Kaveh Khoshkhah

This paper presents a new simulation-based approach to address the stochastic Dynamic Traffic Assignment (DTA) problem, focusing on large congested networks and dynamic settings. The proposed methodology incorporates a random walk model inspired by the theoretical concept of the \textit{equivalent impedance} method, specifically designed to overcome the limitations of traditional Multinomial Logit (MNL) models in handling overlapping routes and scaling issues. By iteratively contracting non-overlapping subnetworks into virtual links and computing equivalent virtual travel costs, the route choice decision-making process is shifted to intersections, enabling a more accurate representation of travelers' choices as traffic conditions evolve and allowing more accurate performance under fine-grained temporal segmentation. The approach leverages Directed Acyclic Graphs (DAGs) structure to efficiently find all routes between two nodes, thus obviating the need for route enumeration, which is intractable in general networks. While with the calculation approach of downstream node choice probabilities, all available routes in the network can be selected with non-zero probability. To evaluate the effectiveness of the proposed method, experiments are conducted on two synthetic networks under congested demand scenarios using Simulation of Urban MObility (SUMO), an open-source microscopic traffic simulation software. The results demonstrate the method's robustness, faster convergence, and realistic trip distribution compared to traditional route assignment methods, making it an ideal proposal for real-time or resource-intensive applications such as microscopic demand calibration.

The emergence of smart cities is set to transform transportation systems by leveraging real‐time traffic data streams to monitor urban dynamics. This complements traditional microscopic simulation methods, offering a detailed digital portrayal of real‐time traffic conditions. A framework for near‐real‐time city‐scale traffic demand estimation and calibration is proposed. By utilising Internet of Things (IoT) sensors on select roads, the framework generates microscopic simulations in congested networks. The proposed calibration method builds upon the standard bi‐level optimization formulation. It presents a significant computational advantage over available methods by (i) formulating the optimization problem as a bounded variable quadratic programming, (ii) acquiring sequential optimization technique by splitting computations into short time frames while considering the dependency of the demand in successive time frames, (iii) performing parallel simulations for dynamic traffic assignment in corresponding time frames using the open source tool Simulation of Urban MObility (SUMO), and (iv) feeding traffic count data of each time frame as a stream to the model. The approach accommodates high‐dimensional real‐time applications without extensive prior traffic demand knowledge. Validation in synthetic networks and Tartu City case study showcases scalability, accuracy, and computational efficiency.

Real-time traffic demand estimation is essential for intelligent transportation systems and traffic forecasting in urban areas. Hence, this paper presents a simulation-based optimization framework for city-scale real-time estimation and calibration of dynamic demand models by focusing on disaggregated microsimulation in congested networks. The calibration approach is based on sequential optimization demand estimation for short time frames and uses a stream of traffic count data from IoT sensors on selected roads. The proposed method builds upon the standard bi-level optimization formulation. The upper-level optimization problem is presented as a bounded variable quadratic programming in each time frame, making it computationally tractable. The DTA (dynamic traffic assignment) problem at the lower level is approached based on the stochastic route choice model. For every time frame, the probabilistic parameters of the route choice model are obtained through several rounds of feedback loop between the OD optimization problem (at the upper level) and parallel samplings and simulations for DTA (at the lower level). At the end of each time frame, the microsimulation state of the network is transferred to the next time frame to ensure the temporal dependency and continuity of the estimations during the time. In addition, the algorithm's convergence is accelerated by applying the fixed-point method to road-segment travel times. The proposed sequential calibration model presents a drastic computational advantage over available methods by splitting the computations into short time frames and under the assumption that the demand in each time frame only depends on current and previous time frames. The model does not depend on reliable prior demand information. Moreover, with receiving the field measurements as a stream and the efficient time complexity of the algorithm in each time frame, the method successfully presents a solution for high-dimensional real-time and online applications. We validated the method with synthetic data and for a real-world case study in Tartu city, Estonia. The results show high accuracy under a tight computational budget.

This paper presents a simulation-based optimization framework for city-scale real-time estimation and calibration of dynamic demand models by focusing on disaggregated microsimulation in congested networks. The calibration approach is based on sequential optimization demand estimation for short time frames and uses a stream of traffic count data from IoT sensors on selected roads. The proposed method builds upon the standard bi-level optimization formulation. The upper-level optimization problem is presented as a bounded variable quadratic programming in each time frame, making it computationally tractable. For every time frame, the probabilistic parameters of the route choice model are obtained through several rounds of feedback loop between the OD optimization problem (at the upper level) and parallel samplings and simulations for DTA (at the lower level). At the end of each time frame, the microsimulation state of the network is transferred to the next time frame to ensure the temporal dependency and continuity of the estimations during the time. In addition, the algorithm's convergence is accelerated by applying the fixed-point method to road-segment travel times. The proposed sequential calibration model presents a drastic computational advantage over available methods by splitting the computations into short time frames and under the assumption that the demand in each time frame only depends on current and previous time frames. The model does not depend on reliable prior demand information. Moreover, with receiving the field measurements as a stream and the efficient time complexity of the algorithm in each time frame, the method successfully presents a solution for high-dimensional real-time and online applications. We validated the method with synthetic data and for a real-world case study in Tartu city, Estonia. The results show high accuracy under a tight computational budget.

In recent years, we have witnessed the emergence of the implementation and integration of significant working solutions in transportation, especially within the smart city concept. A lot of cities in Europe and around the world support this initiative of making their cities smarter for enhanced mobility and a sustainable environment. In this paper, we present a case study of Tartu city, where we developed and designed a daily real-time system for extracting and performing a modal split analysis. Our web-based platform relied on an optimization approach for calibrating our simulation in order to perform the analysis with the use of real data streams from IoT devices installed around the city. The results obtained from our system demonstrated acceptable performance versus the quality of the available data source. In addition, our platform provides downloadable OD matrices for each mode of mobility for the community.

We propose a method for reducing the nonstationary noise in signal time series of Sentinel data, based on a hidden Markov model. Our method is applied on interferometric coherence from Sentinel-1 and the normalized difference vegetation index (NDVI) from Sentinel-2, for detecting the mowing events based on long short-term memory (LSTM). With integrating our noise reduction step to the LSTM neural network architecture, we improved the F 1-score from 0.69 to 0.76.