Download
application of data fusion for route choice modelling by route choice driving simulator n.
Skip this Video
Loading SlideShow in 5 Seconds..
Application of data fusion for route choice modelling by Route Choice Driving Simulator PowerPoint Presentation
Download Presentation
Application of data fusion for route choice modelling by Route Choice Driving Simulator

Application of data fusion for route choice modelling by Route Choice Driving Simulator

153 Vues Download Presentation
Télécharger la présentation

Application of data fusion for route choice modelling by Route Choice Driving Simulator

- - - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript

  1. Application of data fusion for route choice modelling by Route Choice Driving Simulator M. Dell’Orco1- R. Di Pace2– M. Marinelli1-F. Galante3 1 Technical University of Bari - Italy, EU 2 University of Salerno - Italy, EU 3 University of Napoli “Federico II” - Italy, EU

  2. Introduction (1) ATIS (Advanced Traveller Information Systems) are aimed to provide information on traffic conditions to travellers so that they can keep their travel decisions with less uncertainty • Travellers’ reaction to the ATIS is modeled in terms of compliance and route choices • a compliant traveler with ATIS chooses the suggested route • Route choice models under ATIS are often developed and calibrated by using Stated Preferences (SP) surveys

  3. Introduction (2) • Two main types of tools for SP in ATIS contexts are the most popular driving-simulators (DSs) and travel-simulators (TSs) • Both methods are computer-based • DSs are characterised by a greater realism, provided that the respondents are asked to drive in order to implement their travel choices, as it happens in the real world • less trials by the same respondent • In TSs, travel choices are entered after having received a description of travel alternatives and the associated characteristics, without any driving • TSs compensate some lack of realism with a minor cost and with less burden for the respondents • many more trials by the same respondent

  4. ResearchGoals • Here a pilot study is presented (10 respondents), aimed at setting up an SP-tool based on driving simulator developed at the Technical University of Bari The obtained results are analysed • in order to check the accordance with expectations • the results of application of data fusion technique are shown in order to explain how data collected by DSs, can be used to reduce the effect of choice of behavior in unrealistic scenarios in TSs

  5. Employed simulation tools • A PC-based driving simulator of Technical University of Bari has been adopted • The UC-win/Road driving simulator software was used (FORUM8) • The simulation system works on a single computer provided with NVidia Graphic Card (1GB of graphic memory) and a Quad-Core CPU • The simulation is based on a steering wheel (Logitech™ MOMO Racing Force Feedback Wheel), able to provide force feedback, as well as six programmable buttons (ignition, horn, turn signals, etc.), sequential stick shifters and paddle shifters • A 22" wide-screen monitor was used in order to have a good field of view, also showing internal car cockpit with tachometer and speedometer. Environmental sounds are reproduced to create a more realistic situation

  6. Experimentdescription (1) • The network reproduced in the virtual experiment refer to real one in Bari • The network is proposed to respondents in the simulations in a double con-figuration, with (3 repeated trials) and without ATIS (3 repeated trials) • The configuration without ATIS was presented to respondents with some variants, reproducing different congestion levels and travel times, accordingly with their statistical distribution in the real world During the experiment respondents have been asked for choosing a route among three alternatives The simulated networks were part of a real network in Bari

  7. Experimentdescription(2) • The choice set can be viewed as composed by a main route (route 1) that connects the considered origin-destination pair • Depending on traffic conditions, the traffic could spill-back up to a later diversion node (detour toward route 2) or even up to an earlier diversion node (detour toward route 3) • These three different conditions (straight route, later detour, earlier detour) are conventionally classified here as three different levels of congestion (free-flow/low congestion, intermediate congestion, high congestion)

  8. Design of the experiment

  9. Actualshare vs. Observed share

  10. Datafusion: Modelling Route Choice Behavior (1) To incorporate information on system conditions in the choice process, we assume that drivers • have some experience about the attributes of the transportation system • use information to update his experience • choose an alternative according to his updated experience. • The drivers’ knowledge about the transportation system can be expressed in the same way we used for perceived information • So both drivers’ knowledge and information can be expressed in terms of Possibility • To update knowledge of the system, drivers aggregate data coming both from their experience and from current information

  11. Datafusion: Modelling Route Choice Behavior (2) In this work, we have used a route choice model based on uncertainty-based Information Theory as proposed by Dell’Orco at al.* The information fusion model incorporates important aspects such as: • dynamic nature of information integration; the perceived cost of an alternative is influenced by the user’s previous experience and memory; • accuracy of the informative system; the more accurate information is, the more important is the effect on the drivers’ perception • non-linearrelationship between information and perception *Dell’Orco, M. and Marinelli, M.: Fuzzy data fusion for updating information in modelling drivers’ choice behavior. ICIC 2009, LNAI 5755: 1075-1084 (2009)

  12. Datafusion: Modelling Route Choice Behavior (3) • Acquired data has been fused using the method proposed by Yager and Kelman*. • The result of information fusion is a subnormal fuzzy set because its height hfis less than 1 (e.g. 0.67) as reported in figure *Yager, R. R., Kelman, A.: Fusion of Fuzzy Information With Consideration for Com-patibility, Partial Aggregation, and Reinforcement, International Journal of Intelligent Systems 15, 93 -122 (1996)

  13. Datafusion: Modelling Route Choice Behavior (4) • In order to interpret the information given by this fuzzy set, tf must be normalized • To pass from Possibility to Probability we use the probabilistic normalization, (Σipi = 1) along with the Principle of Uncertainty Invariance, systematized by Klir and Wang* • The model allows the quantitative calculation of users’ compliance with information, and thus a realistic updating of expected travel time • We have modeled drivers’ choice behavior according to Uncertainty-based Information Theory • We have applied fuzzy fusion to data acquired in Bari at the end of each simulation *Klir, G. J and Wang, Z.: Fuzzy Measure Theory, Plenum Press, New York (1992)

  14. Results: Observed Share vsPredicted Share

  15. Conclusions and future work • Collected data have to be used in order to increase the effect of reduced realism of TSs • Preliminary application of data fusion technique has been made • Expected travel times are updated according to results of data fusion and the influence of uncertainty on drivers' compliance with provided information is examined according to uncertainty-based Information Theory In future work… • The authors would like to define a methodology of route choice modeling by mixed data set collect by TSs and DSs • Furthermore the authors would like to introduce more test in order to validate the adopted modelling approach

  16. Thankyoufor the attention!! Any questions? Suggestions? rdipace@unisa.it m.marinelli@poliba.it