diff --git a/backend/src/utils/optimizer.py b/backend/src/utils/optimizer.py
index 3bbe703..4857858 100644
--- a/backend/src/utils/optimizer.py
+++ b/backend/src/utils/optimizer.py
@@ -59,20 +59,25 @@ class Optimizer:
def init_ub_time(self, prob: pl.LpProblem, x: pl.LpVariable, L: int, landmarks: list[Landmark], max_time: int):
"""
- Initialize the objective function coefficients and inequality constraints.
- -> Adds 1 row of constraints
- -> Pre-allocates A_ub for the rest of the computations with L + (L*L-L)/2 rows
+ Initialize the objective function and inequality constraints for the linear program.
- This function computes the distances between all landmarks and stores
- their attractiveness to maximize sightseeing. The goal is to maximize
- the objective function subject to the constraints A*x < b and A_eq*x = b_eq.
+ This function sets up the objective to maximize the attractiveness of visiting landmarks,
+ while ensuring that the total time (including travel and visit duration) does not exceed
+ the maximum allowed time. It calculates the pairwise travel times between landmarks and
+ incorporates visit duration to form the inequality constraints.
+
+ The objective is to maximize sightseeing by selecting the most attractive landmarks within
+ the time limit.
Args:
- landmarks (list[Landmark]): List of landmarks.
- max_time (int): Maximum time of visit allowed.
+ prob (pl.LpProblem): The linear programming problem where constraints and the objective will be added.
+ x (pl.LpVariable): A decision variable representing whether a landmark is visited.
+ L (int): The number of landmarks.
+ landmarks (list[Landmark]): List of landmarks to visit.
+ max_time (int): Maximum allowable time for sightseeing, including travel and visit duration.
Returns:
- tuple[list[float], list[float], list[int]]: Objective function coefficients, inequality
+ None: Adds the objective function and constraints to the LP problem directly.
constraint coefficients, and the right-hand side of the inequality constraint.
"""
L = len(landmarks)
@@ -117,14 +122,20 @@ class Optimizer:
def respect_number(self, prob: pl.LpProblem, x: pl.LpVariable, L: int, max_landmarks: int):
"""
- Generate constraints to ensure each landmark is visited only once and cap the total number of visited landmarks.
- -> Adds L-1 rows of constraints
+ Generate constraints to ensure each landmark is visited at most once and cap the total number of visited landmarks.
+
+ This function adds the following constraints to the linear program:
+ 1. Each landmark is visited at most once by creating L-2 constraints (one for each landmark).
+ 2. The total number of visited landmarks is capped by the specified maximum number (`max_landmarks`) plus 2.
Args:
- L (int): Number of landmarks.
+ prob (pl.LpProblem): The linear programming problem where constraints will be added.
+ x (pl.LpVariable): Decision variable indicating whether a landmark is visited.
+ L (int): The total number of landmarks.
+ max_landmarks (int): The maximum number of landmarks that can be visited.
Returns:
- tuple[np.ndarray, list[int]]: Inequality constraint coefficients and the right-hand side of the inequality constraints.
+ None: This function directly modifies the `prob` object by adding constraints.
"""
# L-2 constraints: each landmark is visited exactly once
for i in range(1, L-1):
@@ -137,16 +148,20 @@ class Optimizer:
def break_sym(self, prob: pl.LpProblem, x: pl.LpVariable, L: int):
"""
Generate constraints to prevent simultaneous travel between two landmarks
- in both directions. Constraint to not have d14 and d41 simultaneously.
- Does not prevent cyclic paths with more elements
- -> Adds (L*L-L)/2 rows of constraints (some of which might be zero)
+ in both directions. This constraint ensures that, for any pair of landmarks,
+ travel from landmark i to landmark j (dij) and travel from landmark j to landmark i (dji)
+ cannot happen simultaneously.
+
+ This method adds constraints to break symmetry, specifically to prevent
+ cyclic paths with only two elements. It does not prevent cyclic paths involving more than two elements.
Args:
- L (int): Number of landmarks.
+ prob (pl.LpProblem): The linear programming problem where constraints will be added.
+ x (pl.LpVariable): Decision variable representing travel between landmarks.
+ L (int): The total number of landmarks.
Returns:
- tuple[np.ndarray, list[int]]: Inequality constraint coefficients and
- the right-hand side of the inequality constraints.
+ None: This function modifies the `prob` object by adding constraints in-place.
"""
upper_ind = np.triu_indices(L, 0, L) # Get the upper triangular indices
up_ind_x = upper_ind[0]
@@ -161,15 +176,20 @@ class Optimizer:
def init_eq_not_stay(self, prob: pl.LpProblem, x: pl.LpVariable, L: int):
"""
- Generate constraints to prevent staying in the same position (e.g., removing d11, d22, d33, etc.).
- -> Adds 1 row of constraints
- -> Pre-allocates A_eq for the rest of the computations with (L+ 2 + dynamic incr) rows
+ Generate constraints to prevent staying at the same position during travel.
