Modelling activity travel scheduling for workers

Activity based models have been introduced by researchers to better explain the complex interaction between activity participation and travel demand. The basic idea is that in modern society individuals have to absolve a number of compulsory tasks (work, shopping, child care) and that they have more money and more free time for leisure activities out of home. Traditional models could not answer to change in policies not related to transport infrastructures. However, it is well recognized among decision-makers that only a mix of measures will affect travel behavior. Improvements in transport infrastructures can be combined with changes in lifestyle: flexible working hours, tele-working etc... There is clearly a need for knowing how activities are unchained on the daily (and/or on longer period) basis and which degree of substitution is possible among stops and tours. In this paper we present a scheduling model system to model the complete daily activity travel pattern for workers. The problem has been attacked by others authors. Ben-Akiva and Bowman (1995) calibrate a largescale operational activity based travel demand model with a nested logit structure that includes the day activity pattern model on the higher level and the tour model on the lower level. When and Koppelman (1999) propose an integrated model system of stop and tour formation to model shortrange decisions including generation of daily maintenance activities and the allocation of stops and cars to household members. Bhat and al. (2003) formalize a very similar framework (but the model system is not yet estimated -to our knowledge -) as part of an econometric simulator for daily activity travel patterns called CEMDAP. The choice of model structure and of the discrete number of alternatives to introduce into the model system is based on a comparative study, in which we explore the transferability of the framework proposed to 4 European countries (Cirillo and Toint, 2002). Following those results, the scheduling model system estimated is divided into by three components: 1. pattern level model, which accounts for the scheduling of tours made during the whole day, 2. tour level model, in which the choice of the activity to be pursued is made for each tour and 3. stop level model, which accounts for the number of stops or secondary activities per commute leg. A large number of variables have been estimated including: working schedule variables, activity durations at destinations, households and individual characteristics, level of service and inertia variables. The model is calibrated on a continuous six-week travel diary data set of long-term individual travel behaviour, which is part of the German research project Mobidrive. A similar framework is being transferred to non-workers by the first author