The Smart Bus Project: Redefining Commuting in Cities

Project Description

The need for a real-time public transport information system is growing steadily. People want to plan their city commutes and do not like waiting for long hours, nor take a long route to reach their destination. The proposed hardware solution in this paper computes the shortest path to reach the destination in real time and gives that information to the bus driver. Artificial Neural Networks (ANN) is used to give an accurate estimate of the arrival time (ETA) to the commuter by means of an application. ETA to the next stop is communicated to the commuter using the MQTT (Message Queuing Telemetry Transport) protocol, by the hardware mounted on the bus.The proposed solution also adds a fleet management console to the administrators, making them manage and monitor the fleet of buses in real time. The prototype thus developed makes sure the commuting in cities is pleasant, and hassle free.

My role and contribution

I was the lead designer and developer of this project. I was responsible for the hardware prototype development for the buses, developing the communication infrastructure and using machine learning for decision making based on the current and historic data for shortest path estimation.

Project Details

  • Dijkstra Algorithm is optimized to compute the shortest path in real runtime.
  • The ETA computation to the next stop is done using an Artificial Neural Network implemented on the Beaglebone Black.
  • The commuter can track the bus to the desired location and get the real-time ETA using the android app built with this project.
  • The transport manager can track and monitor the fleet in real-time using a web application with Google Maps.

Implementation

The implementation was carried out using Quectel L10 GPS module and Beaglebone Black as the hardware kits to build the data acquisition and processing hardware, which is mounted on the bus. The communication between the bus- administrator and bus-commuter is facilitated using by the MQTT protocol. The Beaglebone Black was configured to run Debian Wheezy with all the codes to acquire, process and store the data was written in Python. The Beaglebone Black is additionally configured to store the acquired data as logs and also push it to the cloud (in this case - Google Spreadsheet). The results were studied over a period of two months. From the data acquired, a Multi-Layered Perceptron with Adaptive Learning Artificial Neural Network is implemented to predict the estimated time of arrival(ETA), which is communicated to the commuter and the administrator.

Publication

This project is accepted for presentation followed by publication in the proceedings of the 2016 IEEE International Conference on Advanced Networks and Telecommuncations Systems (ANTS), Bengaluru India.