+ Specifically, it removes travel from a landmark to itself (e.g., d11, d22, d33, etc.).
+
+ This function adds one equality constraint to the optimization problem that ensures
+ no decision variable corresponding to staying at the same landmark is included
+ in the solution. This helps in ensuring that the path does not include self-loops.
Args:
- L (int): Number of landmarks.
+ prob (pl.LpProblem): The linear programming problem where constraints will be added.
+ x (pl.LpVariable): Decision variable representing travel between landmarks.
+ L (int): The total number of landmarks.
Returns:
- tuple[list[np.ndarray], list[int]]: Equality constraint coefficients and the right-hand side of the equality constraints.
+ None: This function modifies the `prob` object by adding an equality constraint in-place.
"""
A_eq = np.zeros((L, L), dtype=np.int8)
@@ -181,18 +201,23 @@ class Optimizer:
prob += (pl.lpSum([A_eq[j] * x[j] for j in range(L*L)]) == 0)
- # Constraint to ensure start at start and finish at goal
def respect_start_finish(self, prob: pl.LpProblem, x: pl.LpVariable, L: int):
"""
- Generate constraints to ensure that the optimization starts at the designated
+ Generate constraints to ensure that the optimization starts at the designated
start landmark and finishes at the goal landmark.
- -> Adds 3 rows of constraints
+
+ Specifically, this function adds three equality constraints:
+ 1. Ensures that the path starts at the designated start landmark (row 0).
+ 2. Ensures that the path finishes at the designated goal landmark (row 1).
+ 3. Prevents any arrivals at the start landmark or departures from the goal landmark (row 2).
Args:
- L (int): Number of landmarks.
+ prob (pl.LpProblem): The linear programming problem where constraints will be added.
+ x (pl.LpVariable): Decision variable representing travel between landmarks.
+ L (int): The total number of landmarks.
Returns:
- tuple[np.ndarray, list[int]]: Inequality constraint coefficients and the right-hand side of the inequality constraints.
+ None: This function modifies the `prob` object by adding three equality constraints in-place.
"""
# Fill-in row 0.
A_eq = np.zeros((3,L*L), dtype=np.int8)
@@ -211,27 +236,24 @@ class Optimizer:
for i in range(3) :
prob += (pl.lpSum([A_eq[i][j] * x[j] for j in range(L*L)]) == b_eq[i])
+
def respect_order(self, prob: pl.LpProblem, x: pl.LpVariable, L: int):
"""
Generate constraints to tie the optimization problem together and prevent
stacked ones, although this does not fully prevent circles.
- -> Adds L-2 rows of constraints
+
+ This function adds constraints to the optimization problem that prevent
+ simultaneous travel between landmarks in a way that would result in stacked ones.
+ However, it does not fully prevent circular paths.
Args:
- L (int): Number of landmarks.
+ prob (pl.LpProblem): The linear programming problem where constraints will be added.
+ x (pl.LpVariable): Decision variable representing travel between landmarks.
+ L (int): The total number of landmarks.
Returns:
- tuple[np.ndarray, list[int]]: Inequality constraint coefficients and the right-hand side of the inequality constraints.
+ None: This function modifies the `prob` object by adding L-2 equality constraints in-place.
"""
- # A_eq = np.zeros(L*L, dtype=np.int8)
- # ones = np.ones(L, dtype=np.int8)
- # # Fill-in rows 4 to L+2
- # for i in range(1, L-1) : # Prevent stacked ones
- # for j in range(L) :
- # A_eq[i + j*L] = -1
- # A_eq[i*L:(i+1)*L] = ones
- # prob += (pl.lpSum([A_eq[j] * x[j] for j in range(L*L)]) == 0)
-
# FIXME: weird 0 artifact in the coefficients popping up
# Loop through rows 1 to L-2 to prevent stacked ones
for i in range(1, L-1):
@@ -243,14 +265,19 @@ class Optimizer:
def respect_user_must(self, prob: pl.LpProblem, x: pl.LpVariable, L: int, landmarks: list[Landmark]) :
"""
Generate constraints to ensure that landmarks marked as 'must_do' are included in the optimization.
- -> Adds a variable number of rows of constraints BUT CAN BE PRE COMPUTED
+ This function adds constraints to the optimization problem to ensure that landmarks marked as
+ 'must_do' are included in the solution. It precomputes the constraints and adds them to the
+ problem accordingly.
Args:
+ prob (pl.LpProblem): The linear programming problem where constraints will be added.
+ x (pl.LpVariable): Decision variable representing travel between landmarks.
+ L (int): The total number of landmarks.
landmarks (list[Landmark]): List of landmarks, where some are marked as 'must_do'.
Returns:
- tuple[np.ndarray, list[int]]: Inequality constraint coefficients and the right-hand side of the inequality constraints.
+ None: This function modifies the `prob` object by adding equality constraints in-place.
