Abstract

In The Netherlands one million people use public transportation every working day. The railroad network is overloaded and minor incidents lead to a large amounts of consecutive delays. Travellers are informed of delays by signs along the railway track, but there is no system, helping travellers to find the optimal route by public transportation,which also uses the delay information. The Personal Intelligent Travel Assistant provides the user with a personalised, up to date advice on a handheld device, which takes into account the actual delays.
The travellers themselves provide information about the delays of trains,using positioning technology (like GPS)to estimate whether a user is experiencing a delay. Furthermore the NS website provides information on delays for 83 stations in the Netherlands. This information can be imported into the prototype. Using the actual delay information, each user gets an optimal route advice, which also takes into account his personal preferences.
To find optimal routes,considering train changes and the user preferences, the DYNET algorithm of Eduard Tulp was selected and implemented. By extending the static information on the train schedule with dynamic information on delays, a prototype of a dynamic route guidance system was designed and implemented, which runs on the Sharp Zaurus handheld and a distributed server backend.
A model was designed for distributing the delay data over a set of servers, which can also be used for simulation purposes. Furthermore interfaces were designed for manually adding delays, changing the system time and speeding up the simulation. The system was realised by agents modelled after the Beliefs,Desires and Intentions agent model, using Jade combined with Jadex, forming a FIPA compliant distributed agent platform.