TODO: Add wheelchair accessibility limitations to workflow

[zneedell]

Here's a list of changes that need to happen for us to successfully reproduce the wheelchair accessibility runs within PILATES:

• [x] Add wheelchair attribute to ActivitySim inputs. The easiest way to do this is probably to just modify the persons table in the custom_mpo_06197001_model_data.h5 urbansim data input file to include a wheelchair attribute (important to make sure we do this for the correct year). The starting file is available here, and BEAM expects a column in the persons.csv file titled in_wheelchair that is set to either True or False. @nazanin87.
• [x] Make sure wheelchair attribute is visible to ActivitySim in PILATES. Probably will be true by default if it's in the h5 file but worth confirming. @zneedell
• [x] Change ASim config to use different wait times depending on disability status. Let's create a new branch of the asim config repo where we modify the wait times in the trip and tour mode choice. I think the first step is to create a target requires_wheelchair by adding a line to the annotate_persons script here. As far as I can tell this script should have access to all of the columns that are in the persons file in the input h5 directory. I think doing this should make the boolean requires_wheelchair attribute available whenever we're doing mode choice. The wait times are constructed here and here. We'd need to change the existing logic, e.g.

  odt_skims['RH_SOLO_WAIT'] * (1.0 - odt_skims['RH_SOLO_REJECTIONPROB']) + TNC_missed_trip_penalty * odt_skims['RH_SOLO_REJECTIONPROB']

to something that accounts for the requires_wheelchair attribute, so e.g. something like

np.where(df.requires_wheelchair,odt_skims['RH_SOLO_ACCESSIBLE_WAIT'] * (1.0 - odt_skims['RH_SOLO_ACCESSIBLE_REJECTIONPROB']) + TNC_missed_trip_penalty * odt_skims['RH_SOLO_ACCESSIBLE_REJECTIONPROB'],odt_skims['RH_SOLO_WAIT'] * (1.0 - odt_skims['RH_SOLO_REJECTIONPROB']) + TNC_missed_trip_penalty * odt_skims['RH_SOLO_REJECTIONPROB'])

I'm hoping that requires_wheelchair will automatically make it over to df, but it might take some additional work if not. @nazanin87

• [x] Update PILATES to create separate wheelchair / non-wheelchair skims. New pilates branch wheelchair-ridehail-skims
• [x] Upload new BEAM docker image with wheelchair code. There is now a BEAM image on dockerhub at the location zaneedell/beam:0.wheelchair.0
• [x] Create beam data repo branch with wheelchair inputs. There is now a beam-data repo pilates-wheelchair
• [x] Set up PILATES scenarios with different fleet penetration of accessible vehicles. @nazanin87 with help from @cristian-poliziani

[cristian-poliziani]

Great, thanks Zach and Cristian. I'll take a look at the scripts and documents by tomorrow's meeting.

On Thu, Aug 18, 2022 at 11:41 AM Cristian Poliziani @.***> wrote:

That sounds great, thank you Zach for putting together all that and @Nazanin Rezaei @.***> feel free to write to me or plan a meeting with me when needed.

Cristian

On Thu, Aug 18, 2022 at 11:34 AM Zach Needell @.***> wrote:

Here's a list of changes that need to happen for us to successfully reproduce the wheelchair accessibility runs within PILATES:

• Add wheelchair attribute to ActivitySim inputs. The easiest way to do this is probably to just modify the persons table in the custom_mpo_06197001_model_data.h5 urbansim data input file to include a wheelchair attribute (important to make sure we do this for the correct year). @nazanin87 https://github.com/nazanin87. We'll then need to upload this new file to the AWS AMI we're using to launch these runs ( @cristian-poliziani https://github.com/cristian-poliziani can help with this).
• Make sure wheelchair attribute is visible to ActivitySim in PILATES. Probably will be true by default if it's in the h5 file but worth confirming. @zneedell https://github.com/zneedell
• Change ASim config to use different wait times depending on disability status. Let's create a new branch of the asim config repo https://github.com/LBNL-UCB-STI/activitysim-configs-sfbay where we modify the wait times in the trip and tour mode choice. I think the first step is to create a target requires_wheelchair by adding a line to the annotate_persons script here https://github.com/LBNL-UCB-STI/activitysim-configs-sfbay/blob/main/configs/annotate_persons.csv. I think this should make the boolean requires_wheelchair attribute available whenever we're doing mode choice. The wait times are constructed here https://github.com/LBNL-UCB-STI/activitysim-configs-sfbay/blob/main/configs/trip_mode_choice_annotate_trips_preprocessor.csv#L48-L50 and here https://github.com/LBNL-UCB-STI/activitysim-configs-sfbay/blob/a4067f30c4dec02e764665873f3b7e825cd3f2a9/configs/tour_mode_choice_annotate_choosers_preprocessor.csv#L35-L40. We'd need to change the existing logic, e.g.

