def setUp(self):
goalspec = 'F P_[IC][True,none,==]'
startpoc = (1, 3)
env = init_mdp(startpoc)
keys = ['L', 'IC']
actions = [0, 1, 2, 3]
root = goalspec2BT(goalspec, planner=None)
self.behaviour_tree = BehaviourTree(root)
# # Need to udpate the planner parameters
child = self.behaviour_tree.root
planner = GenRecPropMDP(
env, keys, None, dict(), actions=actions, max_trace=10, seed=123)
child.setup(0, planner, True, 10)
for i in range(10):
self.behaviour_tree.tick(
pre_tick_handler=reset_env(env)
)
print(self.behaviour_tree.root.status)
child.train = False
for i in range(1):
self.behaviour_tree.tick(
pre_tick_handler=reset_env(env)
)
def setUp(self):
goalspec = 'F P_[NC][True,none,==] U F P_[L][03,none,==]'
startpoc = (0, 3)
self.env = init_mdp(startpoc)
keys = ['L', 'NC']
actions = [0, 1, 2, 3]
planner = GenRecPropMDPNear(
self.env, keys, goalspec, dict(),
actions=actions, max_trace=10, seed=123)
root = goalspec2BT(goalspec, planner=planner)
self.behaviour_tree = BehaviourTree(root)
for child in self.behaviour_tree.root.children:
# print(child, child.name)
child.setup(0, planner, True, 20)
# child.planner.env = env
# print(child.goalspec, child.planner.goalspec)
for i in range(20):
self.behaviour_tree.tick(
pre_tick_handler=reset_env(self.env)
)
for child in self.behaviour_tree.root.children:
child.setup(0, planner, True, 10)
child.train = False
# print(child, child.name, child.train)
for i in range(2):
self.behaviour_tree.tick(
pre_tick_handler=reset_env(self.env)
)
def run_planner(planner, behaviour_tree, env, epoch=10):
child = behaviour_tree.root
# Training setup
child.setup(0, planner, True, epoch)
# Training loop
for i in range(epoch):
behaviour_tree.tick(
pre_tick_handler=reset_env(env)
)
print(behaviour_tree.root.status)
# for child in behaviour_tree.root.children:
# child.setup(0, planner, True, 10)
# Inference setup
child.train = False
print(child, child.name, child.train)
# Inference loop
for i in range(1):
behaviour_tree.tick(
pre_tick_handler=reset_env(env)
)
print('inference', behaviour_tree.root.status)
print(env.curr_loc)
def find_cheese(seed, max_trace_len=10, epoch=10):
# Define the environment for the experiment
goalspec = 'F P_[IC][True,none,==]'
# startpoc = (3, 0)
startpoc = (9, 0)
env = init_10x10mdp(startpoc)
keys = ['L', 'IC']
actions = [0, 1, 2, 3]
root = goalspec2BT(goalspec, planner=None)
# print(root)
behaviour_tree = BehaviourTree(root)
# display_bt(behaviour_tree, True)
# print(dir(behaviour_tree))
# # Need to udpate the planner parameters
child = behaviour_tree.root
# for child in behaviour_tree.root.children:
# print(child, child.name, env.curr_loc)
planner = GenRecPropMDP(
env, keys, None, dict(), actions=actions,
max_trace=max_trace_len, epoch=epoch)
child.setup(0, planner, True, epoch=epoch)
# Experiment data
# print(planner.trace_len_data)
data = np.zeros(epoch, dtype=np.uint8)
for i in range(epoch):
behaviour_tree.tick(
pre_tick_handler=reset_env(env)
)
# print(behaviour_tree.root.status)
data[i] = check_bt_status(behaviour_tree.root.status)
# print(planner.trace_len_data)
# for child in behaviour_tree.root.children:
child.setup(0, planner, True, max_trace_len)
child.train = False
# print(child, child.name, child.train)
for i in range(1):
behaviour_tree.tick(
pre_tick_handler=reset_env(env)
)
# data[-1] = check_bt_status(behaviour_tree.root.status)
# print('inference', behaviour_tree.root.status)
# print(planner.trace_len_data)
# print(data)
# print(env.curr_loc)
return (data, planner.trace_len_data)
def run_planner(planner, behaviour_tree, env, epoch=10, maxtracelen=10):
child = behaviour_tree.root
# Training setup
child.setup(0, planner, True, epoch)
child.planner.goalspec = child.name
# Training loop
blackboard = Blackboard()
blackboard.shared_content['current'] = dict()
for i in range(epoch):
reset_env(env)
blackboard.shared_content['current']['epoch'] = i
# print(planner.gtable[(1,3)], planner.gtable[(2,3)])
for j in range(maxtracelen):
behaviour_tree.tick()
# print('Train', i,j, behaviour_tree.root.status)
# if behaviour_tree.root.status == Status.SUCCESS:
# break
if blackboard.shared_content['current']['done']:
break
child.planner.trace = dict()
# print('after trace empty', child.planner.trace)
# print(child.name, child.planner.gtable[(1,3)], child.planner.gtable[(2,3)],child.planner.gtable[(3,3)])
# Inference setup
child.train = False
# Inference loop
for i in range(1):
reset_env(env)
blackboard.shared_content['current']['epoch'] = i
# print(planner.gtable[(1,3)], planner.gtable[(2,3)])
for j in range(maxtracelen):
behaviour_tree.tick()
# print('inference', i,j, behaviour_tree.root.status)
if behaviour_tree.root.status == Status.SUCCESS:
break
if blackboard.shared_content['current']['done']:
break
print(env.curr_loc)
def run_planner_complex(
Planner, behaviour_tree, env, keys, actions, epoch=10, seed=None):
# child = behaviour_tree.root
# Training setup
j = 0
for child in behaviour_tree.root.children:
# print(actions[j], child.name)
planner = Planner(
env, keys, child.name, dict(), actions=actions[0],
max_trace=30, seed=seed)
child.setup(0, planner, True, epoch[j])
j += 1
# Training loop
for i in range(sum(epoch)):
behaviour_tree.tick(
pre_tick_handler=reset_env(env)
)
# print(behaviour_tree.root.status)
for child in behaviour_tree.root.children:
# Inference setup
child.train = False
# print(child, child.name, child.train)
# Inference loop
for i in range(1):
behaviour_tree.tick(
pre_tick_handler=reset_env(env)
)
# print('inference', behaviour_tree.root.status)
if behaviour_tree.root.status is Status.SUCCESS:
return True
else:
return False
def run_planner_complex(Planner,
behaviour_tree,
env,
keys,
actions,
epoch=[10, 10],
seed=None,
maxtracelen=10):
# child = behaviour_tree.root
# Training setup
j = 0
for child in behaviour_tree.root.children:
# print(epoch[j])
planner = Planner(env,
keys,
child.name,
dict(),
actions=actions[j],
max_trace=maxtracelen,
seed=seed,
epoch=epoch[j])
# print(child.name, planner)
child.setup(0, planner, True, epoch[j])
j += 1
# Training loop
blackboard = Blackboard()
blackboard.shared_content['current'] = dict()
blackboard.shared_content['current']['root'] = np.zeros(epoch[0])
for i in range(epoch[0]):
reset_env(env)
blackboard.shared_content['current']['epoch'] = i
# print(planner.gtable[(1,3)], planner.gtable[(2,3)])
for j in range(maxtracelen):
behaviour_tree.tick()
# print('Train', i,j)
if (blackboard.shared_content['current']['done']):
break
blackboard.shared_content['current']['root'][i] = (
behaviour_tree.root.status == Status.SUCCESS) * 1.0
for child in behaviour_tree.root.children:
# self.child.planner.aggrigate_data(len(trace[gkey]), result)
blackboard.shared_content['current'][child.planner.id] = 0
child.planner.tcount = 0
child.planner.trace = dict()
# print('trace after clean up', child.planner.trace)
# print(child.name, child.planner.gtable[(1,3)], child.planner.gtable[(2,3)],child.planner.gtable[(3,3)])
# Inference
for child in behaviour_tree.root.children:
# Inference setup
child.train = False
# print(child, child.name, child.train)
# print(child.name, child.planner.tdata)
# print(blackboard.shared_content['current']['root'])
# Inference loop
# for i in range(1):
# reset_env(env)
# blackboard.shared_content['current']['epoch'] = i
# # print(planner.gtable[(1,3)], planner.gtable[(2,3)])
# for j in range(maxtracelen):
# behaviour_tree.tick()
# # print('inference', i,j, behaviour_tree.root.status)
# if behaviour_tree.root.status == Status.SUCCESS:
# break
# if blackboard.shared_content['current']['done']:
# break
# if behaviour_tree.root.status is Status.SUCCESS:
# return
# else:
# return False
return blackboard.shared_content['current']['root']