Per Trans 2007

Challenges in the Tracking and Prediction  of Scheduled­Vehicle Journeys Dalia Tiešytė, Christian S. Jensen,  {csj,dalia}@cs.aau.dk

Professional Communication Course 2006, Aalborg

Professional Communication, 25­27 October 2006, Aalborg

Scenario

GPS Position

Historical data Position

Minimize costs

Central server

2 18:05 AAU Busterm

Positionrelated info

Maintain accurate state

Professional Communication, 25­27 October 2006, Aalborg

2

Roadmap  Introduction • Challenges   

Efficient tracking Accurate prediction Historical data analysis

• Summary • Questions, suggestions, …

Professional Communication, 25­27 October 2006, Aalborg

3

Introduction • Goal: state correspondence with accuracy  guarantees at minimal costs Vehicle state

Vehicle Server’s  Prediction

Vehicle state? Efficient tracking

Statistical  analysis

Server Prediction Prediction  data

Challenging! Professional Communication, 25­27 October 2006, Aalborg

4

Challenges – Tracking Vehicle Actual  position Shared  prediction

Required  accuracy

Server

Update  policy

Predicted  position Shared  prediction Server’s  prediction External  data

Professional Communication, 25­27 October 2006, Aalborg

5

Challenges ­ Prediction Historical journey patterns Tracking data

A smart algorithm

Predicted journey patterns

External data

Professional Communication, 25­27 October 2006, Aalborg

6

Challenges – Statistical Analysis Historical GPS data Clustering Similarity function

Professional Communication, 25­27 October 2006, Aalborg

Historical journey patterns

7

Summary • Efficient tracking: minimal costs, accuracy  guarantees, dependency on prediction  algorithms. • Accurate prediction: external factors are  difficult to foresee, tracking data is sparse. • Statistical analysis: matching of sub­ journeys, incomplete tracking data,  efficiency

Professional Communication, 25­27 October 2006, Aalborg

8

 Thank you!  Questions? Suggestions?

Professional Communication, 25-27 October 2006, Aalborg 9