have about half of the mandatory scheduling model implemented

need to ask a few questions about the time windowing

example/configs/settings.yaml

@@ -2,6 +2,8 @@ store: mtc_asim.h5
 
 # area_types less than this are considered urban
 urban_threshold: 4
+cbd_threshold: 2
+rural_threshold: 6
 
 households_sample_size: 1000
 

example/configs/tour_departure_and_duration.csv

@@ -1 +1 @@
-Description,Expression,workFree-flow round trip auto time shift effects - departure,start,0.00114Free-flow round trip auto time shift effects - duration,duration,0.00221Part-time worker departure shift effects,(ptype == 2) * start,0.06736Non-working adult duration shift effects,(ptype == 4) * duration,0.1207University student departure shift effects,(ptype == 3) * start,0.05747Household income departure shift effects,income_in_thousands * start,0.00021Destination in CBD departure shift effects,start,0.04717Destination in CBD duration shift effects,duration,0.08679subsequent tour must start after previous tour ends,start < endOfPreviousTour,999First of 2+ work tours departure shift effects,start,0.3033First of 2+ work tours duration shift effects,duration,0.1861Subsequent 2+ work departure tours shift effects,start,0.5381Subsequent 2+ work duration tours shift effects,duration,0.3174Household income -- Early departure interaction,start < 6,0.4854Household income -- Late arrival interaction,end > 22,0.3839Destination in CBD -- Early departure interaction,start < 6,0.4566Destination in CBD -- Late arrival interaction,end > 22,0.2334Rural household -- Early departure interaction,start < 6,0.4039Rural household -- Late arrival interaction,end > 22,0.3451Full-time worker -- duration < 9 hours interaction,duration < 9,1.257Full-time worker -- 10 to 12 departure interaction,(start > 9)*(start < 13),0.5182Part-time worker -- 13 to 15 arrival interaction,(end > 12)*(end < 16),0.5433First of 2+ work tours- duration<8 hrs,duration < 8,1.98Subsequent of 2+ work tours- duration<8 hrs,duration < 8,2.582Work+school tours by worker- duration<8 hrs,duration < 8,0.9126School+work tours by student- duration<8 hrs,duration < 8,2.582Mode Choice Logsum,@@modeChoiceLogsumAlt,1.027Previously-scheduled tour ends in this departure hour,@@prevTourEndsThisDepartureHourAlt,0.8935Previously-scheduled tour begins in this arrival hour,@@prevTourBeginsThisArrivalHourAlt,1.334Adjacent window exists before this departure hour - first tour interaction,@@adjWindowBeforeThisHourAlt,0.1771Adjacent window exists afetr this arrival hour - first tour interaction,@@adjWindowAfterThisHourAlt,0.3627Adjacent window exists before this departure hour - second+ tour interaction,@@adjWindowBeforeThisHourAlt,0.2123Adjacent window exists after this arrival hour - second+ tour interaction,@@adjWindowAfterThisHourAlt,0.1012Remaining work/school tours to be scheduled / number of unscheduled hours,1 / @@remainingHoursAvailableAlt,18.68Departure Constants -- Early (up to 5),1,0.95273Departure Constants -- AM peak 1 (6),1,0.61618Departure Constants -- AM peak 2 (7),1,0Departure Constants -- AM peak 3 (8),1,0.25471Departure Constants -- AM peak 4 (9),1,1.25135Departure Constants -- Midday 1 (10 to 12),1,1.70587Departure Constants -- Midday 2 (13 to 15),1,1.69357Departure Constants -- PM peak (16 to 18),1,1.43999Departure Constants -- Evening (19 to 21),1,1.61051Departure Constants -- Late (22 and later),1,2.88342Arrival Constants -- Early (up to 6),1,0Arrival Constants -- AM peak (7 to 9),1,1.85452Arrival Constants -- Midday 1 (10 to 12),1,0.49597Arrival Constants -- Midday 2 (13 to 14),1,0.37855Arrival Constants -- PM peak 1 (15),1,0Arrival Constants -- PM peak 2 (16),1,0.27608Arrival Constants -- PM peak 3 (17),1,0.69959Arrival Constants -- PM peak 4 (18),1,0.79929Arrival Constants -- Evening (19 to 21),1,0.10357Arrival Constants -- Late (22 and later),1,0.96596Duration Constants -- 0 to 2 hours,1,2.52827Duration Constants -- 3 to 4 hours,1,0.91897Duration Constants -- 5 to 6 hours,1,0.71855Duration Constants -- 7 to 8 hours,1,0.13962Duration Constants -- 9 hours,1,0.05571Duration Constants -- 10 hours,1,0Duration Constants -- 11 hours,1,0.3478Duration Constants -- 12 to 13 hours,1,1.00822Duration Constants -- 14 to 18 hours,1,1.70186
