Reducing the Dimension of Online Calibration in Dynamic Traffic Assignment Systems

Abstract

Effective real-time traffic management strategies often require Dynamic Traffic Assignment systems that are calibrated online. But the computationally intensive nature of online calibration limits their application to smaller networks. This paper presents a principal component based dimensionality reduction of the online calibration problem, which overcomes this limitation. To demonstrate this approach, we formulate the origin-destination flow estimation problem in terms of their principal components. The efficacy of the procedure is tested using real data on Singapore Expressway network in an open loop framework. We observe a reduction in the problem dimension by a factor of 50 with only 2% loss in estimation accuracy. Further, the computational times reduced by an order of 100. Interestingly, the procedure led to better predictions as the principal components capture the structural spatial relationships. This work has the potential to make the online calibration problem more scalable.

Dynamic Toll Pricing using Dynamic Traffic Assignment System with Online Calibration

“…Their optimization system predicts significant time savings with the optimal strategies generated by it, but the actual impacts of such strategies are not tested in the real network or in a simulation environment that is different from the DTA system itself. Yang (14) apply a principle components approach to conduct online calibration using the GLS algorithm. By calibrating principle components of parameters instead of original parameters, this method greatly scales down the computation effort of large-scale online calibration problems.…”

Improving Scalability of Generic Online Calibration for Real-Time Dynamic Traffic Assignment Systems

“…We impose the constraints on the final state estimate by determining the value with the maximal probability in the posteriori distribution subject to the constraints. We determine such estimate by solving the optimality problem defined in equation (22). The constraints in equation (22) ensure that the OD-flows derived from the final state estimate ∆ẑ h|h are non-negative.…”

“…We determine such estimate by solving the optimality problem defined in equation (22). The constraints in equation (22) ensure that the OD-flows derived from the final state estimate ∆ẑ h|h are non-negative. Please refer to [15] for more details regarding the procedure.…”

Evaluation of Traffic Assignment Models through Simulation

“…Thus, Yu & Guo [31] develop a tri-level combined-mode traffic assignment model; Pi et al [32] include heterogeneous traffic on roads, parking availability and travel modes (such as solo-driving, carpooling, ride-hailing, public transit, and park-and-ride); Macedo et al [33] propose an efficient traffic assignment, where users are not only concerned about travel times, but also about global and local pollutant emissions; Jiang et al [34] include the car-truck interaction paradox in assignment problems; and Dimitrov et al [35] model the interaction between buses, passengers and cars on a bus route. Some researchers have applied these methodologies in a specific case study, often with the support of computer simulation [36][37][38][39], while others have attempted to assess different techniques or even propose novel ones [40][41][42][43][44][45]. Among the first ones, Zhang et al [36] integrate an activity-based travel demand model with a dynamic traffic assignment model for the Baltimore Metropolitan Council; Shafiei et al [37] develop a simulationbased dynamic traffic assignment model of Melbourne, Australia; Zhu et al [38] use dynamic traffic assignment for a case study in Maryland; and Kucharski & Gentile [39] apply the Information Comply Model on different situations, including a corridor in the north of Kraków, Poland, and the Sioux-Falls network.…”

“…Among the first ones, Zhang et al [36] integrate an activity-based travel demand model with a dynamic traffic assignment model for the Baltimore Metropolitan Council; Shafiei et al [37] develop a simulationbased dynamic traffic assignment model of Melbourne, Australia; Zhu et al [38] use dynamic traffic assignment for a case study in Maryland; and Kucharski & Gentile [39] apply the Information Comply Model on different situations, including a corridor in the north of Kraków, Poland, and the Sioux-Falls network. Within the latter ones, Zhang et al [40] analyze the calibration of dynamic traffic assignment models applying the extended Kalman filter; Prakash et al [41] present a dimensionality reduction of the assignment models calibration problem, based on principal components; Du et al [42] focus on the dynamic traffic assignment problem on large-scale expressway networks especially under the condition of traffic events (such as severe weather, large traffic accidents etc.) and propose an approximate solution algorithm; Lin & Chen [43] develop a simulation-based multiclass, multimodal traffic assignment model for evaluating traffic control plans of planned special events; Batista & Leclercq [44] study a regional dynamic www.videleaf.com traffic assignment framework for macroscopic fundamental diagram considering stochasticity on both the trip lengths and the regional mean speed; and Bagdasar et al [45] examine discrete and continuous optimization and equilibrium-type problems and make a comparison of them with the Beckmann cost function.…”