Important Considerations for HCM 2010 Calculations ...

The calculations of the 2010 Highway Capacity Manual (HCM) differ significantly from those of prior HCM’s, particularly with regard to treatment of actuated behavior.  This includes the calculation of saturation flow, capacity, v/c, queues, delay and level of service, with all of their associated factors.  It is important that users of earlier versions of the HCM and/or TEAPAC are aware of these differences, especially as to how they affect the calculations differently than before.

Capacity Analysis

The big change in the HCM 2010 for signals is explicit modeling for actuated signals.  The impact of this change in approach on the use of TEAPAC Complete 2010 is not insignificant.  On the surface, all you have to do to produce a 2010 capacity analysis is select 2010 in the Edit-System-Output dialog for Signals before using Results-Analyze.  However, you’ll want to strategize a little more about exactly what you want the analysis to produce in relation to what is on the street.  Here are some important thoughts on that subject – please read them and try to grasp how this will affect what you see in the results.

First, the HCM 2010 signal model is now inherently an actuated model, so the ACTUATIONS input is paramount.  If all ACTUATIONS entries are No, then the analysis will be quite similar to an HCM 2000 analysis – the GREENTIMES entered will be used as phase maximums and Recall-to-Max will be employed to effect the fixed-time operation.  However, if any of the ACTUATIONS entries are Yes, then the actuated model will kick in, with significantly different results than with HCM 2000.  Primarily, the conditions entered will be used to estimate the average phase durations before delays and queues are computed, and the cycle length entered will be primarily irrelevant (unless GREENTIMES are entered in sec/sec or the signal is coordinated), and the average cycle length will be computed and displayed in the capacity analysis along with the average phase times.  This can have a very significant effect on the results as compared to HCM 2000 where the cycle length was always assumed fixed and the average phase times were assumed equal to the timings entered – splits can change considerably due to gapping out, and delays and queues can be reduced dramatically due to lower average cycles.  Pay close attention to the average cycle length and average phase durations, especially as they compare to the entered CYCLE and GREENTIMES.

Defaults exist for all of the actuated inputs that are required to model actuated control, but it will be desirable to make sure these defaults make sense and are at least reasonably consistent with conditions in the field or as anticipated.  Probably the most important of these is PASSAGETIMES, which defaults to 2.0 sec in HCM 2010 (3.0 sec in HCM 2000).  The new default value for PASSAGETIMES in TEAPAC 2010 is 0.0 sec, which tells the program to use the default for the HCM selected, but prior data files may have a non-zero value entered which will override the default and possibly cause inconsistent results in comparison to problems entered from scratch with the new program.  Another important consideration in this regard will be the detector size which is entered in the Edit-Movements menu using FIRSTDETECTS and LASTDETECTS.

A signal is designated as an uncoordinated signal in TEAPAC by entering 0 (zero) as the coordination reference phase in the Edit-Phasing-Offset dialog.  This is the new default in TEAPAC 2010 – it used to be 1, but really didn’t matter in HCM 2000.  Now it matters big time!  If the signal is coordinated (ref phase >0, and this is probably the case in many old TEAPAC files), then the CYCLE input is assumed to be the coordinated background cycle, and the actuated model will estimate actuated phase durations, with any early release time accumulating in the coordinated phase until the coordinated cycle completes, rather than allowing the next cycle to begin as it does if it is an uncoordinated signal.  Thus, the reference phase entry is now crucial in regards to how the average cycle length will be determined and how the average phase durations are distributed in this cycle – check the reference phase entry in Edit-Phasing-Offset and make sure it is appropriate for your conditions!

You’ll find a number of new results in the Capacity Analysis Summary (HCM 2010) (see Appendix D example) that will be useful, not the least of which is a graphical bar diagram showing the average green duration distribution for all phases in both rings.  For example, the Summary now includes the satflow and capacity.  If you want to delve into more details of the analysis you can select the Basic worksheet option in Edit-System-Output for Signals to produce some of the details.

Optimization

Optimization is as easy for the 2010 HCM as it was for the 2000 HCM - with 2010 selected in the Edit-System-Output dialog for Signals, simply use the Results-Design menu to optimize individual signal timings, as before.  However, there are a lot more moving parts in a 2010 capacity analysis as compared to 2000, as described above, so since TEAPAC optimizes critical movement HCM delay with repetitive capacity analyses at the heart of the optimization, there are a lot more things to be aware of that will impact the speed and effectiveness of the optimization.  The most significant thing to be aware of is that optimization will invariably take longer for 2010 than users are used to for 2000, so a few simple steps should be taken to produce optimized results as quickly as is practical.

The easiest thing that can be done is to limit the cycles to be tested and to make judicious choices about what phasings to be considered.  Especially for a first-cut analysis, start the cycle range with a cycle that is rational and implementable (not too small), and use a coarse cycle increment up to another rational upper bound cycle (not too large).  For 2000, a common and rather arbitrary range like 40 to 240 in increments of 5 or 10 seconds was typical; for 2010 something like 60 to 150 in increments of 30, or 80 to 160 in increments of 40 would be a more prudent range to use to significantly reduce the optimization time.

As far as phasing go, several options are available.  First, for the 2010 method, there is no distinction between sequence codes 4, 5 & 6, so these are automatically merged into a single optimization for sequence code 4 and the results are then interpreted as either 4, 5 or 6, as appropriate.  This immediately reduces the maximum number of phasing to optimize down from 64 to 36 without the need for any user adjustment.  Further, if certain phasings are simply not in the realm of consideration, then they can be excluded to save time - this might apply to sequence codes 1, 2 & 3 where protected phasing is not provided to all movements, or to sequence codes 7 & 8 which involve operation that is sometimes considered atypical.  On the other hand, it is always useful to optimize these phasings at least once to determine whether they have any merit in contrast to the 'normal' 4, 5 & 6 - it makes good sense to perform an initial optimization that includes all sequences for a broad range of cycles with a coarse cycle increment to get a feeling for what sequences and cycles might make sense to achieve certain critical levels of service, then perform a more precise optimization for a limited range of sequences and cycles.  The special sequence code abbreviation characters A, B, C & D can be used effectively in this effort - use AA for all 36 codes, BB to exclude split phase and lead-lag, and CC (or simply 44) to only consider standard, eight-phase, lead-only operation.  Read more about the special sequence code abbreviation characters in the manual or help dialog for the Sequences input dialog.

It is also important to be aware of the impact that a number of actuated control input variables can have on interpreting optimization results.  For example, for actuated signals that are not coordinated, the cycle length provided for optimization will be for the sum of the maximums in each ring, but the average cycle for operation and the capacity analysis will frequently be less than the sum of the maximums, and the difference between the optimized cycle and the average cycle can have an important impact on interpreting the optimization results.  (Remember that a coordinated signal is defined by the existence of a reference phase number for the Offset dialog entry.)  Similarly, values input for Recall such as Min or Max can have a profound effect on the ability of the optimizer to achieve the desired target delays and levels of service, much like how input Minimums can have a similar effect for the 2000 method.  Another phenomenon to consider is that a low-volume movement may frequently gap out, preventing the optimizer from giving it enough average green time for longer cycles to achieve a desired delay target.

Lastly, since the HCM 2010 method is inherently an iterative procedure that uses many iterations to determine the average phase durations (and thus the additional time required for optimization), there will be circumstances where the capacity analysis iterations can work at cross-purposes against the optimization iterations and prevent the kind of precise target delay optimizations that we've been accustomed to in TEAPAC for the 2000 HCM.  We will continue to work to speed up the unique TEAPAC optimizer that allows the user to target specific delay for non-priority critical movements, while at the same time improving the precision that is achieved for the target delay.  In the meantime, note that optimization for the HCM 2000 method is still in place, as before, and can be used prior to either an HCM 2000 or HCM 2010 capacity analysis.

Home | About Us | Products | Distributors | Demo Software | Price List | Ordering Information | FAQs | Training/Events Calendar | Publications | Related Sites | Contact Us | Join Our Mailing List

this page last updated June 17, 2011