+Description,Expression,workFree-flow round trip auto time shift effects - departure,roundtrip_auto_time_to_work * start,0.00114Free-flow round trip auto time shift effects - duration,roundtrip_auto_time_to_work * duration,0.00221Part-time worker departure shift effects,(ptype == 2) * start,0.06736Non-working adult duration shift effects,(ptype == 4) * duration,0.1207University student departure shift effects,(ptype == 3) * start,0.05747Household income departure shift effects,income_in_thousands * start,0.00021Destination in CBD departure shift effects,workplace_in_cbd * start,0.04717Destination in CBD duration shift effects,workplace_in_cbd * duration,0.08679subsequent tour must start after previous tour ends,(start < end_of_previous_tour) & (tour_num == 2),-999First of 2+ work tours departure shift effects,(tour_num == 1) * start,0.3033First of 2+ work tours duration shift effects,(tour_num == 1) * duration,0.1861Subsequent 2+ work departure tours shift effects,(tour_num == 2) * start,0.5381Subsequent 2+ work duration tours shift effects,(tour_num == 2) * duration,0.3174Household income -- Early departure interaction,(income_in_thousands > 100) & (start < 6),0.4854Household income -- Late arrival interaction,(income_in_thousands > 100) & (end > 22),0.3839Destination in CBD -- Early departure interaction,workplace_in_cbd & (start < 6),0.4566Destination in CBD -- Late arrival interaction,workplace_in_cbd & (end > 22),0.2334Rural household -- Early departure interaction,home_is_rural & (start < 6),0.4039Rural household -- Late arrival interaction,home_is_rural & (end > 22),0.3451Full-time worker -- duration < 9 hours interaction,(ptype == 1) & (duration < 9),1.257Full-time worker -- 10 to 12 departure interaction,(ptype == 1) & (start > 9) & (start < 13),0.5182Part-time worker -- 13 to 15 arrival interaction,(ptype == 2) & (end > 12) & (end < 16),0.5433First of 2+ work tours- duration<8 hrs,(tour_num == 1) & (duration < 8),1.98Subsequent of 2+ work tours- duration<8 hrs,(tour_num == 2) & (duration < 8),2.582Work+school tours by worker- duration<8 hrs,(mandatory_tour_frequency == 'work_and_school') & is_worker & (duration < 8),0.9126School+work tours by student- duration<8 hrs,(mandatory_tour_frequency == 'work_and_school') & is_student & (duration < 8),2.582Mode Choice Logsum,mode_choice_logsum,1.027Previously-scheduled tour ends in this departure hour,prev_tour_end == start,0.8935Previously-scheduled tour begins in this arrival hour,prev_tour_begin == end,1.334Adjacent window exists before this departure hour - first tour interaction,@@adjWindowBeforeThisHourAlt,0.1771Adjacent window exists afetr this arrival hour - first tour interaction,@@adjWindowAfterThisHourAlt,0.3627Adjacent window exists before this departure hour - second+ tour interaction,@@adjWindowBeforeThisHourAlt,0.2123Adjacent window exists after this arrival hour - second+ tour interaction,@@adjWindowAfterThisHourAlt,0.1012Remaining work/school tours to be scheduled / number of unscheduled hours,1 / @@remainingHoursAvailableAlt,18.68Departure Constants -- Early (up to 5),start < 6,0.95273Departure Constants -- AM peak 1 (6),start == 6,0.61618Departure Constants -- AM peak 2 (7),start == 7,0Departure Constants -- AM peak 3 (8),start == 8,0.25471Departure Constants -- AM peak 4 (9),start == 9,1.25135Departure Constants -- Midday 1 (10 to 12),(start > 9) & (start < 13),1.70587Departure Constants -- Midday 2 (13 to 15),(start > 12) & (start < 16),1.69357Departure Constants -- PM peak (16 to 18),(start > 15) & (start < 19),1.43999Departure Constants -- Evening (19 to 21),(start > 18) & (start < 22),1.61051Departure Constants -- Late (22 and later),start > 21,2.88342Arrival Constants -- Early (up to 6),end < 7,0Arrival Constants -- AM peak (7 to 9),(end > 6) & (end < 10),1.85452Arrival Constants -- Midday 1 (10 to 12),(end > 9) & (end < 13),0.49597Arrival Constants -- Midday 2 (13 to 14),(end > 12) & (end < 15),0.37855Arrival Constants -- PM peak 1 (15),end == 15,0Arrival Constants -- PM peak 2 (16),end == 16,0.27608Arrival Constants -- PM peak 3 (17),end == 17,0.69959Arrival Constants -- PM peak 4 (18),end == 18,0.79929Arrival Constants -- Evening (19 to 21),(end > 18) & (end < 22),0.10357Arrival Constants -- Late (22 and later),end > 21,0.96596Duration Constants -- 0 to 2 hours,duration < 3,2.52827Duration Constants -- 3 to 4 hours,(duration > 2) & (duration < 5),0.91897Duration Constants -- 5 to 6 hours,(duration > 4) & (duration < 7),0.71855Duration Constants -- 7 to 8 hours,(duration > 6) & (duration < 9),0.13962Duration Constants -- 9 hours,duration == 9,0.05571Duration Constants -- 10 hours,duration == 10,0Duration Constants -- 11 hours,duration == 11,0.3478Duration Constants -- 12 to 13 hours,(duration > 11) & (duration < 14),1.00822Duration Constants -- 14 to 18 hours,(duration > 13) & (duration < 19),1.70186

example/models.py

@@ -146,7 +146,7 @@ def destination_choice_spec():
 @sim.table()
 def tour_departure_and_duration_spec():
     f = os.path.join('configs', 'tour_departure_and_duration.csv')
-    return asim.read_model_spec(f, stack=False).head(4)
+    return asim.read_model_spec(f, stack=False)
 
 
 """
@@ -381,6 +381,55 @@ def non_mandatory_tours(persons,
               cast_index=True, onto_on='person_id')
 
 
+"""
+This does the same as the above but for mandatory tours.  Ending format is
+the same as in the comment above except trip types are "work" and "school"
+"""
+
+@sim.table()
+def mandatory_tours(persons):
+
+    persons = persons.to_frame(columns=["mandatory_tour_frequency",
+                                        "is_worker"])
+    persons = persons[~persons.mandatory_tour_frequency.isnull()]
+
+    tours = []
+    # this is probably easier to do in non-vectorized fashion (at least for now)
+    for key, row in persons.iterrows():
+
+        mtour = row.mandatory_tour_frequency
+        is_worker = row.is_worker
+
+        # 1 work trip
+        if mtour == "work1":
+            tours += [(key, "work", 1)]
+        # 2 work trips
+        elif mtour == "work2":
+            tours += [(key, "work", 1), (key, "work", 2)]
+        # 1 school trip
+        elif mtour == "school1":
+            tours += [(key, "school", 1)]
+        # 2 school trips
+        elif mtour == "school2":
+            tours += [(key, "school", 1), (key, "school", 2)]
+        # 1 work and 1 school trip
+        elif mtour == "work_and_school":
+            if is_worker:
+                # is worker, work trip goes first
+                tours += [(key, "work", 1), (key, "school", 2)]
+            else:
+                # is student, work trip goes second
+                tours += [(key, "school", 1), (key, "work", 2)]
+        else:
+            assert 0
+
+    return pd.DataFrame(tours, columns=["person_id", "trip_type", "tour_num"])
+
+
+sim.broadcast('persons', 'mandatory_tours',
+              cast_index=True, onto_on='person_id')
+
+
 """
 Given the tour generation from the above, each tour needs to have a
 destination, so in this case tours are the choosers (with the associated
@@ -450,33 +499,60 @@ def destination_choice(non_mandatory_tours,
 
 
 @sim.model()
-def mandatory_tour_departure_and_duration(persons,
+def mandatory_tour_departure_and_duration(mandatory_tours,
+                                          persons,
                                           households,
                                           land_use,
                                           tour_departure_and_duration_alts,
                                           tour_departure_and_duration_spec):
 
