diff --git a/examples/flatland_2_0_example.py b/examples/flatland_2_0_example.py
index 2a6d90b02de79ca40c09bdcd0c80888386b4c8f8..3bc31626fd017e0e8ad0b33165c3cc90cdd6b8d0 100644
--- a/examples/flatland_2_0_example.py
+++ b/examples/flatland_2_0_example.py
@@ -37,7 +37,7 @@ env = RailEnv(width=50,
max_tracks_in_city=4,
),
schedule_generator=sparse_schedule_generator(),
- number_of_agents=50,
+ number_of_agents=15,
stochastic_data=stochastic_data, # Malfunction data generator
obs_builder_object=GlobalObsForRailEnv())
diff --git a/flatland/envs/rail_generators.py b/flatland/envs/rail_generators.py
index 08982fc6e9786d980d95e820a22d472149d3b632..0bdf5165ac8ccecbd3b6f373f07fc1bef5895442 100644
--- a/flatland/envs/rail_generators.py
+++ b/flatland/envs/rail_generators.py
@@ -4,6 +4,7 @@ import warnings
from typing import Callable, Tuple, Optional, Dict, List, Any
import msgpack
+import networkx as nx
import numpy as np
from flatland.core.grid.grid4_utils import get_direction, mirror
@@ -544,6 +545,7 @@ def sparse_rail_generator(num_cities=5, grid_mode=False, max_inter_city_rails=4,
:param seed: Random seed to initiate rail
:return: generator
"""
+ G = nx.Graph()
def generator(width, height, num_agents, num_resets=0) -> RailGeneratorProduct:
@@ -552,10 +554,19 @@ def sparse_rail_generator(num_cities=5, grid_mode=False, max_inter_city_rails=4,
rail_array = grid_map.grid
rail_array.fill(0)
np.random.seed(seed + num_resets)
+
+ # Graph to be able to create correct start/end pairs for schedule
+
+
+
node_radius = int(np.ceil((max_tracks_in_city + 2) / 2.0)) + 1
+ if 3 > max_tracks_in_city:
+ rail_in_city = 3
+ else:
+ rail_in_city = 3
max_inter_city_rails_allowed = max_inter_city_rails
- if max_inter_city_rails_allowed > max_tracks_in_city:
- max_inter_city_rails_allowed = max_tracks_in_city
+ if max_inter_city_rails_allowed > rail_in_city:
+ max_inter_city_rails_allowed = rail_in_city
# Generate a set of nodes for the sparse network
# Try to connect cities to nodes first
city_positions = []
@@ -577,7 +588,7 @@ def sparse_rail_generator(num_cities=5, grid_mode=False, max_inter_city_rails=4,
node_connection_time = time.time()
inner_connection_points, outer_connection_points, connection_info, city_orientations = _generate_node_connection_points(
node_positions, node_radius, max_inter_city_rails_allowed,
- max_tracks_in_city)
+ rail_in_city)
print("Connection points", time.time() - node_connection_time)
# Connect the cities through the connection points
@@ -651,7 +662,7 @@ def sparse_rail_generator(num_cities=5, grid_mode=False, max_inter_city_rails=4,
len(node_positions),
tries, nb_nodes))
break
-
+ G.add_node(node_idx)
return node_positions, city_cells
def _generate_node_positions_grid_mode(nb_nodes, node_radius, height, width):
@@ -667,6 +678,7 @@ def sparse_rail_generator(num_cities=5, grid_mode=False, max_inter_city_rails=4,
y_tmp = y_positions[node_idx // nodes_per_row]
node_positions.append((x_tmp, y_tmp))
city_cells.extend(_city_cells(node_positions[-1], node_radius))
+ G.add_node(node_idx)
return node_positions, city_cells
def _generate_node_connection_points(node_positions, node_size, max_inter_city_rails_allowed, tracks_in_city=2):
@@ -691,7 +703,7 @@ def sparse_rail_generator(num_cities=5, grid_mode=False, max_inter_city_rails=4,
city_orientations.append(current_closest_direction)
# set the number of tracks within a city, at least 2 tracks per city
connections_per_direction = np.zeros(4, dtype=int)
- nr_of_connection_points = np.random.randint(2, tracks_in_city + 1)
+ nr_of_connection_points = np.random.randint(3, tracks_in_city + 1)
for idx in connection_sides_idx:
connections_per_direction[idx] = nr_of_connection_points
connection_points_coordinates_inner = [[] for i in range(4)]
@@ -735,6 +747,7 @@ def sparse_rail_generator(num_cities=5, grid_mode=False, max_inter_city_rails=4,
:return:
"""
all_paths = []
+
for current_node in np.arange(len(node_positions)):
direction = 0
connected_to_city = []
@@ -767,6 +780,7 @@ def sparse_rail_generator(num_cities=5, grid_mode=False, max_inter_city_rails=4,
neighb_connection_point = tmp_in_connection_point
new_line = connect_cities(rail_trans, grid_map, tmp_out_connection_point, neighb_connection_point,
city_cells)
+ G.add_edge(current_node, neighb_idx)
all_paths.extend(new_line)
direction += 1
@@ -899,6 +913,7 @@ def sparse_rail_generator(num_cities=5, grid_mode=False, max_inter_city_rails=4,
node_available_start[start_node] -= 1
node_available_target[target_node] -= 1
agent_start_targets_nodes.append((start_node, target_node))
+ print(agent_idx, "has connection", nx.astar_path(G, start_node, target_node))
else:
num_agents -= 1
return agent_start_targets_nodes, num_agents
diff --git a/flatland/envs/schedule_generators.py b/flatland/envs/schedule_generators.py
index eb56cc56c434a6b53aff08b98a780a5195806319..86a41af0ec5ef83ec7d306b25a420d2cac32f31a 100644
--- a/flatland/envs/schedule_generators.py
+++ b/flatland/envs/schedule_generators.py
@@ -106,10 +106,14 @@ def sparse_schedule_generator(speed_ratio_map: Mapping[float, float] = None) ->
# Orient the agent correctly
if current_track_nbr % 2 != 0:
- agents_direction.append(city_orientations[current_start_node])
+ current_orientation = city_orientations[current_start_node]
+ agents_direction.append(current_orientation)
else:
- agents_direction.append((city_orientations[current_start_node] + 2) % 2)
+ current_orientation = (city_orientations[current_start_node] + 2) % 2
+ agents_direction.append(current_orientation)
+ if not rail.check_path_exists(start, current_orientation, target):
+ print("No path")
if speed_ratio_map:
speeds = speed_initialization_helper(num_agents, speed_ratio_map)