Truck Assignment Record of Discussion

Introduction

SWIM has had a long standing issue of with assigning trucks, this is more fully described at the bottom of this page. The description at the bottom of the page is a question that was posted to the TMIP (Travel Model Improvement Portal) at the end of 2020 / beginning of 2021. The question posted to TMIP (copied at the bottom of this page) generated a lot of expert feedback from across the industry and across the world. This page summarizes that feedback in an attempt to capture and record as much of that feedback and expertise as possible. The summarization of responses is broken into Themes and Specific Guidance. The Themes section captures conceptual approaches and ideas that usually came from multiple sources. The Specific Guidance section covers ideas that were closer to a prescriptive approach for Oregon to consider (as opposed to just a concept to consider).

Themes

The following bullets are summaries of the main concepts that were shared from the TMIP community on how to better approach Truck Assignment.

• Acquiring additional data (example GPS vehicle traces) and re-estimating assignment to better account for penalties.
• Local streets might have different terms and penalties than the national highway system (as an example). Overall support for the idea proposed in the question to the group, where roadways off the national highway system might get a different impedance multiplier than higher functional class roads (like the national highway system).
• Multiple parties suggested moving away from an equilibrium assignment and instead using a discrete route choice model / multinomial logit choice (having similar benefits to a stochastic approach, where multiple paths are represented for each O-D pair). Some related and provided research - https://www.caliper.com/pdfs/traffic-assignment-and-feedback-research-to-support-improved-travel-forecasting.pdf
• Pre-loading trucks (any vehicle class) seems generally to be discouraged, although a number of peers identify that it can be a very practical solution (simple representation) of behavior that is very difficult to fully enumerate.
• Custom treatments (extra time penalties) are sometimes added to segments known to be problematic (areas with high curvature, or other geometric or highly specific situations). "Many highway networks, particularly as to local streets, are not very well suited for getting the accurate impedances for trucks. Tweaks may be necessary."
• Assign a different posted or free-flow speed for trucks versus cars (which Oregon / SWIM has), specific advice is to use this to consider speed adjustments for geometric items like grade, curvature, railroad crossings, neighborhood roads / number of access points (some documented examples of this were shared).
• When considering truck versus car travel speeds, the performance management (NPMRDS) data is a great and standard source for the differences between car and truck speeds. Data sources available to each agency will naturally vary, but the NPMRDS data is the one that’s available for free to all DOTs and MPOs in the US.
• "There is no substitute for a high-level of spatial precision."
• A “truck” model is a big simplification of a “freight” model. “Truck” models can be quite useful, but their limitations need to be recognized.

Here is a list of all the different terms that were provided that could potentially be included in representing path decisions (assignment):

• distance,
• free-flow time,
• congested travel time,
• reliability of travel time,
• road hierarchy,
• legal restrictions
• smaller lane widths,
• driveways/access interference,
• inappropriate turn radii,
• pedestrian conflicts.

All trucks/shippers are not created equal. Here is a list of items on consider if vehicles can be broken up into different classes:

• Potential split between articulated and rigid vehicles.
• Possible range of different size categories.
• Potential split between loaded and unloaded vehicles.
• Are drivers following navigation software? Maybe the alternative route is better, but navigation software is programmed to look at different characteristics? If a lot of drivers are using technology, can you find out what the navigation software is considering and use this to support your solution? Market penetration of navigation assistance could be important.
• How many drivers are personally paying the toll or feel like the toll cuts into their income?
• Many trucks travel directly from the point of production to the point of consumption, but many others do not.

Specific Guidance

The following topics and subsections are specific ideas from the TMIP community that Oregon should consider as they think about new ways to approach Truck assignment (and assignment overall).

Specific Route Weighting Factors to Consider (Scott Smith, Volpe)

In our Freight and Fuel Transportation Optimization Tool (FTOT), which routes over the national FAF network, we use an impedance penalty for the higher functional classes (e.g., 1.0 for Interstate, 1.3 for a local road). Since it is a national network, it does not include all links.

We have a public release of FTOT, https://github.com/VolpeUSDOT/FTOT-public

The documentation is buried in a 141MB zip file (which also contains scenarios). Key sentences:

“Weights are used to encourage flows on lower functional class roadways (e.g. interstates and highways). Truck_Interstate includes FAF Function Class 1. Truck_Principal_Arterial includes FAF Function Classes 2 and 3. Truck_Minor_Arterial includes FAF Function Class 4. Truck_Local includes all other FAF Functions Classes (excluding those above). More information on functional classes and the FAF road network is available at the following URL: https://ops.fhwa.dot.gov/freight/freight_analysis/faf/faf4/netwkdbflow/index.htm “

“The impedances apply a scaling factor to the base dollar cost. For example, FTOT is designed to prefer faster/larger links on the road network. Therefore, a multiplier of 1 is used for interstate highways, versus a multiplier of 1.3 for local roadways, …”