-    print tour_departure_and_duration_spec.to_frame().tail()
+    choosers = sim.merge_tables(mandatory_tours.name, tables=[mandatory_tours,
+                                                              persons,
+                                                              households,
+                                                              land_use])
 
-    choosers = sim.merge_tables(persons.name, tables=[persons,
-                                                      households,
-                                                      land_use])
+    print "Running %d mandatory tour scheduling choices" % len(choosers)
 
-    # filter based on results of CDAP
-    choosers = choosers[choosers.cdap_activity == 'M']
-    print "%d persons run for workplace tour departure and duration model" % \
-          len(choosers)
+    # assert there's only a first or second mandatory tour - that's a basic
+    # assumption of this model formulation right now
+    assert choosers.tour_num.isin([1, 2]).value_counts()[True] == len(choosers)
 
-    choices, _ = \
-        asim.simple_simulate(choosers,
-                             tour_departure_and_duration_alts.to_frame(),
-                             tour_departure_and_duration_spec['work'],
-                             mult_by_alt_col=False)
+    first_tours = choosers[choosers.tour_num == 1]
+    second_tours = choosers[choosers.tour_num == 2]
+
+    spec = tour_departure_and_duration_spec.work.head(27)
+    alts = tour_departure_and_duration_alts.to_frame()
+
+    print choosers.mandatory_tour_frequency.value_counts()
+    print spec
+
+    # this is a bit odd to python - we can't run through in for loops for
+    # performance reasons - we first have to do a pass for the first tours and
+    # then for the second tours - this is mainly because the second tours are
+    # dependent on the first tours' scheduling
+
+    print "Running %d mandatory first tour choices" % len(first_tours)
+
+    alts["end_of_previous_tour"] = -1
+
+    # FIXME - a note to remember that this also needs the mode choice logsum
+    alts["mode_choice_logsum"] = 0
+
+    first_choices, _ = \
+        asim.simple_simulate(first_tours, alts, spec, mult_by_alt_col=False)
+
+    print "Running %d mandatory second tour choices" % len(second_tours)
+
+    # FIXME need to set end_of_previous_tour to the ends computed above
+    second_choices, _ = \
+        asim.simple_simulate(second_tours, alts, spec, mult_by_alt_col=False)
+
+    choices = pd.concat([first_choices, second_choices])
 
     # as with non-mandatory tour generation, this stores the INDEX of
     # the alternative in the tour_departure and_duration_alts dataframe -
     # to actually use it we'll have ot go back and grab the start and end times
-    print "Choices:\n", choices.value_counts()
+    print "Choices:\n", choices.describe()
 
     sim.add_column("persons", "mandatory_tour_departure_and_duration", choices)
 
@@ -568,6 +644,12 @@ def home_is_urban(households, land_use, settings):
     return s < settings['urban_threshold']
 
 
+@sim.column('households')
+def home_is_rural(households, land_use, settings):
+    s = usim_misc.reindex(land_use.area_type, households.home_taz)
+    return s > settings['rural_threshold']
+
+
 @sim.column('households')
 def car_sufficiency(households, persons):
     return households.auto_ownership - persons.household_id.value_counts()
@@ -782,6 +864,16 @@ def under16_not_at_school(persons):
             persons.cdap_activity.isin(["N", "H"]))
 
 
+@sim.column("persons")
+def is_worker(persons):
+    return persons.employed_cat.isin(['full', 'part'])
+
+
+@sim.column("persons")
+def is_student(persons):
+    return persons.student_cat.isin(['high', 'college'])
+
+
 @sim.column("persons")
 def workplace_taz(persons):
     # FIXME this is really because we ask for ALL columns in the persons data
@@ -838,3 +930,10 @@ def roundtrip_auto_time_to_school(persons, sovam_skim, sovmd_skim):
                      sovmd_skim.get(persons.school_taz,
                                     persons.home_taz),
                      index=persons.index)
+
+
+@sim.column('persons')
+def workplace_in_cbd(persons, land_use, settings):
+    s = usim_misc.reindex(land_use.area_type, persons.workplace_taz)
+    return s < settings['cbd_threshold']
+