TMODEL2 has been modeling nodes for 15 years, and we are still learning better ways of modeling intersection operation. We are constantly trying out new techniques and occasionally cringing when we look back at what we did only a few years ago. The path to improvement sometimes takes us one step back for each two steps forward. With this newsletter article, we want to convey what we see as current best practice for modeling intersections with TMODEL2. We hope that you give us your feedback, as always!
UTPS, and other transportation planning models derived from it, does not model nodes. We at TModel Corporation think node modeling is very important, because the majority of the delay on the arterial and collector street system occurs at intersections or ramp merge locations. As in all modeling, it is important that the rules or equations and parameters used in modeling nodes be based upon logical, rational reasoning. Therefore node delay should be based upon the entering volume, conflicting volumes, capacity, type of control, and the approaches being controlled. Each of these factors is considered during the development of the model and the equations and parameters that are used. We give some examples of parameters below; however, keep in mind that your area probably operates differently and should use different parameter values.
Node Class - We feel that the node delay equation should be dependent upon the type of intersection control. That is, as the volume to capacity ratio (v/c) increases, the amount of delay should increase at different rates for different types of control. Because a signal is different than a stop sign or an off-ramp diverge, we use Node Class as a way of differentiating the type of control. This permits a separate node delay equation to be specified for each class or control. An example of possible node classifications is shown in Table 1 below.
Node delay coefficients can be developed to correspond to the node classes 1 through 100. The delay coefficients should reflect additional travel time that is incurred per vehicle as the v/c ratio increases. Examples of node delay coefficient values that may be used with the node classification scheme in Table 1 are shown in Table 2.
You should note that you can override the node delay function by placing zeros in all of the columns except for the EA column. The node delay coefficient function will revert to the default parameters (derived from TRC 212) if all of the entries in a row are zero. However, if you have one non-zero entry, such as the EA corresponding to Node Class 1, the equation evaluates to zero delay time.
In the March 1993 newsletter, we had discussed the Node Delay Coefficients. A windows module has been included with the last two updates that allows you to experiment with impacts of adjusting the delay coefficients. This module, called NODEDLY.EXE, is also available on the BBS. We test all of our changes to the Node Delay Coefficients file with this module. This is shown in figure 1 below.
Please note the values for Kiss & Ride and Park & Ride nodes. There are delays associated with the capacity for these nodes also. For example, the Park & Ride node capacity can be set at the number of available parking spaces with a steep delay curve when volume approaches the capacity of the number of spaces.
Node Area - We have not been using node area for any one use in the models we have been developing or with which we have been working. That is not to say that we do not use it. Node Area has been used to contain a unique identification number for matching with GIS records, as a key for renumbering, and to denote the subarea location of the node. It has also been used to contain the number of additional turn lanes at the node and the number of accidents for graphic representation.
Node Type - This field has been used mostly for the development of codes to derive capacities. Typically, we cross reference the classifications of the streets entering the node as a way of approximating the operation of that node.
If one considers two arterial streets, usually the volumes are about equal on the North/South and East/West approaches. If the signal timing is evenly split between the approaches, the capacity of the node should be about 50% of the entering link capacity minus lost time.
If one considers an arterial intersecting with a local street, the arterial has much higher capacity, more traffic, and a higher green time to cycle length ratio. At the extreme, if the local street never had any green time, the capacity of the intersection would be 100% of the arterial street entering link capacity.
From this method, we have derived starting percentages to estimate the intersection capacities. Typical values are shown in Table 3 below.
These Type codes can be filled in one by one, or you can use the new option in section 1.7.1.9, ASSIGN NODE TYPES. This new option will assign the node types based upon a matrix of the intersecting link classes. The desired node type is entered in the cells as indexed by the link classes. The automatic assignment of node type should save you significant time and error.
This article on node definitions is intended to get you started. We are planning on continuing this topic in the next newsletter. Other important considerations are the special delay links (SDLs), refinement of node capacity, and turn penalties. Watch for more next time.