Trip Scheduling (#51)

* Differentiate examples between quality and validity of example models (ActivitySim#571) * change examples that are no longer tied to an agency to fictional place names * change name of full example also * add back missing output dir * restore test output dir also * more empty dirs that got lost * clean up docs * example_mtc -> prototype_mtc * Prototype MTC extended * add all the ignored files * add test output dirs * remove superfluous example_test * prototype_sf * prototype_arc * prototype_marin * move dirs * psrc * semcog * sandag_xborder * placeholder_sandag * placeholder_multiple_zone * no more coveralls * repair docs * clean up example names * black and isort (ActivitySim#578) * black and isort * stop emitting output dir it fails later tests * trace files in nested directories because windows * swap files for xborder model * repair ARC MP * print checksum even when not used * add hashes for sandag_xborder_full * fix dtype in university hack * fix persons to match tours * repair ARC * Bump version: 1.0.4 → 1.1.0 * Added mwcog small area * missed one edit * reformatting * Update README.MD * call as module * github actions tests * pre commit hooks * pyproject toml * limit numpy * docs for mamba instead of conda * ignore generated files * add nbmake to test env * fix dupe line * repair test multiple zones for github actions * publish develop docs * fix docbuild env * Update .travis.yml * fixed testing files * fixed testing files (again) * fixed test script again * publish docs to branch name * updated testing scripts (note shorter travis script for now) * fixed slash (windows vs. linux testing issue) * added output folders * updated travis script to run all tests, should pass * docs cleaning * docs re-style * rebuild * dynamic versioning docs * version switcher * blacken * fix switcher url * fix conf * switcher update * master to main * deployment actions * actions * build wheel * fix for testpypi * blacken * manual switcher * branch docs service [makedocs] * syntax [makedocs] * travis depends * checkout v3, fix versioning in docs * only build develop docs once * failsafe version * documentation repairs * python-simplified-semver * front cards * end testing w travis * add mwcog test to gh-actions * add mwcog to docs * point to data not copy it * sort dependencies * req sh 2.2.4 * account for variance across platforms in trip dest est * copy bike skims for sandag test * param on rtol * fix sandag_2 test files * fix test file names * added pre-processor option to trip scheduling * trip scheduling relative mode initial commit * moved everything to mwcog example * adding output analysis notebook * adding additional segmentation * testing and documentation * blacken * not assuming scheduling mode is set * fixing merge trip scheduling * still fixing merge * reverting regression trips * fixing bad merge and updating test Co-authored-by: Jeffrey Newman <jeffnewman@camsys.com> Co-authored-by: Andrew Rohne <andrew@siliconcreek.net>
SANDAG · Dec 21, 2022 · 9506f16 · 9506f16
1 parent 7386a4d
commit 9506f16
Show file tree

Hide file tree

Showing 9 changed files with 414 additions and 62 deletions.
diff --git a/activitysim/abm/models/non_mandatory_scheduling.py b/activitysim/abm/models/non_mandatory_scheduling.py
@@ -14,8 +14,6 @@
 from .util.school_escort_tours_trips import create_pure_school_escort_tours
 
 from .util.vectorize_tour_scheduling import vectorize_tour_scheduling
-from activitysim.core.util import assign_in_place
-
 
 logger = logging.getLogger(__name__)
 DUMP = False

diff --git a/activitysim/abm/models/trip_scheduling.py b/activitysim/abm/models/trip_scheduling.py
@@ -8,7 +8,15 @@
 
 from activitysim.abm.models.util import estimation
 from activitysim.abm.models.util.trip import cleanup_failed_trips, failed_trip_cohorts
-from activitysim.core import chunk, config, inject, logit, pipeline, tracing
+from activitysim.core import (
+    chunk,
+    config,
+    inject,
+    logit,
+    pipeline,
+    tracing,
+    expressions,
+)
 from activitysim.core.util import reindex
 
 from .util.school_escort_tours_trips import split_out_school_escorting_trips
@@ -36,13 +44,15 @@
 
