User Tools

Site Tools


multi_resolution_modeling

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

multi_resolution_modeling [2014/10/18 21:56] (current)
Line 1: Line 1:
 +====== Multi-Resolution Modeling (MRM) ======
  
 +====Integration with Microscopic Simulation ====
 +
 +A vast amount of effort has been invested to enable the integration between DynusT and macroscopic travel demand models as described previously. It was the model design objective to complete the macro-meso-micro integration to enable flexible and robust modeling capabilities combining models with different resolutions and modeling strengths as show below. In this section, further discussions are presented for the integration of DynusT and a selected microscopic simulation model. ​
 +
 +{{ :​macro-meso-micro_integration_sm.jpg | Macro-Meso-Micro Integration}}
 +
 +Mesoscopic and microscopic models are complementary to each other, and with proper integration,​ both can jointly accomplish optimal modeling capabilities. However, the capability to integrate mesoscopic and microscopic models is almost nonexistent. Translating the mesoscopic model and results to a microscopic model has been shown to be troublesome and time-consuming without a streamlined process. Microscopic models have proven to be difficult to calibrate and apply because of their richness in parameters and their dependency on large sets of fine-grained,​ accurate input data. Microscopic models also require large amounts of computational memory and efficiency and therefore large networks are both troublesome and time-consuming to create. ​ Mesoscopic models on the other hand, have shown their ability to accurately model the dynamics of traffic demand in large scale networks but lack the detailed resolution needed to analyze vehicle interactions. This section describes the functionality of a newly developed tool that integrates a mesoscopic simulation tool (**DynusT**) with a microscopic simulation model (VISSIM) through the use of PTV America’s ® open COM interface.
 + 
 +Historically,​ multiple scenarios were modeled in **DynusT** through the use of its Dynamic Traffic Assignment (DTA) capabilities. DTA, a component of both Advanced Traveler Information Systems (ATIS) and Advanced Traffic Management Systems (ATMS) uses either historic or real-time data to estimate and predict network traffic conditions. **DynusT** allows the user to use either pre-trip or enroute information to model ATIS and ATMS strategies, and has the ability to simulate multiple user classifications (MUC). These MUC's (e.g. auto, truck, HOV/HOT) can be further defined in terms of their responsiveness to available information. Once all components are determined, DTA is run in the form of User Equilibrium (UE) where vehicles are assigned on various paths until convergence is reach or vehicles cannot improve their travel time by switching routes. Output files are generated in **DynusT** and are used as the basis for conversion to a microscopic network.
 +
 +Output data from **DynusT** in the form of time-dependent shortest paths and flows, were then manually fed into VISSIM as model input parameters. Roadway networks had to be created manually in VISSIM, and all input parameters for calibration were individually fed into the microscopic model. The most tedious and time-consuming part was converting dynamic routes from **DynusT** to static routes in VISSIM. Since **DynusT** runs DTA, hourly traffic volumes are continuously changing over time. Researchers and engineers are continuously scratching their heads on how to import dynamic path files to the micro level without compromising routing or traffic volumes. The new tool developed enables mesoscopic users to create microscopic models with high levels of resolution and detail without the lingering task of data transfer or network recreation. Paths from the mesoscopic level are exported as time-dependent static routes, allowing for a more realistic time-based distribution of traffic. The end result is a tool that literally reduces the time to convert a mesoscopic model to the microscopic level. The capability of this new tool also allows for detailed intersection-level analyses based on the network-wide traffic assignment results. ​
 +
 +DTA models like **DynusT** complement microscopic models in that **DynusT** estimates the system-wide traffic flow distributions resulted from changes in demand (e.g., peak spreading, or evacuation, etc.) or supply (e.g., improved link capability, signal timing, etc.) Due to simulation granularity,​ **DynusT** won’t provide level of resolutions or sensitivity to certain operational strategies (e.g., lane configuration) or to provide certain MoEs (e.g. emission estimates, control delay, etc.) as microscopic models do. 
 +
 +Generally, **DynusT** integrates with microscopic models in an offline fashion. Once the dynamic traffic assignment is completed on the entire network, the user would define the sub-area of interest and the assigned flows that go through the sub-area would be processed to create both the demand tables, and vehicle and path files that are consistent with these flows. As shown in the figure below, the vehicles with trajectories traversing through the sub-area in the original network will be kept in a consistent fashion in the sub-area data. 
 +
 +{{:​start:​overallconcept.png|Sub-Area Cut}}
 +
 +The next integration step is to create the microscopic network that corresponds to the sub-area. At this moment, a software tool has been created through the collaboration with Texas Transportation Institute (TTI) to automatically take the **DynusT** sub-area dataset (including network, routes and assigned flows) and convert it to VISUM/​VISSIM format. This automatic process would save a significant amount of resources for the users and to ensure the correctness and compatibility between the **DynusT** and VISSIM datasets. ​
 +
 +For the integration with CORSIM, at this moment, the user needs to create the sub-area network separately in TSIS, and then enter the entry flow rates and intersection turning ratios from **DynusT** outputs. Although TSIS allows a user to load the vehicles through its vehicle and path files, previous experience has shown that it is difficult to convert routes from **DynusT** to TSIS format due to wide difference in node numbering scheme between the two software package. On the other hand, VISUM/​VISSIM allows identical node numbers and network structure to be directly transferred from **DynusT**.
 +
 +More information about integrating with VISSIM and CORSIM will be presented in the following sections. ​
 +
 +==== Integration with VISSIM ====
 +
 +
 +
 +
 +
 +
 +=== Overall Concept ===
 +
 +In collaboration with Texas Transportation Institutes'​s Center for Intelligent International Transportation Research (CIITR) in El Paso, and PTV America,​Inc.,​ a **DynusT** to VISSIM Conversion (DVC) toolkit has been developed and is currently under extensive testing. The combination of **DynusT** and DVC allows a sub-area to be defined in **DynusT** and the sub-area along with the assigned time-varying routes and flows to be automatically and consistently imported to VISSIM. ​
 +
 +The integration of **DynusT** and VISSIM provides expanded dimensions of modeling capabilities for existing **DynusT** and VISSIM users to take advantages of the modeling strengths of both models. In the modeling framework shown below, **DynusT** can be used to perform the DTA analysis in a regional network, capturing the flow distributions resulted from major demand or supply side scenarios/​alternatives. The assigned time-varying flows and routes can then be imported into VISSIM for a user-defined sub-area for the modeler to perform detailed operational strategy analysis. ​
 +
 +The "​feedback loop" between VISSIM and **DynusT** could be used as an important modeling task and thus needs to be exercised with careful modeling judgment. If the modeler believes that the operational strategies being introduced and tested in VISSIM creates only localized impact that is not significantly enough to induce system-wide flow re-distribution,​ then the feedback to **DynusT** can be omitted. Otherwise, performing the feedback loop to update the DTA flow distributions may be desirable. ​
 +
 +It is also important to note that the traffic flow dynamics between **DynusT** and VISSIM needs to be tuned to consistency before conducting integrated analysis. In other words, both VISSIM and **DynusT** need to simulate traffic and produce similar MoE results in order for the integrated framework to produce meaningful and comparable results. The consistency can be established primarily by calibrating both models using the field data. The steps to calibrate **DynusT** in the dimensions of OD matrices, traffic flow models and signal discharge are discussed in the [[start:​calibration|Calibration]] section. ​
 +
 +Another thing to note is that once the modeler decides to perform the feedback loop, the strategies that are implemented in VISSIM also need to be implemented in **DynusT** to ensure the strategy consistency. For example, if a freeway interchange configuration is being changed in VISSIM, such a configuration needs to be also coded in **DynusT** to ensure the consistency in the network geometric configuration. ​
 +
 +{{ :​dvc_flowchart.jpg?​500 | Modeling Framework }}
 +
 +
 +
 +==== DynusT-VISSIM Conversion (DVC) Tool  ====
 +
 +The DVC Tool was developed through a joint collaboration with the Center for Intelligent International Transportation Research (CIITR), Texas Transportation Institute recognizing the growing needs for integrated DTA models like **DynusT** and microscopic simulation models such as VISSIM. The current version of DVC, written in Python script, processes the **DynusT** traffic simulation and assignment outputs and creates the COM objects for VISUM including all links and nodes, and more importantly,​ vehicle paths and flows through the use of VISUM'​s "​extended route import"​ function. The genereted network in VISUM format then uses its export function to convert the new subcut roadway network to VISSIM format. Dynamic routes defined in **DynusT** are converted to time-dependent static routes and flows for VISSIM. The next version will take advantage of the VISSIM 5.0 Abstract Network Model (ANM) feature to bypass VISUM and to improve the robustness and interoperability of DVC. 
 +
 +The following figure shows that a sub-area is cut from **DynusT**. ​
 +
 +{{ :exm1.jpg |Sub-area cut from DynusT}}
 +
 +DVC converts the sub-area to the VISUM network as shown below. Since mesoscopic simulation does not have the same level of resolution as microscopic simulation, special care was taken on deciding how vehicles would be generated in VISSIM. Therefore, scripting logic was needed to locate and define generation links. All non-freeway links, which were defined as generation links in **DynusT**, were split and a new bi-directional link was added to simulate either a smaller localized road or driveway. This newly generated network produces two more links for each defined path of travel (i.e. generation roadway and destination roadway), however, the tradeoff is a more realistic simulation of traffic conditions at the microscopic level. ​
 +
 +{{ :exm2.jpg | DVC to VISUM}}
 +
 +VISUM converts the network to VISSIM as shown below. Additional modifications are still needed to complete the conversion process. Vehicle types and traffic compositions can be defined at this point to replicate a more enriched representation of vehicles. Operational procedures including right-of-way (priority rules/​conflict areas), speed limits and speed reduction areas are also needed to complete the conversion process. ​
 +
 +{{ :​exm3.jpg?​246 |VISUM Conversion to VISSIM}}
 +
 +=== Things to Remember ===
 +
 +1. VISUM 10.0 and VISSIM 5.0 are recommended to use this tool. 
 +
 +2. Make sure the hardlock key is plugged in.
 +
 +3. Make sure that “Trajalt – flag to write AltVeh.dat, AltPath.dat,​ AltEnQ.dat is set to 1 on parameter.dat. This would ensure that DynusT generates necessary files for DVC to use. 
 +
 +4. It is advised that a user revisits his/her DynusT model first and make sure that – for the subarea of interest, do not allow vehicles to be generated/​loaded on links vehicle are not supposed to be loaded in VISSIM (i.e. Underneath a freeway at a diamond interchange – these are usually set as generation links but may not as realistic in microsimulation)
 +
 +5. One can tell if the program is running by looking at the CPU usage on task manager. If it is at 50% or higher, then the program is running. If it drops right away, then there was a problem. Further improvement in this regard is underway. ​
 +
 +6. Current conversion rate is approx 1,000 vehicles per minute (if using combined demand). It is considerably longer if using separate truck demand. So if a user's model generates 50,000 vehicles, it should take about an hour to convert.
 +
 +7. Do remember to remove the offset from DynusT before the conversion as DynusT has directional link for each link but VISUM takes only the center line of both directional links. The descriptions are not in the current user’s manual – on DynusT go to the tools menu then click “Save LinkXY.dat without offset”. This needs to be done before using the DVC tool.
 +
 +
 +
 +=== Best Practices ===
 +
 +There are several issues that may arise when converting from DynusT datasets to VISSIM format. First, all freeway underpasses should not be considered generation links in DynusT. Changing these sections of roadway links to non-generation links prevents the DVC tool from automatically splitting the links. Second, nodes at these freeway interchanges (i.e. diamond interchanges) should not be considered generation or destination nodes. Again, this is to prevent the DVC program from creating either unwarranted vehicle loading or unrealistic paths. ​
 +
 +Another issue that may arise deals with the subarea cut function in DynusT. Typically, user's often cut only freeway sections of the network. However, after performing a subarea cut, there may be instances where the freeway links are the only links left in a particular zone. The current version of DynusT must have generation links associated with all zones. Therefore, it is recommended that at least one "​arterial"​ link be left in each zone. This is to prevent the program from converting freeway links into generation links.
 +
 +
 +
 +=== Further Information ===
 +
 +A [[http://​www.youtube.com/​user/​jeffshelton2|DVC Demo Video]] is provided for a user to understand the entire conversion process. The {{:​dvc_users_manual.pdf| DVC user's manual}} is also available herein. To inquire how to access or acquire the DVC tool, please contact [[j-shelton@tamu.edu|Mr. Jeff Shelton]] at Texas Transportation Institute. ​
 +
 +==== Integration with CORSIM or Synchro ====
 +
 +The integration with CORSIM or Synchro has not been automated at this moment. However, an alternative approach has been developed to facilitate the integration. First the CORSIM or Synchro network of interest needs to be created manually. The turning ratio needed for each intersection can be obtained from **DynusT** by preparing the SigOpt.dat file that lists the node numbers by which the time-varying turning ratios will be generated, as shown in the figure below. ​
 +
 +{{:​start:​sigopt.png|SigOpt.dat}}
 +
 +After the simulation, the SigOptOut.dat will be generated and this contains all the turning movement counts for all the inbound approaches for the intersections specified in SigOpt.dat. Such turning volume/​ratio information can then be entered into CORSIM intersection properties dialog boxes.  ​
 +
 +The format of SigOptOut.dat is explained using the following example:
 +
 +“Time=> ​  ​60.0” is the time instance at which the movement counts is generated.
 +
 +The following block indicates that for the inbound approach coming from node 1324 to intersection 6992, 6 vehicles are counted to make the turn to node 7015 and 2 vehicles are counted to make the turn to node 7020. 
 +
 +From    1324 To    6992  remaining PCE   1.0 K    4.1
 +                             To Node   1324 Outflow N     0
 +                             To Node   6991 Outflow N     0
 +                             To Node   7015 Outflow N     6
 +                             To Node   7020 Outflow N     2
 +
 +Similar blocks of information are given for each inbound approach of each specified intersection. ​
 +
 +If multiple time instances are reported, the counters are reset. At the next reporting time instance, the net intersection counts (only vehicles crossing the intersection between the previous and the current time instance) are reported. ​
 +
 +{{:​start:​optout.png|Output}}
 +
 +[[xyntarakis@pbworld.com|Mr. Michail Xyntarakis]] ​ at Parsons Brinkerhoff has also developed a Synchro to DynusT converter.
 +
 +
 +====NEXTA Subarea Cut Procedure====
 +
 +DynusT is capable of modeling extensive areas such as large regional networks. ​ While this capability is valuable for large-scale regional planning, it becomes inefficient when modeling scenarios for more localized, jurisdictional or corridor-based purposes in simulation and modeling. ​ To handle such cases, a subarea network component was added to the functionality of NEXTA. ​ This allows a user to “cut” a section from a larger network and create a subarea network for analysis. ​ Once the dynamic traffic assignment is completed on the large network, the subarea network procedure will maintain the assigned flows that go through the subarea and are processed to create both the demand tables, and vehicle and path files that are consistent with these flows.
 +
 +
 +
 +
 +===Subarea Cut Steps===
 +  - Ensure that the line “1 - vehtrajectory.dat written at the VehJO interval, 0: no” in the parameter.dat file is set to 1, and that the base network has been run once with this setting. ​ If this has already been done prior to beginning the subarea cut, skip this step.
 +  - In NEXTA, load network to be cut.
 +  - In the menu, select “Advanced Utilities”,​ then “Create Subarea”. ​ Using the mouse, draw a polygon by selecting the location of the polygon corners to define the boundary of the subarea. Note: Saving is strongly recommended before continuing on to step 3.  Saving here creates subarea.dat in the base network, which makes performing a cut of the same subarea simpler if it is needed in the future (e.g. if changes are made to the original network, or if multiple scenarios will be tested). ​ Subarea.dat can also be copied into different versions of the network (e.g. from a subarea to the original network), as long as the utilized feature points in zone.dat are identical. {{ :​start:​3a1.jpg |Sub-Area Cut}} {{ :​start:​3b.jpg |Sub-Area Cut}}
 +  - In the menu, select “Advanced Utilities”,​ then “Delete Nodes and Links outside Subarea”. {{ :​start:​4a1.jpg |Sub-Area Cut}}
 +  - In the menu, select “Advanced Utilities”,​ then “Save Subarea Network to Another Folder”. ​ Create and open a new folder to save the subarea network to, and save the subarea network in this new folder. {{ :​start:​5a1.jpg |Sub-Area Cut}}
 +  - Close NEXTA, and verify that the necessary input files (e.g., network.dat,​ movement.dat,​ origin.dat, destination.at,​ xy.dat, zone.dat, etc.) have been created in the subarea folder.
 +  - Reopen NEXTA and open the newly created subarea network from the new subarea folder created in step 5.
 +  - In the menu, select “Advanced Utilities”,​ then “Import Vehicle Trajectory File from Entire Network”. ​ Select the VehTrajectory.dat file in the original uncut network folder from step 1. {{ :​start:​8a1.jpg |Sub-Area Cut}} {{ :​start:​8b.jpg |Sub-Area Cut}}
 +  - In the menu, select “Project”,​ then “Assignment/​Simulation Settings”,​ and change the generation mode to “Veh + Path File”, and the iterative consistent assignment to “One-shot”. {{ :​start:​9a1.jpg |Sub-Area Cut}}
 +  - In the menu, select “Project”,​ then “Scenario Data”, and change the MUC distribution to 100% habitual (i.e.: set habitual to 100, and all other values to 0). {{ :​start:​10a1.jpg |Sub-Area Cut}} {{ :​start:​10b1.jpg |Sub-Area Cut}}
 +  - Run DynusT. {{ :​start:​12.jpg |Sub-Area Cut}}
 +This concludes the steps of performing a sub-area cut in NEXTA for DynusT.
multi_resolution_modeling.txt · Last modified: 2014/10/18 21:56 (external edit)