On centroid connectors in static traffic assignment: Their effects on flow patterns and how to optimize their selections

In this paper, we investigate to what extent the results of static traffic assignment (STA) are influenced by centroid connectors and how to optimize their selections. Three networks are used to evaluate the impact of different centroid connector configurations on the resulting traffic flow pattern: a synthetic grid network, the California SR-41 corridor network and a large Sacramento area network. From the STA results of these three networks, we observe large fluctuations on resultant link volumes, maximum volume capacity (V/C) ratios, average V/C ratios and total travel time with respect to randomized connector selections. The fluctuations seem to indicate that STA results are unstable with respect to arbitrary connector selections, and this cannot be improved by simply changing the number of connectors. In fact, more connectors often result in serious under-estimation of total travel time and average link load. We infer that, if provided little information of access/egress nodes of trips, randomly generated connectors lead to artificial over- or under-utilization on network links. We therefore propose a connector optimization algorithm in which the connectors and their travel times are chosen in such a way that the maximum V/C ratio of some characteristic links is minimized. As the numerical example on the SR-41 network indicates, this connector optimization algorithm reduces the artificial over- and under-utilization of network links, and obtains a flow pattern more consistent with the one derived from a refined network where trip access/egress nodes are known in details.

► Traffic assignment results are highly influenced by centroid connector placements.
► Too many or few centroid connectors cause under or overestimation of network travel times.
► The placement of centroid connectors can be optimized by reducing the congestion levels of some critical links.