MobilitApp: Deep learning to predict citizens’ transport modes while ensuring data security

Biofertilisers with biopesticide properties and biostimulants based on wastewater sludge from the production of biogas
March 14, 2023
A gamified virtual reality application for the rehabilitation of patients with chronic disabilities
February 22, 2023


MobilitApp is an application capable of predicting the mode of transport used by citizens through a tool based on deep learning from three mobile phone sensors.

MobilitApp includes algorithms that enable data to be gathered from multimodal journeys (changes in transport modes) with total privacy. This will help to easily analyse flows and habits of citizens’ movements to determine users’ origins and destinations and where and when changes in transport modes occur. The aim is to offer a better public transport service and at the same time boost more sustainable mobility.

The application was designed for the city of Barcelona in collaboration with the Metropolitan Transport Authority (ATM). Since February 2023, it has started to be integrated into the code of the application MOU-TE of the ATM. In September, pilot tests are planned in collaboration with UPC Sustainable, to analyse the mobility habits and the pollution footprint of voluntary students from all of the UPC.

MobilitApp uses three sensors: the accelerometer, the magnetometer and the gyroscope. In addition, it reads a sample GPS at the start and end of each journey, and in each change in mode of transport that is detected. The mode of transport that is used is predicted with an accuracy of 98% through a deep learning model, trained with a sufficiently large, representative database. This database includes unimodal journeys labelled by volunteers on journeys, using 12 modes of transport: metro, bus, tram, train, electric bicycle, electric scooter, running, walking, motorbike, car and even resting.

MobilitApp will help to study the combinations between transport means (e-scooter + bus, bicing [bike share] + metro, etc). The analysis of the flow of mobility can be broken down by variables such as gender, age range, time bands, day of the week and zones of the map, among others.

MobilitApp, which is available on Android and soon also on iOS, has been developed by the research group Smart Services for Information Systems and Communication Network (SISCOM) of the Universitat Politècnica de Catalunya - BarcelonaTech (UPC), in collaboration with ATM. It is funded by MOBILYTICS (Anonymization technology for AI-based analytics of mobility data, TED2021-129782B-I00, strategic projects on the ecological transition and the digital transition); and COMPROMISE (Enhancing communication protocols with machine learning while protecting sensitive data, PID2020-113795RB-C31/AEI/10.13039/501100011033), which are both Ministry of Science and Innovation projects.




Related Projects