1. Modeling Passenger convenience in Metro transit
    This work builds upon existing research in the area, studied during our joint survey of related work, and applies the work to the context of the Paris metro and New Delhi metro system. This work captures ‘personalized’ experience of passengers during a multi-leg journey and models the convenience for commuters. The work proposes a mathematical model of commuter convenience and validates it using data collected from metro commuters. The convenience model uses 3 convenience measures namely seat availability, wait time and comfort. The work also aims to identify the best interaction paradigm for enabling timely data collection and dissemination and outlines a middleware architecture for this (aiming at acceptable response times to mobile apps). The work was accepted as a conference paper and presented at Intelligent Transportation Systems Conference in 2015. Details of the convenience model are presented in the paper.

  2. Preparation of the recommendation engine
    Both sides are currently collaboratively working on developing a user (commuter) based recommendation engine for recommending personalized metro routes to the commuters. The recommendation engine takes into account the feedback ratings provided regarding the convenience measures on the scale of 1 to 4 by the commuter for their metro multi-leg journey. The recommendation engine tries to identify similar commuters to provide the rating of the convenience measures on their desired multi-leg journey. To accomplish this, a study of different types of recommendation engines has been done to identify the best-suited recommendation engine. The same is then used on real-world dataset of commuters’ feedback on the convenience measures to propose personalized metro routes to the commuters. Currently integration of a recommendation engine in under active implementation.

  3. Preparation for real-word studies
    An Android application - MetroCognition, for gathering commuters convenience during a metro transit based on the 3 above described measures, has been developed, deployed, and made available for testing on Google Play Store. The Android application also collects connectivity dataset that will be used in the Sarathi specific middleware. This Android application supports the metro systems of 2 cities, Delhi and Paris. We are currently extending the same to be available to the public with various UI updates and feature implementations such as integration of recommendation engine. The complete use-case for the use of recommendation engine is as follows:

    1. The commuter/user of the app opens the MetroCognition app installed on their phone.
    2. They provide the source and destination of the commute.
    3. The recommendation engine ranks the list of identified paths to the user between the provided source and destination. The most recommended path is the one listed at the top.
    4. The user can select any path from the list and check the convenience measures attached to the select path.

  4. Preparation of Sarathi specific middleware
    In the context of Sarathi, Inria is responsible to provide the middleware platform. On top of this platform, the recommendation engine will be deployed and run. The middleware platform must guarantee acceptable levels of response times and delivery success rates between the data sources (recommendation engine and/or metro travelers) and mobile users (metro commuters). Therefore, prior to the platform implementation using specific middleware technologies, it is essential to study the deployment context (metro), the different ways of interaction, and the level of response times and delivery success rates.