🚀 Feature: Highway Reliability Implementation Plan

[i-am-sijia]

User Story

To implement highway reliability in highway assignment and skimming, use the reliability skims in the demand model.

Progress:

• [x] Sufficiently defined
• [x] Approach determined
• [x] Tests developed
• [x] User story satisfied
• [ ] Doc strings
• [ ] General documentation
• [x] Passing tests

Priority

High

Level of Effort

Medium

Resolution Ideas

This is a re-implementation of SANDAG's highway reliability and the travel-model-two TM2.1 Cube based reliability. Both implementations have similar structure:

• Static reliability, which is based on the facility type, free flow speed, lanes, distances to upstream and downstream interchanges. This is calculated during the network preprocessing steps

• LOS based reliability, which is based on the assigned volume, V/C ratio, facility type. This is calculated after the assignment. And the SANDAG and TM2.1 approaches seem to differ as following:

  • SANDAG: The VDF function is defined with reliability component, e.g.,. So the original travel time from base assignment includes reliability. To get the actual travel time without reliability, after the assignment, split travel time into time and reliability by creating @auto_time and @reliability link attributes, then run another assignment with skimming.

  • TM2.1 Cube version: The VDF function does not include reliability component. After the assignment in AverageNetworkVolumes.job, calculates the LOS based reliability, updates the total reliability (LOS based + static), which then gets skimmed in the highway skimming in the beginning of the next iteration

The question is do we want to implement the SANDAG way or TM2.1 way?

• SANDAG way uses EMME blended path skimming, which ensures the skim matches the equilibrium achieved in the assignment. And since the VDF includes reliability, the final "best path" includes reliability. TM2.1 uses Cube skimming, which is just from one single shortest-path, without equilibrium. TM2.1 VDF does not include reliability, thus the assignment is not reliability-based.

Project

MTC tm2py conversion

Who should be involved?

Users: @lmz Reviewers: @inrokevin

Risk

Low

Tests

Static reliability is part of the highway network preparation step, LOS based reliability is part of the highway assignment step.

• [ ] Test for...

[i-am-sijia]

We will implement the SANDAG way. It is important to include reliability in the VDF as it ensures the path fed into the mode choice is considering the reliability. It does not make much sense to send a highly unreliable path to the mode choice when there exists a slightly slower but reliable path.

[lmz]

Hi @jfdman and @dhensle - Would you mind looking at the above; is the difference between TM2.1 and SANDAG intentional? Thanks!