This page details how congested times are fed back from assignment to the destination and mode choice models.

In TRMG2, each time period has its own feedback loop. MD could converge in 3 iterations while the PM requires 5. After each assignment, a congested sov skim is calculated and compared to the skim used in the destination and mode choice models. If the %RMSE is less than the target (0.1%), that period is finished and proceeds to transit assignment and summaries. If the %RMSE is larger than the target, then the new congested times are fed back to destination and mode choice and the cycle continues. This can happen up to 5 times for each period.

The feedback parameters can be modified here in the parameter editor as shown below. 0.1% is a very tight convergence appropriate for official runs. For early scenario testing and quick what-if analysis, 1% is more than adequate and significantly reduces run time.

Warm Start

To speed up model run times, the model comes with congested times pre-loaded. These were created from a previous run and then saved to the following tables in the master/networks folder:

You can change which set of congested times you want to use here: