Simulation of a Fully Automated Bicycle Sharing System Showed Higher Bicycle Usage, with over 96 Percent of Bicycles Used in a Week, Compared to 92 Percent with a Station-Based Bike Sharing System Having a Fleet Size 3.5 Times Larger.
Researchers Assessed the Mobility Impacts of a Fully Automated Bike Share System Concept for the Boston, Massachusetts Area Using Agent-Based Discrete Event-Based Simulation.
Date Posted

On the performance of shared autonomous bicycles: A simulation study

Summary Information

For cities with sufficient density, where resources are often within walking distance, micro-mobility systems could be a potential efficient mobility solution for short-distance trips and first- and last-mile connections to public transit. Researchers in Massachusetts investigated a fully automated bicycle-sharing system concept, where shared bicycles were capable of driving to user-requested locations, using the Boston metropolitan area as the study network. Agent-based microsimulation provided an in-depth understanding of the fleet behavior of fully automated bicycle-sharing systems in realistic scenarios with a rebalancing system to prevent excessive bicycle accumulation or depletion in some areas. The simulation study assessed fleet performance and user experience and compared a fully automated bicycle-sharing system with station-based and dockless bicycle-sharing schemes.


An agent-based, discrete event-based simulation model was developed for the study, using historical usage data from a public bike-sharing system in the Boston area for the period October 7-14, 2019. Geospatial data from building locations were used to generate users’ origins and destinations. The impact of demand variations was analyzed by repeating the simulations with a randomized distribution of origins and destinations in buildings within 300 meters around shared-bike pick-up or return stations. Four different bike sharing system scenarios were considered in the microsimulation:

  • station-based system
  • dockless system
  • fully automated system with no rebalancing (lower bound for fully automated system performance)
  • fully automated system with ideal rebalancing (upper-bound for fully automated system performance)


  • For a nominal scenario with 1,000 fully automated bicycles, results show that compared with a fleet size of 3,500 station-based bicycles and a dockless system with 8,000 bicycles, the fully automated bike-sharing system could potentially improve overall performance and user experience even with no rebalancing. For example, the fully automated system could serve between 0.47 percent and 1.01 percent more of the demand than the station-based and dockless systems.
  • Productivity of the system, measured by the average number of daily trips per bicycle, was improved in the fully automated system scenario, with an average of 8.84–8.88 daily trips per bicycle, while station-based bikes were used 2.51 times per day, and dockless bicycles were used 1.10 times per day.
  • For the fully automated bike-sharing system with no rebalancing, the percentage of system bicycles used in a week increased was 97.20 percent, compared with 91.59 percent for the station-based bike share system and 70.93 percent for the dockless system. For the fully automated system with rebalancing, utilization averaged 96.43 percent during the week.
  • The average user wait time for the fully automated system was found to be shorter than the added walk time at origins and destinations when using a station-based system. For the fully automated system with no balancing, average wait time was 11.35 percent shorter than station-based walk time, and was 28.99 percent shorter with ideal rebalancing. However, when compared with the dockless system, the fully automated system had 16.34 percent longer total trip times when no balancing system was used. With ideal rebalancing, the fully automated system had a 6.62 percent shorter total trip time on average compared to the dockless system.
Goal Areas
Results Type
Deployment Locations