Spreadout weekends (#39)

Spreadout weekends

config/sample_config.json

@@ -1 +1 @@
-{"Michael": {"email": "[email protected]", "divisions": {"A": {"min": 0, "max": 100}, "B": {"min": 0, "max": 100}}}, "Jane": {"email": "", "divisions": {"A": {"min": 0, "max": 100}, "B": {"min": 0, "max": 100}}}, "Bob": {"email": "[email protected]", "divisions": {"A": {"min": 0, "max": 100}, "B": {"min": 0, "max": 100}}}, "Alice": {"email": "[email protected]", "divisions": {"A": {"min": 0, "max": 100}, "B": {"min": 0, "max": 100}}}, "James": {"name": "James", "email": "", "divisions": {"A": {"min": 0, "max": 100}, "B": {"min": 0, "max": 100}}}, "Alex": {"name": "Alex", "email": "", "divisions": {"A": {"min": 1, "max": 100}}}}
+{"Michael": {"email": "[email protected]", "divisions": {"A": {"min": 0, "max": 100}, "B": {"min": 0, "max": 100}}}, "Jane": {"email": "", "divisions": {"A": {"min": 0, "max": 100}, "B": {"min": 0, "max": 100}}}, "Bob": {"email": "[email protected]", "divisions": {"A": {"min": 0, "max": 100}, "B": {"min": 0, "max": 100}}}, "Alice": {"email": "[email protected]", "divisions": {"A": {"min": 0, "max": 100}, "B": {"min": 0, "max": 100}}}}

paper/scheduler.tex

@@ -7,6 +7,7 @@
 \usepackage{hyperref}
 \usepackage{bbm}
 \usepackage{physics}
+\usepackage{xcolor}
 
 \newcommand{\mc}{\mathcal}
 \newcommand{\bb}{\mathbb}
@@ -44,7 +45,9 @@ \subsection{Variables}
 		X_{c, b, d} \in \{0, 1\}: &\text{ clinician $c$ covers block $b$ for division $d$} \\
 		Y_{c, w} \in \{0, 1\}: &\text{ clinician $c$ covers weekend $w$}
 	\end{align}
-Note that weekends are not distinguished according to division, since the clinician assigned covers all divisions.
+Note that weekends are not distinguished according to division, since the clinician assigned covers all divisions. \\
+
+\textcolor{red}{\textbf{IMPORTANT: This assumes that each clinician works at each of the divisions $1, \ldots, D$. Some of the formulas below are not directly applicable to the case where some clinician only works a subset of those divisions}}
 
 \subsection{Constraints} \label{constraints}
 	\begin{itemize}
@@ -72,14 +75,14 @@ \subsection{Constraints} \label{constraints}
 			\begin{equation}
 				\sum_d \left(X_{c, b, d} + X_{c, b+1, d}\right) \leq 1 \text{ for each $c$ and each $b \leq B - 1$}
 			\end{equation}
-		\item A clinician cannot work two consecutive weekends
-			\begin{equation}
-				Y_{c, w} + Y_{c, w+1} \leq 1 \text{ for each $c$ and each $w \leq W - 1$}
-			\end{equation}
 		\item Prevent on-off-on-off-on block assignments (across all divisions) to improve the spread of assignments for any given clinician
 			\begin{equation}
 				\sum_d \left(X_{c, b, d} + X_{c, b + 2, d} + X_{c, b + 4, d}\right) \leq 2 \text{ for each $c$ and each $b \leq B - 4$}
 			\end{equation}
+		\item Improve spread of weekend assignments by preventing more than 1 assignment for a given clinician in a span of 4 weekends
+			\begin{equation}
+				Y_{c, w} + Y_{c, w+1} + Y_{c, w+2} + Y_{c, w+3} \leq 1 \text{ for each $c$ and each $w \leq W - 3$}
+			\end{equation}
 	\end{itemize}
 
 \subsection{Objectives} \label{objectives}

src/services/scheduler.py

@@ -408,23 +408,24 @@ def setup_problem(self, shuffle=False):
         self._build_clinician_variables(divisions, clinicians)
         self._build_coverage_constraints(divisions, clinicians)
         self._build_minmax_constraints(divisions)
-        self._build_consec_constraints(clinicians)
-        self._build_spread_constraints(clinicians)
+        self._build_consec_blocks_constraints(clinicians)
+        self._build_spread_blocks_constraints(clinicians)
+        self._build_spread_weekends_constraints(clinicians)
         self._build_longweekend_constraints(clinicians)
         self._build_weekend_constraints(clinicians)
         self._build_adjacency_variables(divisions)
-        appeasement_objs = self._build_appeasement_objectives(clinicians)
 
