Predictive Route Guidance

A number of researchers have emphasized that guidance should be generated based on predicted traffic conditions which take into account drivers' response to the provided guidance. Predictive route guidance can minimize the inconsistency between provided information and drivers' experience and avoid problems such as over-reaction [Kaysi et al.(1993)]. In the literature there are 3 general approaches to guidance generation: (i) assignment/simulation; (ii) optimal feedback control; and, (iii) fuzzy-neural network. While pros and cons exist for all these approaches, the assignment/simulation based approach is receiving extensive attention because of its flexibility to incorporate behavior models and provide network-wide solutions.

In a predictive system, route guidance is provided by taking into account drivers' current position, destination and projected travel time on alternative paths. The time that a driver arrives at the decision node is very relevant in providing this guidance. In calculating the expected travel time that a driver would experience on the shortest path and alternative paths, projected time-variant link travel times are used (instead of current link travel times), i.e.:

where:

travel time on path i given departure time t;

travel time on link j of the path given that the driver enters the link at time t.

An idealized guidance system based on the assignment/simulation approach for traffic prediction is implemented in TMS. The main characteristics of this guidance system are:

• simulation-based traffic prediction and guidance generation which consists of network state estimation, network state prediction, and guidance generation; and,
• periodic updates of guidance in a rolling horizon implementation.

• length of prediction horizon;
• frequency of guidance updates;
• time resolution of guidance between updates; and,
• computational delay.

• Rolling Horizon Model
  • Rolling horizon length
  • Rolling horizon step size
  • Guidance Resolution
  • Computational delay
• Network State Estimation
• Network State Prediction
  • Dynamic OD prediction:
  • Traffic prediction:
• Guidance generation