Modelling highway or public transport congestion is essentially a route-choice aggregation problem. It is also the 'Achilles Heel' of Activity-Based Models because all the heterogeneity captured by modelling the activity of individuals has to be lost during the aggregation imposed by current assignment models. The other option is to use a traffic simulation model butthis is inappropriate for large multi mode transport networks. The activity modeller is forced into adopting one of these two extremes as there hasbeen nothing in-between - until now. This paper explores the middle ground between these two extremes and develops a methodology which provides a continuum-of-modelling-complexity, suitable for modelling congestion, whileretaining progressively more of the heterogeneity of travel, captured by activity-based models. At the coarse end of the continuum the methodology defaults to normal assignment and at the other end it comes close to micro-simulation. The methodology was implemented in software and tested in a model of Truro and the results compared and contrasted to illustrate some of the issues and complexities. The paper discusses how these ideas can be implemented into working practices and researched further. The continuum-of-modelling-complexity was considered as modelling greater (or lesser) precision in the following variables: travel group size, value of time, activity duration and segmentation; the time-of-day of travel (either departuretime or arrival time); the time-of-day at each node, link and junction along the travel path; time-dependent queuing and junction delay. The unit of travel was the travel group size. At the fine end of the continuum, travel groups could be individuals while at the course end of the continuum they could be normal trip matrices. At an intermediate level they could be 10-minute interval travel groups with a common departure time, activity duration, value of time etc. Travel groups could be travelling from origin todestination or for the finer end of the continuum they could be from nodeto node. The time-of-day that the travel group visited each node, link and junction along the path was noted as their path was built. Travel groupscould be sub divided so as to arrive at each node, link or junction slightly earlier or slightly later. Travel groups were aggregated by time-of-day for each junction turning movement in (user defined) small time intervals. The junction turning movements by time of day were input to a time-dependent queuing model starting from early morning until late at night to calculate the queue length and delay profile for each junction turning movement by time-of-day. The junction delay depended upon the arrival time of the path at the junction - so if a path arrived at a junction at 8.25am, thedelay it experienced would be that which related to the (say) ten-minute time-of-day interval 8.20 to 8.29. The queue length and delay for each junction turning movement by time-of-day was used to find paths for the next iteration. Iterations were completed until equilibrium in the usual way. The methodology was applied to a case study in Truro with different levels of precision. The results from the case study were compared with the results from applying a conventional assignment model. The paper describes the methodology, results from the case study and draws conclusions for practitioners, indicating avenues for further research. This paper moves the state-of-the-art considerably forward by offering activity-based modellers a solution to the hitherto stumbling block of assignment aggregation. The innovation was in dealing with a user-define set of (assumedly homogeneous market segments of) travel groups, which can vary down to be individuals andup to matrices. Building paths and aggregating travel groups by time-of-day and using this to get better estimates of delay by deriving junction turning flows and delays by time-of-day using time-dependent queuing theory.For the covering abstract see ITRD E135582.
Samenvatting