Projects on Public Transportation

  • Currently we have two on-going projects related to public transportation
  • The first deals mostly with public passenger sensing. Also supports meansfor interaction through message pushing
  • Contribution => Appliance of external sensing techniques tocompute important strategic structures at a public transportationlevel; Passenger interaction using mobile equipment; Real-time queryand analysis of passenger related data; Simulation potential
  • The second is an advisory system used to promote a more sustainabledriving behavior Contribution => Apply efficient feedback techniques to promote”smoothness” of driving; Historical feedback is better assimilatedthan instantaneous/real-time continuous feedback (most systems)


Wireless Sensing and Interaction platform

Exploit pervasiveness of Bluetooth technology (nextrelease to support additional wireless technologies)

  • Deployment of scanners scattered throughout strategic points => bus-stops
  • Gain the ability to determine passenger waiting patterns
  • Identification of passengers at entry & exit points opens innovative methods for O/D Matrix computation High granularity & Low-cost

Deployment of scanners throughout high-frequency public bus line

  • Installation of 12 stations at/nearby public transit bus stops
  • Initial objective to provide city-wide monitoring and determine approximate number of Bluetooth devices (+7000 devices in just 4 days in 1 station). Estimation of trip mean duration. Comparison with company data.


Fuel Consumption Advisory System

  • Despite technological advances, fuel consumption of vehicles remains a big problem
  • Our research aims on quantifying to which extent the use of adequate stimulus may improve driving behavior
  • Main focus on improving driving profiles by promotion of smoothness of driving
  • Basically, this means moderate speeds and low accelerations.

We estimate the use of fuel consumption using only Android based phones

  • Use of accelerometer & GPS allows to collect enough information to approximate fuel consumption
  • The Vehicle Specific Power (VSP) model correlates strongly with both fuel consumption and emission levels
  • A model that relies only on velocity, acceleration and road gradient=> Additional parameters are air and rolling resistance (may be constants, but may be changed according to other factors such as vehicle weight)
  • After initial calibration of the VSP model, we no longer need to measure instantaneous fuel consumption using instrumentation (usually through OBD system) => OBD-II socket present in vehicles since 1996 in USA; EOBD around 2000 in EU
  • Removal of road gradient influence on results presentation (”egocentric” feedback) => Most raw fuel consumption calculation fail to take this into account.

Our approach is to induce desired driver behavior by giving information and “education” on the current driving style

  • Achieving right level of “temporal granularity” difficult => too coarse and many opportunities lost; too fine-grained and information overload. Minimum requirements (Van der Voort):

1. Provide clear, accurate and non-contradictory information

2. Take into account vehicle’s context

3. Place no burden upon driver that may influence its driving task

  • Especially for high stress driving jobs (such as public transit buses), instantaneous feedback is distracting and potentially ignored => Historical feedback preferred


Project on Mobility

  • Study of mobility behaviors and patterns
  • Have the ability of automatically determine method of transportation
  • Purpose is to be able to predict any method of transportation change
  • Identify the windows of opportunity for applying motivational techniques (and evaluate them) => how can we convince people to change? Pilot showed that the use of GPS specific devices (receiver only) will probably lead to poor results => Tendency to forget device
  • Use of smart-phones recommended => Also facilitates use of data filtering => Greatly facilitates potential interactions with the user
  • Makes possible application of machine learning algorithms “on-line”
  • Possible to merge results with the Wireless Sensing platform => Bluetooth scanning
  • Need to constantly recharge is an issue => People don’t want to do this => Limitation of the sampling frequency must be ensured