Agenda

1. Outputs from UC Berkeley
   • Part 1 - Connected corridors
   • Part 2 - Connected cars
   • Part 3 - Decision Support System
   • Other activities
   • Lessons learned
2. Status of the Ph.D. project
   • Thesis premises
   • Thesis outline
   • Current work
   • Gantt chart
3. Plan for the third year

1. PATH research center, Berkeley, CA http://www.path.berkeley.edu/
2. Integrated Corridors Management (ICM) Project – traffic optimisation in a corridor (a freeway and parallel arterial routes) http://www.its.dot.gov/icms/. The project aims in a lcalised traffic management of the most congested regions in California.
3. Learned about the applied research on:
   • Highway modelling
   • Highway control
   • Boundary conditions modelling
   • Arterial modelling
   • Arterial control
   • Traffic prediction
   • Traveller response
4. Worked on:
• Freeway Traffic Estimation System - using a cell transmission model (CTM) for simulating the flow of vehicles on a freeway, and an ensemble Kalman filter (EnKF) to assimilate measurements GPS probes into the estimation model

Project demonstration for client - Caltrans

• Richmond field station (RFS), Richmond, CA http://rfs.berkeley.edu/
• Connected Car (CC) Project - real-world experiments on connected vehicles with DSRC on board (OBU) and roadside units (RSU)
• Learned about on-going research and projects:
  • Cooperative Adaptive Cruise Control (CACC) more info
  • Multi-Modal Intelligent Traffic Signal Systems (MMITSS) more info
  • Real-Time, Freeway End-of-Queue Alerting System
  • Animal Warning System
• California PATH director Tom West, Program Leaders: Ching-Yao Chan, Steven Shladover, Headquarters Researchers: Somak Datta Gupta, Christopher Nowakowski, Ashkan Sharafsaleh

Workshop on Connected Vehicle Reference Implementation Architecture program (CVRIA) organized by the Intelligent Transportation Systems Joint Programs Office (ITS JPO). Multiple vendors and implementers of connected vehicle environment (CVE) commented on standards for new technologies and applications

Traffic information and management system for California

1. Monitoring of real-time traffic condition
2. Estimation and prediction of traffic behavior
3. Suggesting response strategies
4. Evaluation of the response strategies with simulations
5. Selection of the best response
6. Input:
   • Traffic events (accident, road closure)
   • Weather conditions (rain expected)
   • Social events (games, holidays)
7. Output:
   • Comparison of performance measures for different control strategies
   • Selection of the best control strategy

Graphlab workhop, San Francisco, July 1st, 2013, 2013

Connected Cars Aps (Uber, Ford, Telefonica) San Francisco, October 2 2013

GraphConnect conference, October 3-4, 2013 , 2013

Transportation Conference, UC Berkeley November 16, 2013

Transportation Conference, UC Berkeley November 16, 2013

Part of a student group at UC Berkeley: VUD Lab

1. The big picture of the transportation:
   • Traffic information systems - online systems that gather data from different sources and show real-time traffic.
     • Previous systems using historical data were good for planning.
     • Today's systems aim in short time prediction about what will happen in half an hour. They only rely on real-time data and learn typical traffic patterns so and can't response to unpredicted traffic situations.
     • Future systems aim in using real-time data to control people - manage the demand: ride sharing, routing, prices. Currently demand management is done through economic means.
   • Connected and self-driving cars.
     • Starting of autonomous cars at PATH (Shladover)
     • Dharma Initiative
     • Google self-driven cars (2017)
2. Practical knowledge about:
   • The structure of US universities and methodology in research.
   • Current Traffic Information and Management Systems in California and plans for the future development
   • Research trends in transportation.
3. Conclusion:
   • There is a paradigm shift in transportation systems: from centralised to distributed (driver-centred).
4. Ph.D. context:
   • Vehicular network can implement and improve distributed traffic information and management systems.

Title: "Community-based vehicular network for traffic information systems"

Premises:

1. Vehicular networks can be used effectively in traffic information and management systems.
   • VANET-based traffic inormation systems efficiently distribute traffic data.
   • Problem solved: Estimation of traffic condition in urban areas (currently is hard due to lack of traffic data).
   • VANETs enable control over individual vehicles (e.g. routing).
   • Problem solved: Lack of control of drivers behavior.
2. Distributed community detection enables better communication in the vehicular network.
   • Dynamic formation of communities groups vehicles who travel together and share communication for a longer time.
   • Problems solved:
     • Control of communities for demand management (routing, ride-sharing, fleet management).
     • Information from communities for traffic control (density estimation).
     • Communities can be used for analysis of patterns in traffic demand.

• Chapter 1. Introduction
  • Motivation and context, Contributions, Thesis outline
• Chapter 2. Vehicular ad hoc networks
  • Introduction
  • Definitions and specifications
  • Simulations of vehicular networks
  • Vehicular traces
• Chapter 3. Traffic Information Systems and Management
  • Introduction
  • Architecture (centralised,distributed)
  • Examples of Information Systems (Luxembourg, California)
  • Traffic management (Traffic congestion, Traffic demand, Routing)
  • Potential of vehicular networks (in information collection and traffic management)
• Chapter 4. Realistic simulation platform for VANET-based TIS
  • Introduction
  • Existing platforms
  • Generation of realistic vehicular traces (Vehilux, Gawron, Evaluation)
  • Architecture (NS3, SUMO, Traci, VANET application)
• Chapter 5. TrafficEQ - VANET-based TIS
  • Introduction
  • Problem definition
  • System overview
  • Data disemination
  • Traffic Estimation
  • Routing strategies
  • Evaluation metrics
• Chapter 6. Evaluation of TrafficEQ
  • Introduction
  • Research questions
  • Simulation settings
  • Simulation results
  • Discussion and conclusions
• Chapter 7. Community detection
  • Introduction
  • Notations
  • Social networks
  • Dynamic networks
  • Community definition
  • Community detection algorithms
  • Algorithms applicable for VANETs
  • Community analysis
• Chapter 8. Crowds
  • Introduction
  • Overview
  • Community assignment
  • Evaluation metrics
  • Evaluation of Crowds
• Chapter 9. Conclusion and perspectives

• Label propagation community detection
• Based on Leung's algorithm
• Vehicle assigns to a community based on similarity
  • Topological metric (as in Leung's algorithm)
  • Mobility similarity
• Mobility metric
• DSD (Degree of Spatial Dependency)
• sDSD (stable Degree of Spatial Dependency)
• Evaluation metrics
  • Stability index
• Comparison
  • Epidemic
  • Leung
  • Sand/SHARC
• Evaluation of Crowds

• Set plan, meetings and deadlines
• Deliverables:
  1. Journal article about TrafficEQ
  2. Results from crowds algorithm
  3. Journal article about crowds algorithm
  4. Collaboration paper with UC Berkeley
• Thesis:
  • Content
  • Submission deadline
  • Committe
  • Dissertation date

End