 DEPARTURE_MODE = "departure"
 DURATION_MODE = "stop_duration"
+RELATIVE_MODE = "relative"
 PROBS_JOIN_COLUMNS_DEPARTURE_BASED = [
     "primary_purpose",
     "outbound",
     "tour_hour",
     "trip_num",
 ]
 PROBS_JOIN_COLUMNS_DURATION_BASED = ["outbound", "stop_num"]
+PROBS_JOIN_COLUMNS_RELATIVE_BASED = ["outbound", "periods_left"]
 
 
 def set_tour_hour(trips, tours):
@@ -181,6 +191,7 @@ def schedule_trips_in_leg(
     failfix = model_settings.get(FAILFIX, FAILFIX_DEFAULT)
     depart_alt_base = model_settings.get("DEPART_ALT_BASE", 0)
     scheduling_mode = model_settings.get("scheduling_mode", "departure")
+    preprocessor_settings = model_settings.get("preprocessor", None)
 
     if scheduling_mode == "departure":
         probs_join_cols = model_settings.get(
@@ -190,10 +201,14 @@ def schedule_trips_in_leg(
         probs_join_cols = model_settings.get(
             "probs_join_cols", PROBS_JOIN_COLUMNS_DURATION_BASED
         )
+    elif scheduling_mode == "relative":
+        probs_join_cols = model_settings.get(
+            "probs_join_cols", PROBS_JOIN_COLUMNS_RELATIVE_BASED
+        )
     else:
         logger.error(
             "Invalid scheduling mode specified: {0}.".format(scheduling_mode),
-            "Please select one of ['departure', 'stop_duration'] and try again.",
+            "Please select one of ['departure', 'stop_duration', 'relative'] and try again.",
         )
 
     # logger.debug("%s scheduling %s trips" % (trace_label, trips.shape[0]))
@@ -232,18 +247,35 @@ def schedule_trips_in_leg(
         ADJUST_NEXT_DEPART_COL = "latest"
     trips.next_trip_id = trips.next_trip_id.where(~is_final, NO_TRIP_ID)
 
+    network_los = inject.get_injectable("network_los")
+    locals_dict = {"network_los": network_los}
+    locals_dict.update(config.get_model_constants(model_settings))
+
     first_trip_in_leg = True
     for i in range(trips.trip_num.min(), trips.trip_num.max() + 1):
 
-        if outbound or scheduling_mode == DURATION_MODE:
+        nth_trace_label = tracing.extend_trace_label(trace_label, "num_%s" % i)
+
+        # - annotate trips
+        if preprocessor_settings:
+            expressions.assign_columns(
+                df=trips,
+                model_settings=preprocessor_settings,
+                locals_dict=locals_dict,
+                trace_label=nth_trace_label,
+            )
+
+        if (
+            outbound
+            or (scheduling_mode == DURATION_MODE)
+            or (scheduling_mode == RELATIVE_MODE)
+        ):
             # iterate in ascending trip_num order
             nth_trips = trips[trips.trip_num == i]
         else:
             # iterate over inbound trips in descending trip_num order, skipping the final trip
             nth_trips = trips[trips.trip_num == trips.trip_count - i]
 
-        nth_trace_label = tracing.extend_trace_label(trace_label, "num_%s" % i)
-
         choices = ps.make_scheduling_choices(
             nth_trips,
             scheduling_mode,
@@ -265,6 +297,12 @@ def schedule_trips_in_leg(
             )
             choices = choices.fillna(trips[ADJUST_NEXT_DEPART_COL])
 
+        if scheduling_mode == RELATIVE_MODE:
+            # choices are relative to the previous departure time
+            choices = nth_trips.earliest + choices
+            # need to update the departure time based on the choice
+            update_tour_earliest(trips, choices)
+
         # adjust allowed depart range of next trip
         has_next_trip = nth_trips.next_trip_id != NO_TRIP_ID
         if has_next_trip.any():
@@ -437,9 +475,9 @@ def trip_scheduling(trips, tours, chunk_size, trace_hh_id):
         ]
         estimator.write_choosers(trips_df[chooser_cols_for_estimation])
 
-    probs_spec = pd.read_csv(
-        config.config_file_path("trip_scheduling_probs.csv"), comment="#"
-    )
+    probs_spec_file = model_settings.get("PROBS_SPEC", "trip_scheduling_probs.csv")
+    logger.debug(f"probs_spec_file: {config.config_file_path(probs_spec_file)}")
+    probs_spec = pd.read_csv(config.config_file_path(probs_spec_file), comment="#")
     # FIXME for now, not really doing estimation for probabilistic model - just overwriting choices
     # besides, it isn't clear that named coefficients would be helpful if we had some form of estimation
     # coefficients_df = simulate.read_model_coefficients(model_settings)
@@ -499,6 +537,11 @@ def trip_scheduling(trips, tours, chunk_size, trace_hh_id):
                 failed = choices.reindex(trips_chunk.index).isnull()
                 logger.info("%s %s failed", trace_label_i, failed.sum())
 
