Abstract
Coupling activity-based models with dynamic traffic assignment appears to form a promising approach to investigating travel demand. However, such an integrated framework is generally time-consuming, especially for large-scale scenarios. This paper attempts to improve the performance of these kinds of integrated frameworks through some simple adjustments using MATSim as an example. We focus on two specific areas of the model—replanning and time stepping. In the first case we adjust the scoring system for agents to use in assessing their travel plans to include only agents with low plan scores, rather than selecting agents at random, as is the case in the current model. Secondly, we vary the model time step to account for network loading in the execution module of MATSim. The city of Baoding, China is used as a case study. The performance of the proposed methods was assessed through comparison between the improved and original MATSim, calibrated using Cadyts. The results suggest that the first solution can significantly decrease the computing time at the cost of slight increase of model error, but the second solution makes the improved MATSim outperform the original one, both in terms of computing time and model accuracy; Integrating all new proposed methods takes still less computing time and obtains relatively accurate outcomes, compared with those only incorporating one new method.
Highlights
Activity-based models (ABM), which attempt to simultaneously investigate individual daily activities and travel behaviour, has gradually dominated in studies of travel demand modelling, see (Rasouli and Timmermans 2014) for a recent review of Activity‐based models (ABMs)
The improvement work focused on two aspects: (1) the reduction of the number of iterations to converge (2) and the reduction of computing time for each iteration, with the application of a more focussed replanning algorithm and a varying time step approach in the execution module of MATSim
For the first solution that reduces the number of iterations, the improved MATSim incorporating the new framework and replanning module can improve the original one in computing time at the cost of slight decrease in model accuracy; While, for the second solution that reduces the time of running each iteration, the varying time step based approach outperforms the original one in both computing time and model accuracy
Summary
Activity-based models (ABM), which attempt to simultaneously investigate individual daily activities and travel behaviour, has gradually dominated in studies of travel demand modelling, see (Rasouli and Timmermans 2014) for a recent review of ABM. Replanning Replanning Execution outcomes from this paper should be of interest to modellers working on ABM This is because, on one hand, MATSim (including its extensions and variants) has become one of the most-used ABMs (as reviewed above) with a relatively large number of users across the world; on the other hand, the proposed improvements should be applicable to other ABMs (e.g., TRANSIMS) similar to MATSim. As discussed above, one of the main time-consuming parts of MATSim is the co-evolutionary module which aims to optimize daily plans of each agent through a number of iterations. The reroute module is used to find the shortest path given the current origin and destination locations in the agents’ plans; the reschedule module is used to adjust the activity-related elements in the plan, such as departure time, transport mode and activity location Both strategies are applied to adapt the daily plans to the dynamic traffic flow, aiming at optimizing the plans (Horni et al 2016). Only time allocation mutation is used, and it can randomly change the departure time of an activity, in a specific range, for example, [− 30 min, + 30 min] (Balmer 2007)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
