load balancing in electric grids.)

The difficulty of the balancing task depends upon the skewness of the distribution of demand in space and time. Where desired trip origins and destinations are randomly distributed, the system can be expected to self-organize – keeping vehicles distributed fairly evenly throughout the service area. But where demand is highly skewed – for example, when it is dominated by morning and evening commutes – keeping the system balanced requires effort and costs money.

One way to balance the system is to move riderless vehicles to where they are needed, for example by loading bikes on trucks. This can be done by setting up the system in the small hours of the morning, letting it become gradually less balanced during the day, and then resetting it the following evening. Alternatively, vehicles can be moved continually – in effect, resetting it less sweepingly but at shorter time intervals. In either case, balancing is easier when there are buffers of excess vehicles and parking spaces in the system to absorb minor imbalances.

A more elegant approach is to exploit elasticities in times and locations of trips, and to manage demand through dynamic pricing. Under this strategy, it becomes more expensive for customers to pick up vehicles from locations where demand for them is currently high, and less expensive to pick them up from locations where demand is low. Similarly, it becomes more expensive to drop off vehicles at locations where parking spaces are currently heavily in demand, and less expensive to drop them off at locations where spaces are less in demand. Price signals thus motivate customer behavior patterns that keep supply and demand in equilibrium. Here the cost of system balancing is not that of moving riderless vehicles, but of providing the necessary price incentives.

All of these strategies require the support of a sophisticated, networked information technology. Both for billing of customers and for monitoring the distribution of vehicles and parking spaces in the system, it is necessary to track vehicle pickups and dropoffs in real time. The system must also compute optimum balancing strategies, and either make price adjustments or send instructions to redistribution truck operators.

Mobility-on-demand systems can and should coexist with privately owned vehicles. Through appropriate standards, and use of appropriate information technology, they can share parking spaces and recharging infrastructure. Such a joint system is likely to be more effective in meeting all aspects of demand, and it facilitates economies of scale in both vehicle supply and infrastructure development.

Principal Investigator
William J. Mitchell, MIT Professor of Architecture and Media Arts and Sciences

Design Team
Claire Abrahamse, SMarchS, Department of Architecture
Ryan Chin, PhD Candidate, Smart Cities, Media Lab
Chih-Chao Chuang, Smart Cities, MS Candidate
Will Lark Jr., PhD Candidate, Smart Cities, Media Lab
Michael Chia-Liang Lin, PhD Candidate, Smart Cities, MIT Media Lab
Dimitris Papanikolaou, MS Candidate, Smart Cities, Media Lab
Raul-David "Retro" Poblano, MS Candidate, Smart Cities, Media Lab
Andres Sevtsuk, PhD Candidate, Dept. of Urban Studies & Planning

Schneider Electric
Plymouth Rock Studios

TEDx Boston Video on Sustainable Urban Mobility-on-Demand

Case Studies: