The Effective Integration of Analysis, Modeling, and Simulation Tools

MULTI-RESOLUTION AMS TOOL INTEGRATION

INTRODUCTION

Existing AMS tools have the capability to meet the needs of specific and relatively simplistic activities performed by different agencies. However, these tools have limited capacity to replicate complex system performance across different domains. For example, macroscopic TDMs are incapable of modeling detailed signal control, making them unsuitable to model traffic management and control alternatives. Running speed is a constraint for large networks in a microscopic simulation model. Real-time information and simulation of nonrecurring events are only available in the DTA and simulation models. Table 2 provides a high-level overview of AMS capabilities.

Table 2. Capabilities of AMS tools.

Although it is not suggested to use a single model across purposes for different domains, transportation models can be integrated on many levels because most of these domains are correlated with others on multiple dimensions. Figure 3 illustrates possible integration links between multiple AMS domains in transportation. The user may need to perform an integrated process when a comprehensive and complex transportation analysis is needed. The software tools for most of the integration links between the model domains in figure 3 are either unavailable, still in the research/development phase, or can be purchased as proprietary software programs. In many cases, users need to manually edit or create their own utility tools for integration purposes.

Figure3. Illustration. Level of aggregation for various AMS tool domains.

USE CASES

Several existing example use cases pertain to supporting transportation decisionmaking to solve complex challenges facing the transportation system. These challenges are multi-modal and diverse by geographic location. Temporal elements have multiple interrelated influences on the system including traffic demand, transportation capacity/supply, economic market forces, land use allocations and decisions, environmental conditions, and the desire to reduce crash frequency, to name a few. Table 3 presents a summary list of example use cases, typical user agency(s) that may be involved, current practice, and improved practice. The improved practice should be driven by the development and use of the AMS data hub.

Table 3. Use cases for AMS tool integrated modeling applications.

NEEDED CAPABILITIES FOR EFFECTIVE MODEL INTEGRATION

A review of AMS tools and modeling practices has shown that current integrated modeling practices in transportation are mostly ad hoc and relatively inefficient. Inherent challenges to the modeling process include data constraints and inefficiencies, interoperability, calibration/validation, and adequate modeling behavioral response/feedback.

Data Constraints and Inefficiencies

Each stand-alone model program generally has sufficient data management techniques; however, each model program generally has a unique data format that impedes the ability to seamlessly integrate with other software programs. Effective model integration must bring diverse datasets into conformity. Currently, some users are developing ad hoc tools for integration. The lack of clear guidance and data standards is a barrier to entry for resource-constrained agencies that want to apply integrated modeling techniques. The following subsections provide additional detail regarding data constraints.

Data Availability

The types of input data required are often not readily available to modelers or are difficult to obtain, particularly for arterial streets and freeway off-ramps. A wave of new data sources, particularly probe-based data sources such as automatic vehicle identification, Bluetooth^®, and data from private vendors along with real-time data from traffic signal controllers are becoming increasingly available for use in transportation modeling; however, these data sources are not yet widely integrated into standard modeling practices and still require significant labor resources to compile, postprocess, error check, and convert to a useful format.

Data Quality

Detector failure rates have been found to be as high as 60 percent, nearly double the 30 percent failure rate that is frequently cited as the average condition.⁽⁵⁾ Poor data quality requires more time for scrubbing raw sensor data and raises questions about the credibility of modeling results.

Data Format

The lack of standardized data formats for common elements such as demand data (i.e., O-D, turning movement, vehicle trajectories, etc.), junction control data (i.e., signal timing), and, to a lesser extent, roadway network data requires manual manipulation or customization of utility tools to interface between models. This increases the risk of error in transposing data inputs. The additional effort also takes away from time that could be spent on calibrating/validating, running additional scenarios, and performing sensitivity tests.

Many AMS tools have proprietary components that are encumbered by copyrights or other restrictions. These limitations create impediments for exchanging data across independent software packages and require the development and application of individual utility tools.

Data Exchange

It is important to standardize data exchange between tools at different levels of resolution in order to fully integrate multiple stand-alone models or simulation programs. An open standard would enhance the interoperability of analytical/simulation tools to coordinate and work together in a common (virtual) environment. This ability requires the cross resolution model integration architecture to be built on a common understanding of the transportation network, traffic demand, traffic control devices, and traffic sensor data.

System Coupling

Although a full integration approach within a single platform is desirable for cross resolution modeling, it requires a common data model and single-user interface for geographic information system (GIS) and other data analysis software. Thus, it is more practical to develop a future integrated modeling approach using either a loose coupling approach or a tight coupling approach. The utility tools serving the transfer of data files between different resolutions of common models need to be well defined and designed in order to be scalable, modular, interoperable, and extendable. Individual models should be developed on a modular basis so that each module can perform its function and be easily connected and extended to meet future modeling needs.

Interoperability

Software developers are enhancing the functionality of their suites of programs to enable integrated and cross resolution modeling to act as a one-stop shop for modeling needs. Examples include PTV Visum/PTV Vissim^® by PTV Group^®, TransCAD^®/TransModeler^® by Caliper^®, and Aimsun^® by Transport Simulation Systems^®. However, given the lack of industry-wide data standards and protocols, it is generally not feasible from a time and resource perspective for practitioners to work across multiple software-developer packages, and even if they did, the process would be far from seamless. This underscores the driving need to establish guidelines, standards, and protocols to address data handling and exchange.

CALIBRATION/VALIDATION

Few guidance or support tools exist for the calibration/validation of performance at a network level and for cross resolution model integration. Most validation is generally performed by qualitatively assessing the results and output for reasonableness. To the extent quantitative validation is performed, it is generally based on link or turning movement volumes. It is often difficult to validate performance measures such as travel time and route choice given the lack of available data, particularly when broken down for individual O-D pairs and routes.

The lack of calibration and validation can lead to a credibility issue with the public and decisionmakers and result in agencies shying away from applying data-driven models.

BEHAVIORAL RESPONSE

One of the most significant gaps in the functionality of current models is the lack of behavioral response associated with congestion, pricing, and traveler information (and other demand management strategies). Transportation models in many cases require significant manipulation and calibration to produce reasonable results for networks that are oversaturated and/or include demand management strategies that affect travelers’ choice of mode, departure time, and route. By only focusing on the supply side, many transportation analyses do not reflect the true effects of operating conditions, particularly when it comes to examining performance over multiple days for the purposes of estimating reliability. Ongoing research efforts are addressing this shortcoming and are expected to produce enhanced modeling tools and techniques that will establish a foundation for the future state-of-the-practice. As such, the development of standards, tools, and guidance as part of this project should look ahead to meet future modeling needs, not just current practice, which may soon be outdated in some cases.