Soft Mobility
Challenge provided by the City of Porto and Associação Porto Digital

Optimization of soft-mobility drop-off points

What if we could increase the number of e-scooter journeys by up to 12%? Easy-to-find e-scooters in strategic locations can be the answer.

A recent review [1] found that soft mobility (including e-scooters), depending on the city and culture, is used differently for entertainment purposes or commuting. Specifically for commuting, the usage of soft mobility has the potential to help with the last mile problem. This is also backed up by the same review, which shows most trips have a distance of 0.72–2.4 km and last, on average, between 8–12 minutes. On the other hand, according to the Tom Tom traffic index [2], Porto’s traffic is worse than, for example, Madrid’s traffic, where residents spend 41 hours per year in traffic jams in comparison to 52 hours in Porto. A good transportation system is crucial, and soft mobility might be vital in improving how citizens move around the city.


The goal of the challenge was to study and analyze the soft mobility pattern in Porto to improve the overall experience and usability. More concretely, the teams were challenged to create an optimization model for optimizing the drop-off locations of the e-scooters.

United Nations SDG 

GOAL 11: Sustainable Cities and Communities

  • Target 11.2.1: Provide access to safe, affordable, accessible, and sustainable transport systems for all.

  • E-Scooter Transport Data, provided by Associação Porto Digital
  • E-Scooter Location Data, provided by Associação Porto Digital
  • GTFS for E-Scooter Parking and Metro Stations, provided by Associação Porto Digital
  • Entry and Exit validation data from public transportation in the Metropolitan Area of Porto, provided by Associação Porto Digital
  • Origin-Destination (OD) matrices of Movement of People from/to the Porto Metropolitan Area, provided by Associação Porto Digital


Most of the teams enriched the dataset provided by the City of Porto by downloading Points of Interests (POIs), the road network, schools, and city boundaries from OpenStreetMaps. One team used the taxi trajectories to evaluate if a specific place is a hotspot. Demographical data regarding each parish was also used.

Methods and Techniques

Most teams started by creating candidate points with two different approaches:  selecting them manually (e.g., current drop-off points, bus stops, and metro stops) or automatically determining them through a clustering algorithm. For example, one team used HDBSCAN, and others used k-means clustering.

All the teams, later on, applied an optimization algorithm as defined by the constraints given by the challenge provider. The optimization methodologies used were Constrain Integer Programming, PuLP Linear optimization, and mixed integer programming.

Main Insights from Data

In the data, the teams observed that the e-scooters mainly were used for recreation purposes and not as the last mile travel solution. This was seen in the data as having a more than double drop-off rate near recreational places rather than bus stops, while there were more bus stops than recreational points. This was also visible when compared to the days of the week and hours that e-scooters were used. In Figure 1, it is possible to see that the most use was at the end of the day and during weekends.

The teams did not see an expected peak during morning rush hour, and they also noted that there were many standard routes that people used (e.g., in the old town or by the sea), and this, in turn, could be used to improve safety for the most used routes. 

Figure 1  - The number of trips is higher during the weekend than during the weekdays. This could be because the scooters are mostly used for entertainment purposes.


All the teams proposed a similar product: a dashboard for service providers, government bodies, or regulatory agencies to optimize the drop-off zones and monitor their use. One team additionally proposed to create live information on the performance of each drop-off zone and indicate if it should be changed. Figure 2 shows an example of a dashboard where the user could regulate the weights (that is, the importance) for each factor to be considered when optimizing.

Figure 2  - An example of a dashboard for optimizing the drop-off zones.

Social Impact

By creating an optimized e-scooter network, the e-scooters should be easier to find, more useful to citizens, and constitute a better alternative to cars. As metrics to measure social impact, the teams proposed: 

  • Total number of scooter trips
  • Number of scooter journeys per day and increase compared to baseline
  • Percentage of scooters correctly dropped in the drop-off zones
  • Number of cars circulating in the city
  • Maintenance cost
  • Lifespan of e-scooters
  • Pickups and dropoffs: the number of e-scooters needed to be relocated

Teams estimated an increase in the number of scooter journeys between 2.53% and 12%. One team estimated this expansion based on their model, and another team made the estimation based on previous successful cases. The same team also calculated that there was a potential for a reduction in CO2 emission per transportation mile of 80.2 tons.


[1] Şengül, B., & Mostofi, H. (2021). Impacts of E-Micromobility on the Sustainability of Urban Transportation—A Systematic Review. Applied Sciences.

[2] TomTom 2022, Tom Tom Traffic Index. Available at: https://www.tomtom.com/en_gb/traffic-index/

Top 5 Solutions
Open-source code

More about this category

World Data League - a competition for data scientists
World Data League ©Copyright 2022