Modular agent-based modeling of an urban road network including its management: to assess the impact of different transportation asset management decisions

Benjamin A. Cohen

doi:10.17918/kqbx-6e23

The objective of this PhD research was to explore how the complex interdependencies of a dense urban environment can be simulated through a hybrid multi-layered model, integrating agents at different resolutions. The model incorporated: (1) the graph model of an existing road network within Philadelphia, recognizing the engineering and operational characteristics of various road segments within this network; (2) vehicles and drivers as mesoscopic agents, where the vehicle types, counts and behaviors were calibrated by actual data; (3) a time and use based deterioration model and repair techniques for different elements of the road network; and, (4) an owner agent making various asset management decisions for the road network leading to four distinct scenarios. The mesoscopic traffic simulator therefore simulated feedback mechanisms linking the drivers' decisions and behaviors to the roadway steward's decisions, revealing the short-term consequences of management policies. This provided a realistic understanding of the "user costs," corresponding to different decisions on maintenance and replacement scheduling. The model proved to be a low cost, modular, sociotechnical, cloud-based model of a surface transportation urban infrastructure asset management simulator. The main computational engine selected for the model was an open-source mesoscopic traffic simulator, SUMO, which can represent driver agents, routing algorithms, and traffic control systems for a road network. This traffic model was defined and calibrated with data and reports from the local municipal planning organization. The deterioration of each roadway segment was modeled as a function of time and the simulated traffic's vehicular loads. The roadways' speed limit was a function of the deterioration and as a result, over time, the roadways became less favorable to the vehicles trying to optimize their daily path. At the beginning of each year the owner-agents assess the current deterioration levels and develop a plan and schedule either maintenance or replacement activities in order to meet specific condition objectives while also satisfying real-world constraints. The long-term planning was accomplished through the use of linear programming, while the short-term yearly schedule was achieved through a custom prioritized scheduling algorithm. A total of four different scenarios were generated to observe and compare the effects of different management policies and approaches. These included a baseline Do-Nothing scenario, a scenario based on a reactive short-term yearly scheduling scheme, and two scenarios that incorporated both short-term and long-term approaches to planning and scheduling with different constraints. A computer program was written in the Python language and adapted to take advantage of elastic cloud computing processing power to integrate the different modular components in order to simulate the daily activities within the network during a five-year period. The data from the simulations were then processed, analyzed, and visualized utilizing a combination of Python and Excel. The results for all the scenarios were compared with regards to the roadways' overall and yearly condition ratings as well as user-costs. Given the relatively short run-time of the simulation in comparison to the life-span of the assets, the short-term reactive scenario performed better than the two long-term based scenarios in the overall condition of the network. However, for longer-periods of simulation we should expect changes in the resulting operational performance and cost for different scenarios. The study was successful in demonstrating the potential of integrative modeling of complex sociotechnical infrastructures, and the computational requirements for simulating entire urban regions with interdependent infrastructures.

Modular agent-based modeling of an urban road network including its management: to assess the impact of different transportation asset management decisions

Files and links (1)

Abstract

Metrics

Details

Modular agent-based modeling of an urban road network including its management: to assess the impact of different transportation asset management decisions

Files and links (1)

Abstract

Metrics

Details

Drexel University Social media