Exploiting the relation between activity data and traffic data within the dynamic demandeEstimation problem

Abstract

of the results:

The first step of the analysis was focused on the comparison of the trend of the observed demand, obtained from traffic data, with the trend of the aggregate Activity Function. In figure 1(a) and 1(b) is possible to observe the measured flows on a specific detector on Wednesday and Thursday.

(a) (b)

(c) Figure 1: (a) Observed traffic Flow on a detector Wednesday 10/09/2008; (b) Observed traffic Flow on a detector Thursday 11/09/2008; (c) Scatter between observed measured on Wednesday and Thursday.

Figure 1(a) and 1(b) show a very similar trend, so using this function to obtain the target matrix for the DDE problem, the matrices used to determinate the Traffic Demand on Thursday and Wednesday are very similar, as confirmed by the scatterplot between the functions in figure 1(c). A demand estimation from the above traffic counts would therefore provide very similar results.

(a) (b)

(c) (d) Figure 2: (a) Aggregate plot of the observed activities for Wednesday; (b) Scatter Activity Data and Traffic Data on Wednesday; (c) Aggregate plot of the observed activities for Thursday; (d) Scatter Activity Data and Traffic Data on Thursday;

Figure 2 presents the differences between the activity pattern for Wednesday and Thursday obtained by the surveys. The particular trend in Figure 2(a) is related to the fact that on Wednesdays in Belgium the schools close sooner with respect to the other days. The Activity's Trend shows as the trip demand is very different for the two days. The scatter between the Activity Data Function and Traffic Data Functions (Figure 2(b) 2(c)) shows that traffic demand function not ever represent the demand trend, and also when the trend of the demand is similar (Figure 1(c)) there is a Bias between the two functions, that it is transferred to the seed matrix. On the other hand, also Activity Functions present biases, related for survey coverage; an evident gap in the respondents’ population is that trips originating and ending in the study areal are not included. It is so necessary to obtain a new function taking into account the Activity Function and correcting them using the aggregate information from the traffic count.

Contribution to the existing literature:

The discrepancy between traffic data and activity data may therefore be corrected by considering that travel patterns modeled during the early afternoon significantly differ.

Once we defined the appropriate Activity Function to relate to the Traffic Counts, it is possible to disaggregate the Activity Function observing each single component. In this way is also possible to take into account in the OD estimation spatial information, and not only temporal information. On a

specific link is possible, for example, observe only the demand peak of the evening, this could be explained observing the activity in the specific area.

In the full paper we will show how demand estimation can be improved by separating seed matrices by activity and by associating a different functional relation to these activities in time. Further improvements are foreseen also considering that activity chains can be interpreted and used as constraints to the reproduced traffic patterns.

Bibliography:

[1] E. Cascetta, D. Inaudi and G. Marquis (1993). Dynamic Estimators of Origin-Destination Matrices using Traffic Counts. Transportation Science 27, 363-373. [2] X. Zhou, C. Lu, , K. Zhang (2012). Dynamic Origin-Destination Demand Flow Estimation Utilizing Heterogeneous data sources under Congested Traffic Conditions, Available online at: http://onlinepubs.trb.org/onlinepubs/conferences/2012/4thITM/Papers-A/0117-000097.pdf. wAccessed January 2013. [3] R. Frederix, F. Viti, R. Corthout, C. M. J. Tampère, (2011). New Gradient Approximation Method for Dynamic Origin-Destination Matrix Estimation on Congested Networks. Transportation Research Record, 2263:19-25. [4] G. Cantelmo, E. Cipriani, A. Gemma, M. Nigro (2014). An Adaptive Bi-Level Gradient Procedure for the Estimation of Dynamic Traffic Demand. Intelligent Transportation Systems, IEEE Transactions, Volume:PP , Issue: 99 [5] D.F. Ettema and H.J.P Timmermans (1997). Activity-Based Approaches to Travel Analysis, Published by Emerald Group Publishing Limited [6] S. T. Doherty, E. J. Miller, K. W. Axhausen, and T. Gärling (2002).. A Conceptual Model of the Weekly 22 Household Activity-Travel Scheduling Process. In: Stern, E., Salomon, I., Bovy, P. (Eds.), Travel Behaviour: Patterns, Implications and Modelling. Elsevier, Oxford, 2002, pp. 233–264. [7] J.L. Bowman, M. Ben-Akiva (2001). Activity-based Disaggregate Travel Demand Model System with Activity Schedules. Transportation Research Part A 35 (1), pp. 1–28. 28 [8] T.A. Arentze, H.J.P Timmermans (2004). A Learning-based Transportation Oriented Simulation System. Transportation Research Part B 38 (7), pp. 613-633. [9] T.A. Arentze, H.J.P Timmermans (2004). A Multi-state Supernetwork Approach to Modeling Multi-Activity Multi-modal Trip Chains. International Journal of Geographical Information Science 18 (7), 32 2004, pp. 631-651 [10] M. Balmer, K. Meister, M. Rieser, K. Nagel and K. W. Axhausen (2008). Agent-based simulation of travel demand: Structure and computational performance of MATSim-T, Innovations in Travel Modeling (ITM’08), Portland. [11] M. Feil, M. Balmer, K. W. Axhausen (2009). New approaches to generating comprehensive all day activity-travel schedules. Report: ETH institutional repository. [12] P.J. Jin, F.Yang, M.Cebelak, B. Ran, C.M. Walton (2013), Urban Travel Demand Analysis for Austin TX USA using Location-based Social Networking Data, TRB 92nd Annual Meeting Compendium of Papers.