For the modeling of individual decisions, utility maximisation models are increasingly used. The decision of the road user is made as a discrete choice between different available alternatives depending on the interpreted utility differences.
The use of modal choice as well as the route selection and assignment results then chnage benefit components. Thus, a feedback is necessary for the application of sequential models. The (known) disadvantages of the sequential approach is the inconsistency in the overall system.
Dependencies of individual decisions, as observed in reality, are neglected due to the complexity.
To overcome the disadvantages outlined, the modal split and route choice should simultaneously take into account the mutually influencing change in demand and supply parameters.
The mission of the project was to develop a combined mode and route choice model prototypically as a multi-agent system, and if necessary to collect data.
The main objective of this project is to develop and test a combined mode and route choice model in the framework of a multi-agent-system simulation.
The project specifically contributes into two research areas:
- A traffic survey that includes an automated completion and evaluation ofdetailed transport mode and route choice data for selected individuals. Data was collected using both a conventional trip diary which the individual filled out, and a parallel GPS based automated tracking of trips undertaken.
- An agent-based combined mode and route choice model illustrating the capability levelof the studied approach.
Based on the analyses of already existing data as well as new data collected during the project individual choice models are developed specifically to be implemented in a multi-agent-system. One agent is representing a single person. The variability in individual decisions is simulated using the individual characteristics and behavioural attribute values. A survey of 1500-2000 motorised and bicycle trips to empirically support the approach is technically supported (using GPS). The system is tested for urban situations.
The project consists of the following activities:
- Evaluation of the effectiveness of traffic information in traffic situations (Agent Simulation)
- Abstract commuters Scenario (route selection model) with and without traffic forecast (Agent Simulation)
- Expo.02 traffic simulations
- Transport model DETEC: models of network - road and public transport - using the VISA transport model
- Traffic simulations with VISSIM (city of Wankdorf)
- Transport model of the canton of Bern
- Demand modelling Rail 2000 II - Zurich region
- Modelling of the performance of parking facilities
- Modelling of shopping behavior with agent systems
The project results revealed the following:
The survey addressed the individual’s travel behaviour, focusing on mode and route choices. It combined a conventional personal trip diary (including personal data questionnaire) with an automated collection of the distances travelled using a GPS data logger. The information was used as a basis for the model development.
At the time of survey automated data collection and derivation of transport mode based on GPS was a relatively innovative process. The GPS data were processed using newly developed algorithms (contributed by international research partners). Those algorithms automatically derive trip characteristics (for example trip stages and the assumed transport mode). The collected GPS- and model derived data was crosschecked with the information recorded in the personal trip diaries. It was found that the geopositioning aspects of the GPS data were very helpful (for example travel time und length) while the automatic determination of the trip mode was less accurate.
Combined transport mode and route choice model
A simulation model for agent decision making about multimodal routes was developed. Existing data from earlier projects was used to test for model robustness and validity. It was shown by comparing the model results with surveyed and analysed data from other models that the developed agent-based simulation forms indeed a successful approach for showing the dependency of route choice decisions on different individual factors as well as illustrating the effect of combined mode and route choice decisions on the overall dynamic system.
In addition to the more reproduction of model results from traditional transport modeling approaches, a significant advantage of a Multi Agent System appears to be that – because the simulated agents are able to directly control their travel mode/path -, there is a greater degree of flexibility during iterative analysis. In each iteration an agent makes a route/mode choice based on past experiences of the trip time on a particular route and adapts its choice accordingly. Simulation results are only based on those individual choices. It would also be possible to incorporate decision aspects such as travel cost, passenger comfort and ‘personal’ preference.
This approach paints a realistic portrait of our everyday experiences with trip time and delays on certain routes as well as of the impact that these experiences have on our willingness to t