+                if (failed.sum() > 0) & (
+                    model_settings.get("scheduling_mode") == "relative"
+                ):
+                    raise RuntimeError("failed trips with relative scheduling mode")
+
                 if not is_last_iteration:
                     # boolean series of trips whose leg scheduling failed
                     failed_cohorts = failed_trip_cohorts(trips_chunk, failed)

diff --git a/activitysim/abm/models/util/probabilistic_scheduling.py b/activitysim/abm/models/util/probabilistic_scheduling.py
@@ -172,10 +172,22 @@ def _preprocess_scheduling_probs(
         )
     elif scheduling_mode == "stop_duration":
         chooser_probs = _preprocess_stop_duration_probs(choosers)
+    elif scheduling_mode == "relative":
+        # creating a dataframe with just the trip_id as index and alternatives as columns
+        probs_cols = [
+            c
+            for c in probs_spec.columns
+            if ((c not in probs_join_cols) & (c.isnumeric()))
+        ]
+        chooser_probs = choosers.loc[:, probs_cols]
+        chooser_probs = chooser_probs.div(chooser_probs.sum(axis=1), axis=0)
+        assert (
+            ~chooser_probs.isna().values.any()
+        ), f"Missing probabilities for trips \n {chooser_probs[chooser_probs.isna().any(axis=1)].index}"
     else:
         logger.error(
             "Invalid scheduling mode specified: {0}.".format(scheduling_mode),
-            "Please select one of ['departure', 'stop_duration'] and try again.",
+            "Please select one of ['departure', 'stop_duration', 'relative'] and try again.",
         )
 
     # probs should sum to 1 with residual probs resulting in choice of 'fail'
@@ -196,6 +208,12 @@ def _postprocess_scheduling_choices(
     # convert alt choice index to depart time (setting failed choices to -1)
     failed = choices == choice_cols.get_loc("fail")
 
+    if scheduling_mode == "relative":
+        if failed.any():
+            RuntimeError(
+                f"Failed trips in realtive mode for {failed.sum()} trips: {choosers[failed]}"
+            )
+
     # For the stop duration-based probabilities, the alternatives are offsets that
     # get applied to trip-specific departure and arrival times, so depart_alt_base
     # is a column/series rather than a scalar.
@@ -334,7 +352,11 @@ def make_scheduling_choices(
     if failed.any():
         choices = choices[~failed]
 
-    if all([check_col in choosers_df.columns for check_col in ["earliest", "latest"]]):
+    if all(
+        [check_col in choosers_df.columns for check_col in ["earliest", "latest"]]
+    ) & (scheduling_mode != "relative"):
+        # check to make sure choice does not come before previously scheduled trip or after end of tour
+        # does not apply if choices are relative to previous trip depart
         assert (choices >= choosers_df.earliest[~failed]).all()
         assert (choices <= choosers_df.latest[~failed]).all()
 

diff --git a/activitysim/examples/prototype_mwcog/configs/trip_scheduling.yaml b/activitysim/examples/prototype_mwcog/configs/trip_scheduling.yaml
@@ -1,10 +1,27 @@
 
 # int to add to probs column index to get time period it represents.
 # e.g. depart_alt_base = 5 means first column (column 0) represents period 5
-DEPART_ALT_BASE: 1
+DEPART_ALT_BASE: 0
 
 MAX_ITERATIONS: 100
 
 #FAILFIX: drop_and_cleanup
 FAILFIX: choose_most_initial
 
+# --- relative trip scheduling settings
+PROBS_SPEC: trip_scheduling_probs_purpose_stops.csv
+
+scheduling_mode: relative
+
+probs_join_cols: 
+  - periods_left_min
+  - periods_left_max
+  - outbound
+  - tour_purpose_grouped
+  - half_tour_stops_remaining_grouped
+
+preprocessor:
+  SPEC: trip_scheduling_preprocessor
+  DF: choosers
+  TABLES:
+    - tours
diff --git a/activitysim/examples/prototype_mwcog/configs/trip_scheduling_preprocessor.csv b/activitysim/examples/prototype_mwcog/configs/trip_scheduling_preprocessor.csv
@@ -0,0 +1,8 @@
+Description,Target,Expression
+,periods_left,(df.latest - df.earliest)
+# binning the periods differently for inbound and outbound
+,periods_left_min,"np.where(df['outbound'], periods_left.clip(upper=25), periods_left.clip(upper=34))"
+,periods_left_max,"np.where(((periods_left >= 25) & (df['outbound'])) | ((periods_left >= 34) & (~df['outbound'])), 47, periods_left)"
+,tour_purpose,"reindex(tours.tour_type, df.tour_id)"
+,tour_purpose_grouped,"np.where(tour_purpose.isin(['work','school','univ']), 'mand', 'non_mand')"
+,half_tour_stops_remaining_grouped,(df.trip_count - df.trip_num).clip(upper=1)