odt_skims['RH_SOLO_WAIT'] (1.0 - odt_skims['RH_SOLO_REJECTIONPROB']) + TNC_missed_trip_penalty odt_skims['RH_SOLO_REJECTIONPROB']

to something that accounts for the requires_wheelchair attribute, so e.g. something like

np.where(df.requires_wheelchair,odt_skims['RH_SOLO_WHEELCHAIR_WAIT'] (1.0 - odt_skims['RH_SOLO_WHEELCHAIR_REJECTIONPROB']) + TNC_missed_trip_penalty odt_skims['RH_SOLO_WHEELCHAIR_REJECTIONPROB'],odt_skims['RH_SOLO_WAIT'] (1.0 - odt_skims['RH_SOLO_REJECTIONPROB']) + TNC_missed_trip_penalty odt_skims['RH_SOLO_REJECTIONPROB'])

I'm hoping that requires_wheelchair will automatically make it over to df, but it might take some additional work if not. @nazanin87 https://github.com/nazanin87

• Update PILATES to create separate wheelchair / non-wheelchair skims. @zneedell https://github.com/zneedell
• Set up PILATES scenarios with different fleet penetration of accessible vehicles. @nazanin87 https://github.com/nazanin87 with help from @cristian-poliziani https://github.com/cristian-poliziani

— Reply to this email directly, view it on GitHub https://github.com/LBNL-UCB-STI/PILATES/issues/25, or unsubscribe https://github.com/notifications/unsubscribe-auth/AYPABKCG4FF2HETIWRAHQCTVZZ623ANCNFSM566EDPDA . You are receiving this because you were mentioned.Message ID: @.***>

[zneedell]

Draft PR that should allow this on the skims end in PILATES: https://github.com/LBNL-UCB-STI/PILATES/pull/26

[cristian-poliziani]

That sounds great, thank you Zach for putting together all that and @Nazanin Rezaei @.***> feel free to write to me or plan a meeting with me when needed.

Cristian

On Thu, Aug 18, 2022 at 11:34 AM Zach Needell @.***> wrote:

Here's a list of changes that need to happen for us to successfully reproduce the wheelchair accessibility runs within PILATES:

• Add wheelchair attribute to ActivitySim inputs. The easiest way to do this is probably to just modify the persons table in the custom_mpo_06197001_model_data.h5 urbansim data input file to include a wheelchair attribute (important to make sure we do this for the correct year). @nazanin87 https://github.com/nazanin87. We'll then need to upload this new file to the AWS AMI we're using to launch these runs ( @cristian-poliziani https://github.com/cristian-poliziani can help with this).
• Make sure wheelchair attribute is visible to ActivitySim in PILATES. Probably will be true by default if it's in the h5 file but worth confirming. @zneedell https://github.com/zneedell
• Change ASim config to use different wait times depending on disability status. Let's create a new branch of the asim config repo https://github.com/LBNL-UCB-STI/activitysim-configs-sfbay where we modify the wait times in the trip and tour mode choice. I think the first step is to create a target requires_wheelchair by adding a line to the annotate_persons script here https://github.com/LBNL-UCB-STI/activitysim-configs-sfbay/blob/main/configs/annotate_persons.csv. I think this should make the boolean requires_wheelchair attribute available whenever we're doing mode choice. The wait times are constructed here https://github.com/LBNL-UCB-STI/activitysim-configs-sfbay/blob/main/configs/trip_mode_choice_annotate_trips_preprocessor.csv#L48-L50 and here https://github.com/LBNL-UCB-STI/activitysim-configs-sfbay/blob/a4067f30c4dec02e764665873f3b7e825cd3f2a9/configs/tour_mode_choice_annotate_choosers_preprocessor.csv#L35-L40. We'd need to change the existing logic, e.g.

odt_skims['RH_SOLO_WAIT'] (1.0 - odt_skims['RH_SOLO_REJECTIONPROB']) + TNC_missed_trip_penalty odt_skims['RH_SOLO_REJECTIONPROB']

to something that accounts for the requires_wheelchair attribute, so e.g. something like

np.where(df.requires_wheelchair,odt_skims['RH_SOLO_WHEELCHAIR_WAIT'] (1.0 - odt_skims['RH_SOLO_WHEELCHAIR_REJECTIONPROB']) + TNC_missed_trip_penalty odt_skims['RH_SOLO_WHEELCHAIR_REJECTIONPROB'],odt_skims['RH_SOLO_WAIT'] (1.0 - odt_skims['RH_SOLO_REJECTIONPROB']) + TNC_missed_trip_penalty odt_skims['RH_SOLO_REJECTIONPROB'])

I'm hoping that requires_wheelchair will automatically make it over to df, but it might take some additional work if not. @nazanin87 https://github.com/nazanin87

• Update PILATES to create separate wheelchair / non-wheelchair skims. @zneedell https://github.com/zneedell
• Set up PILATES scenarios with different fleet penetration of accessible vehicles. @nazanin87 https://github.com/nazanin87 with help from @cristian-poliziani https://github.com/cristian-poliziani

— Reply to this email directly, view it on GitHub https://github.com/LBNL-UCB-STI/PILATES/issues/25, or unsubscribe https://github.com/notifications/unsubscribe-auth/AYPABKCG4FF2HETIWRAHQCTVZZ623ANCNFSM566EDPDA . You are receiving this because you were mentioned.Message ID: @.***>