Roadmap

Roadmap for Implementation

This page contains high-level tasks or Milestones (referred to as 'epics' in some project tracking systems), for the implementation of ActivitySim. It is intended to be collaboratively created and edited by the partners in this project.

We will use GitHub issues and tasks to reference to these Milestones, and Git pull requests will also link to these. This approach enables ongoing tracking of the project from its high-level objectives, to the specific tasks that need to be completed to achieve them, to the commits that resolve them.

Milestones

Below are the Milestones for the implementation phase of this project, which will subsequently be fleshed out to include a list of issues/tasks to complete each one:

1. Extend UrbanSim data types to include skims, trips, tours, transit paths
2. Extend UrbanSim model types to include N-level Nested Logit
3. Extend UrbanSim model types to include within-household interactions
4. Register every data structure to be used in ActivitySim
5. Register every model to be implemented in ActivitySim
6. Create an instance of every data structure (name, structure, elements)
7. Create an instance of every model (model structure, reference to data)
8. Create test data and unit tests for each model instance
9. Create run management process to execute all model components in model system
10. Complete a simulation run exercising all models, using test data
11. Use data from one MPO to expand model specifications to minimal configurations
12. Code all variables needed for minimal specifications
13. Test model system using minimal configuration on data from one MPO
14. Profile computational performance and prioritize performance tuning
15. Refine and generalize virtual transit path building procedures
16. Refine skim-handling procedures to optimize performance
17. Test model system using full configuration on data from one MPO
18. Write Documentation for model system API
19. Write User Documentation

Proposed Implementation Phasing and Budget

Phase 1 - a first pass over all the relevant tasks, including all data structures and models for a simple trip version (no tours), then a simple tour (no intermediate stops, no joint tours) using MTC / SF data. Model system runs end to end on SF and MTC data. Simple model specifications. Model results summaries: web maps/charts/tables.

Schedule: 6 months for basic version and initial release. Conda / PIP installer Cost: $150K

Phase 2: add complex tours, then joint tours. Extend model validation and reporting.

Schedule: 6 months, with updated installers Cost: $150K

Phase 3: Finish out model specifications for reference model (MTC), implement full calibration procedures.

Schedule: 6 months Cost: $150K

Phase 4: Refinement: performance tuning, increased testing, more complete documentation; adding features/specifications as needed; training materials; and deployment among all agency partners, including integration with various traffic assignment platforms

Schedule: 6 months Cost: $150K

Phase 5: Ongoing maintenance (repeating): continual testing of all partner's models, updating of code and supporting libraries as needed, increased testing, improved documentation and tutorials, new features as prioritized

Schedule: 6 months Cost: $25K per partner (scaling with new partners)