scheduling test - may be shifted to another repo

test/scheduler.py

+from DecMCTS import Tree
+import numpy.random as rand
+
+# Creat a bunch of jobs
+# Each job has a time to do (in seconds?)
+# Each job has an x-y location (0-1 m)
+# Each job pair has a time to travel between them (fixed velocity m/s)
+n_jobs = 20
+job_time_min = 0
+job_time_max = 10
+vel = 0.1
+rand.seed(1)
+job_locs = [(x,y) for x,y in zip(rand.rand(n_jobs), rand.rand(n_jobs))]
+job_times = [job_time_min+(job_time_max-job_time_min)*t
+             for t in rand.rand(n_jobs)]
+def dist(p1,p2):
+    return ((p1[0]-p2[0])**2 + (p1[1]-p2[1])**2)**0.5
+travel_time = {(j1,j2):dist(job_locs[j1], job_locs[j2])/vel 
+               for j1 in range(n_jobs)
+               for j2 in range(n_jobs)}
+# Data needed for any calculations
+data = {
+    "job_locs": job_locs,
+    "job_times": job_times
+    "tt": travel_time
+}
+
+# State For a particular robot
+class State:
+    def __init__(self, start_pos):
+        self.jobs = []
+        self.time_so_far = 0
+        self.time_wasted = 0
+        self.current_pos = start_pos
+
+    def append(self, new_job):
+        self.jobs.append(new_job)
+        self.time_so_far += data["tt"][self.jobs[-1], self.jobs[-2]] \
+            + job_times[new_job]
+        self.time_wasted += data["tt"][self.jobs[-1], self.jobs[-2]]
+        self.current_pos = job_locs[new_job]
+
+
+def state_storer(data, parent_state, action):
+    if parent_state == None:
+        return State((0,0)) # This state is also used Null action when calculating local reward
+    state = deepcopy(parent_state)
+    state.append(action)
+    return state
+
+# All actions not done by the current robot are considered available
+# actions of other robots are not considered
+def avail_actions(data, states, robot_id):
+    return [j for j in list(range(n_jobs)) if j not in states[robot_id].jobs]
+
+# reward is inversely proportional to total time taken
+def reward(dat, states):
+    time_wasted = [states[robot].time_wasted for robot in states]
+    return 1/sum(each_robot_sum)
+
+# Communicate top n nodes of the MCTS tree
+comm_n = 5
+
+# Create the trees
+n_robots = 2
+trees = [None]*n_robots
+for i in range(n_robots):
+    trees[i] = Tree(data, reward, avail_actions, state_storer, comm_n, i)