-        num_clin = len(clinicians)
-        num_div = len(divisions)
+        block_conflicts_obj = self._build_block_objective(clinicians)
+        weekend_conflicts_obj = self._build_weekend_objective(clinicians)
+        adjacency_obj = self._build_adjacency_objective(clinicians)
 
         # make sure to normalize objectives, and weigh them equally
         self.problem.setObjective(
-              (1 / 3) * (1 / (num_clin * self.num_blocks * num_div)) * appeasement_objs[0]
-            + (1 / 3) * (1 / (num_clin * self.num_weekends)) * appeasement_objs[1]
-            + (1 / 3) * (1 / (num_clin * self.num_blocks * num_div)) * self._build_adjacency_objective(clinicians)
+              (1 / 3) * (1 / len(block_conflicts_obj)) * block_conflicts_obj
+            + (1 / 3) * (1 / len(weekend_conflicts_obj)) * weekend_conflicts_obj
+            + (1 / 3) * (1 / len(adjacency_obj)) * adjacency_obj
         )
-
+        
     def _build_clinician_variables(self, divisions, clinicians):
         # create clinician BlockVariables
         for div in divisions:
@@ -473,7 +474,7 @@ def _build_minmax_constraints(self, divisions):
                 self.problem.add(sum_ <= max_)
                 self.problem.add(sum_ >= min_)
 
-    def _build_consec_constraints(self, clinicians):
+    def _build_consec_blocks_constraints(self, clinicians):
         for clinician in clinicians:
             for block_num in range(1, self.num_blocks):
                 # if a clinician works a given block, they should not work any
@@ -484,20 +485,10 @@ def _build_consec_constraints(self, clinicians):
                 )
                 self.problem.add(sum_ <= 1)
 
-            # at most 1 consecutive weekend of work
-            for week_num in range(1, self.num_weekends):
-                sum_ = pulp.lpSum(
-                    [_.get_var() for _ in clinician.get_weekend_vars(
-                        lambda x, week_num=week_num: x.week_num in (
-                            week_num, week_num + 1)
-                    )]
-                )
-                self.problem.add(sum_ <= 1)
-
-    def _build_spread_constraints(self, clinicians):
+    def _build_spread_blocks_constraints(self, clinicians):
+        # on-off-on-off-on constraint for block assignment
         # we need at least 5 consecutive blocks to implement this constraint
-        # on-off-on-off-on
-        if self.num_blocks <= 5: return
+        if self.num_blocks < 5: return
 
         for clinician in clinicians:
             for block_num in range(1, self.num_blocks - 3):
@@ -508,6 +499,21 @@ def _build_spread_constraints(self, clinicians):
                 )
                 self.problem.add(sum_ <= 2)
 
+    def _build_spread_weekends_constraints(self, clinicians):
+        # spreading out weekend assignments
+        # needs at least 4 weekends
+        if self.num_weekends < 4: return
+
+        for clinician in clinicians:
+            for week_num in range(1, self.num_weekends - 2):
+                # constraint: X_i + X_{i+1} + X_{i+2} + X_{i+3} <= 1
+                sum_ = pulp.lpSum(
+                    [_.get_var() for _ in clinician.get_weekend_vars(
+                        lambda x, week_num=week_num: x.week_num in list(range(week_num, week_num + 4))
+                    )]
+                )
+                self.problem.add(sum_ <= 1)
+
     def _build_longweekend_constraints(self, clinicians):
         if self.long_weekends:
             # (roughly) equal distribution of long weekends
@@ -587,7 +593,7 @@ def _build_adjacency_objective(self, clinicians):
             )
         return pulp.lpSum(adjacency_vars)
 
-    def _build_appeasement_objectives(self, clinicians):
+    def _build_weekend_objective(self, clinicians):
         wa_variables = []
         for clinician in clinicians:
             wa_variables.extend(
@@ -602,6 +608,9 @@ def _build_appeasement_objectives(self, clinicians):
                  ]
             )
 
+        return pulp.lpSum(wa_variables)
+
+    def _build_block_objective(self, clinicians):
         ba_variables = []
         for clinician in clinicians:
             ba_variables.extend(
@@ -616,10 +625,7 @@ def _build_appeasement_objectives(self, clinicians):
                  ]
             )
 
-        return (
-            pulp.lpSum(wa_variables),
-            pulp.lpSum(ba_variables)
-        )
+        return pulp.lpSum(ba_variables)
 
     def assign_schedule(self):
         """