From the configuration xml file: <Truck_Interstate_Weight>1.0</Truck_Interstate_Weight> <Truck_Principal_Arterial_Weight>1.1</Truck_Principal_Arterial_Weight> <Truck_Minor_Arterial_Weight>1.2</Truck_Minor_Arterial_Weight> <Truck_Local_Weight>1.3</Truck_Local_Weight>

Consider Treating External Trips Differently than Internal

First do the following as a simple test:

1. Separate the Truck trip-tables into E-E, E-I, I-E, I-I . Assign the E-E truck trip-table to a specific E-E network, essentially the freeway network ( you determine that)
2. For the E-I trip-table, basically comprised of mostly long trips, assign to the entire network where trucks are allowed on, BUT factor the local roadways’ impedance by a factor greater than 1.0
3. I would separate the I-I Trucks into several tables based on predetermined internal trip-lengths, for example: - less than 1 to 5 mile trips will be allowed to use only the non-freeway limited-access network (you have to use some method to determine these trips) use impedance factors as necessary - trips greater than 5 and to 10 miles to use access/egress links to minor/major arterials and no freeway. Links (you have to use some method to determine these trips) use impedance factors as necessary - trips greater than 10 miles to use access/egress links to minor/major arterials and freeway links (you have to use some method to determine these trips) use impedance factors as necessary)
4. Try testing various loading strategies, eg: loading E-E the E-I, I-E then I-I truck trips first, in the same order, then the auto trips.

These are all trial-and- error tests. Remember, that regional and statewide models shown only strive to get reasonable results at the freeway link level. Anything expected on lower level links will lead to frustration.

It's really important to get correct truck values of time (Alan Horowitz)

The link below should get you to the preprint of a TRR article on the subject. We eventually concluded that the values of time in the bottom half of table 3 worked best for the most part, but we were pretty sure of the range. The point of the paper is that truckers are fairly rational in their route selection in order to minimize cost of haul. The full citation of the paper is:

Qinfen Mei, Mazen I. Hussein, Alan J. Horowitz, “Establishing Values of Time for Freight Trucks in Order to Better Understand the Impact of Toll Policies”, Transportation Research Record, Journal of the Transportation Research Board, Number 2344, 2013, pp 135–143.

Microsoft Word - EstablishingValuesOfTimeForTrucks_v21.docx wpmucdn.com

Related comment from a different professional - My teams always try to use a variety of VOTTs (for travelers as a whole, rather than trucking specifically). If you have just one, you’ll have big shifts in truckers using a tolled link (0% jumping to 100% or the reverse, for many OD pairs). In reality, everyone’s VOTT changes from day to day & trip to trip, and truckers are similar. Their loads & drop off times change. So it’s truly a continuous distribution that we just approximate with 3+ VOTTs.

I wouldn’t want to use tolling results if a single VOTT is used for that class of user (truckers in this case), but I realize many people use a single VOTT for trucks (since having multiple classes is time-consuming in network loading, and they should already have at least 3 for passenger vehicles).

Original Question to TMIP Community

Written by Alex Bettinardi 12/31/2020, Accepted/Posted to TMIP 1/2/2021

I have a transportation assignment question for the wild world as 2020 comes to a close.

I'm wondering what ideas / treatments have been applied to try and keep trucks on major freight corridors as opposed to picking an optimal path by simply time, distance, and cost.

As background, this is for Oregon's Statewide model, where we assign freight and passenger traffic separately. Over the years we have tested various weights of distance versus time, but neither really solves the issue. In either case (distance priority or time priority) there can be odd network configurations where a shorter time or distance path is through a neighborhood versus on the freeway. So if we shift the priority some sections look better as far as attracting the right amount of freight traffic and some sections look worse. Shift it again, and different areas look better and different areas look worse. To be clear, we block links where trucks are not allowed to travel (and this helps), but there are many links where trucks can travel to reach their shipping destination but they are not primary routes, they are just side routes to finalize the last mile of delivery. But assignment doesn't know that and uses them at the first instance of those routes becoming optimal.

We recently had a peer state give us an interesting idea - pre-load the assignment with trucks first, so that trucks make a longer trip length decision without knowledge or the easy ability to shift based on local congestion changes. This idea was very interesting, but after looking at it more, it seems like that would be problematic for analysis in Oregon. Specifically as Oregon continues to move forward with tolling analysis. If trucks see a congestion free network first, they would be the first to jump ship at a policy made bottleneck like a toll. So where some might expect trucks to be the last vehicles to try and avoid the toll; if they are assigned first, they would likely be the first (most likely) vehicles to re-route around the toll.

My current plan/thought is to tag links with national freight highway designation. Any link tagged as a freight route would see the normal link impedance in assignment (1x). All links not tagged as a freight route would have impedance multiplied up by something like 2x (would need to play with the multiplier). This would allow trucks to dodge tolls on other major freight routes, but it would put a high penalty on dodging a toll through a small commercial street. It would also push the trucks to get on freight corridors as fast as possible and stay there for as long as possible - which I believe would be pretty close to reality.

A long question for the end of a long year. Please share your experience and thoughts with me. Many thanks

Alex Bettinardi