A strategy-based transit assignment model for unreliable networks

In this paper we present a model how passengers navigate through overcrowded public transit networks. The model is then integrated into a transit assignment procedure. Variations of this model can be used to reflect different levels of information. This allows estimating the benefit a portable journey planner might have for passengers who need to navigate through unreliable networks. In many cities public transportation systems have reached their capacity limit. During the peak hours vehicles are crowded and often passengers are left behind on the platform. In some cases subway stations need to be temporarily closed in order to prevent overcrowding on the platforms. Passengers are aware of these facts and take them into account when they choose their route and their departure time. When a public transit network reaches its capacity limit it is no longer reliable from the point of view of the passenger. It no longer suffices to choose a path through the network, since it may contain a ride on a vehicle that might be overcrowded. Instead, a passenger needs to have a strategy to navigate through such an unreliable network. He has to determine what to do, if he fails to board a vehicle. A schedule-based transit assignment model that is based on strategies was presented by Hamdouch and Lawphongpanich (Hamdouch and Lawphongpanich, 2008). In that model a strategy consists of an ordered set of choices at each station. When the capacity of a vehicle is too small to accommodate all passengers, a passenger who is unable to board the first vehicle of his choice tries to board the second best and so forth. An optimal strategy minimizes the expected generalized cost of travel. The expected cost depends on the cost of the possible outcomes and their probability. As opposed to a single path, a strategy is always reliable. In this paper we analyze two use cases for this strategy concept. The first use case is a portable journey planner that guides a passenger through an unreliable network. The input is information about the schedule and the crowding situation inside vehicles and at stations. We analyze to what extent additional information improves the resulting strategy compared to a passenger who has less information or other restrictions. Passengers may be unaware of the capacity restrictions or have less knowledge of the schedule and navigate based on headways; or - due to the complexity of the network - they restrict their choices to certain connections that are deemed attractive. The results may be used as an indicator how portable journey planners that access online information about the crowding situation can improve the travel times of the passengers. The second use case is transit assignment. We present a transit assignment model that is based on strategies. It seems plausible that in unreliable networks passengers leave earlier than their desired departure time in order to account for the unreliability. We show that this effect can be reproduced in our model