"""
ones = np.ones(L, dtype=np.int8)
A_eq = np.zeros(L*L, dtype=np.int8)
@@ -264,52 +291,22 @@ class Optimizer:
prob += (pl.lpSum([A_eq[j] * x[j] for j in range(L*L)]) == 2)
- # Prevent the use of a particular solution. TODO probably can be done faster just using resx
- # def prevent_config(self, prob: pl.LpProblem, x: pl.LpVariable, resx):
- # """
- # Prevent the use of a particular solution by adding constraints to the optimization.
-
- # Args:
- # resx (list[float]): List of edge weights.
-
- # Returns:
- # tuple[list[int], list[int]]: A tuple containing a new row for A and new value for ub.
- # """
-
- # for i, elem in enumerate(resx):
- # resx[i] = round(elem)
-
- # N = len(resx) # Number of edges
- # L = int(np.sqrt(N)) # Number of landmarks
-
- # nonzeroind = np.nonzero(resx)[0] # the return is a little funky so I use the [0]
- # nonzero_tup = np.unravel_index(nonzeroind, (L,L))
-
- # ind_a = nonzero_tup[0].tolist()
- # vertices_visited = ind_a
- # vertices_visited.remove(0)
-
- # ones = np.ones(L, dtype=np.int8)
- # h = np.zeros(L*L, dtype=np.int8)
-
- # for i in range(L) :
- # if i in vertices_visited :
- # h[i*L:i*L+L] = ones
-
- # return h, len(vertices_visited)-1
-
-
- # Prevents the creation of the same circle (both directions)
def prevent_circle(self, prob: pl.LpProblem, x: pl.LpVariable, circle_vertices: list, L: int) :
"""
- Prevent circular paths by by adding constraints to the optimization.
+ Prevent circular paths by adding constraints to the optimization.
+
+ This function ensures that circular paths in both directions (i.e., forward and reverse)
+ between landmarks are avoided in the optimization problem by adding the corresponding constraints.
Args:
- circle_vertices (list): List of vertices forming a circle.
- L (int): Number of landmarks.
+ prob (pl.LpProblem): The linear programming problem instance to which the constraints will be added.
+ x (pl.LpVariable): Decision variable representing the travel between landmarks in the problem.
+ circle_vertices (list): List of indices representing the landmarks that form a circular path.
+ L (int): The total number of landmarks.
Returns:
- tuple[np.ndarray, list[int]]: A tuple containing a new row for constraint matrix and new value for upper bound vector.
+ None: This function modifies the `prob` object by adding two equality constraints that
+ prevent circular paths in both directions for the specified circle vertices.
"""
l = np.zeros((2, L*L), dtype=np.int8)
@@ -544,12 +541,8 @@ class Optimizer:
return prob, x
- def solve_optimization(
- self,
- max_time: int,
- landmarks: list[Landmark],
- max_landmarks: int = None
- ) -> list[Landmark]:
+
+ def solve_optimization(self, max_time: int, landmarks: list[Landmark], max_landmarks: int = None) -> list[Landmark]:
"""
Main optimization pipeline to solve the landmark visiting problem.
@@ -563,14 +556,12 @@ class Optimizer:
Returns:
list[Landmark]: The optimized tour of landmarks with updated travel times, or None if no valid solution is found.
"""
- # 1. Setup the optimization proplem.
+ # Setup the optimization proplem.
L = len(landmarks)
prob, x = self.pre_processing(L, landmarks, max_time, max_landmarks)
- # 2. Solve the problem
+ # Solve the problem and extract results.
prob.solve(pl.PULP_CBC_CMD(msg=False, gapRel=0.1))
-
- # 3. Extract Results
status = pl.LpStatus[prob.status]
solution = [pl.value(var) for var in x] # The values of the decision variables (will be 0 or 1)
@@ -588,10 +579,6 @@ class Optimizer:
timeout = 40
while circles is not None :
i += 1
- # print(f"Iteration {i} of fixing circles")
- # l, b = self.prevent_config(solution)
- # prob += (pl.lpSum([l[j] * x[j] for j in range(L*L)]) == b)
-
if i == timeout :
self.logger.error(f'Timeout: No solution found after {timeout} iterations.')
raise TimeoutError(f"Optimization took too long. No solution found after {timeout} iterations.")
@@ -611,7 +598,6 @@ class Optimizer:
if circles is None :
break
-
# Sort the landmarks in the order of the solution
order = self.get_order(solution)
tour = [landmarks[i] for i in order]
diff --git a/backend/src/utils/refiner.py b/backend/src/utils/refiner.py
index 013d9fc..a07e9c6 100644
--- a/backend/src/utils/refiner.py
+++ b/backend/src/utils/refiner.py
@@ -4,7 +4,6 @@ from math import pi
import yaml
from shapely import buffer, LineString, Point, Polygon, MultiPoint, concave_hull
-
from ..structs.landmark import Landmark
from .get_time_distance import get_time
from .take_most_important import take_most_important