def test_custom_reward(self):
from gym.envs.registration import register
register(id='simglucose-adolescent3-v0',
entry_point='simglucose.envs:T1DSimEnv',
kwargs={
'patient_name': 'adolescent#003',
'reward_fun': custom_reward
})
env = gym.make('simglucose-adolescent3-v0')
ctrller = BBController()
reward = 1
done = False
info = {'sample_time': 3, 'patient_name': 'adolescent#002', 'meal': 0}
observation = env.reset()
for t in range(200):
env.render(mode='human')
print(observation)
# action = env.action_space.sample()
ctrl_action = ctrller.policy(observation, reward, done, **info)
action = ctrl_action.basal + ctrl_action.bolus
observation, reward, done, info = env.step(action)
print("Reward = {}".format(reward))
if observation.CGM > 180:
self.assertEqual(reward, -1)
elif observation.CGM < 70:
self.assertEqual(reward, -2)
else:
self.assertEqual(reward, 1)
if done:
print("Episode finished after {} timesteps".format(t + 1))
break
def test_gym_random_agent(self):
from gym.envs.registration import register
register(id='simglucose-adolescent2-v0',
entry_point='simglucose.envs:T1DSimEnv',
kwargs={'patient_name': 'adolescent#002'})
env = gym.make('simglucose-adolescent2-v0')
ctrller = BBController()
reward = 0
done = False
info = {'sample_time': 5, 'patient_name': 'adolescent#002', 'meal': 0}
observation = env.reset()
for t in range(150):
env.render(mode='human')
print(observation)
# action = env.action_space.sample()
ctrl_action = ctrller.policy(observation, reward, done, **info)
action = ctrl_action.basal + ctrl_action.bolus
observation, reward, done, info = env.step(action)
if done:
print("Episode finished after {} timesteps".format(t + 1))
break
def run(es=None,policy=None,patient_id=1,Initial_Bg=0):
# es=self.es[0]
print(es)
env = gym.make('simglucose-adult{}-CHO{}-v0'.format(Initial_Bg,patient_id+1))
ctrller = BBController()
# ctrller = PIDController(P=0.001, I=0.00001, D=0.001)
# ctrller = PIDController(P=PID_parameters[patient_id][0], I=PID_parameters[patient_id][1], D=PID_parameters[patient_id][2])
reward = 0
done = False
info = {'sample_time': 5,
'patient_name': 'adult#001',
'meal': 0}
observation = env.reset()
pre_glucose = 0
pre_rate = 0
##########################mitigation##################
Monitor = 2#2 #0:CAWT; 1:MPC 2:DT
Mitigation_Enable = True
threshold_col = threshold_col_list[patient_id]+'_CV0'
if Mitigation_Enable:
#====read threshold files############
thresholds=pd.read_csv("/home/uva-dsa/Research/Medical/test/openaps_monitor/Reseult/scripts-12rules_uvasimulator/thresholds.csv")
threshold=thresholds[threshold_col]
mitigate_H1_flag = False
mitigate_H2_flag = False
if Monitor == 2:
import numpy as np
from sklearn.externals import joblib
model_name = '/home/uva-dsa/Research/Medical/test/openaps_monitor/Reseult/script_jupyternotebook/saved_model_Uva/DT/DT3_NoFIlabel_model_{}.sav'.format(threshold_col)
print("Load model {} now!".format(model_name))
clf = joblib.load(model_name)
delBg = 0
delIob = 0
delInsulinRate = 0
pre_insulinRate = 0
pre_iob =0
pre_bg =0
######################################################
for t1 in range(150):
t = t1 -30 #wait 30*5=150 minutes
# env.render(mode='human')
glucose = observation.CGM
# print(observation)
glucose_refresh = True
rate_refresh = True # update the glucose reading and rate output command
#Fault injection Hook################
#glucose:HOOK#
#hold the glucose value when fresh signal is false
if glucose_refresh != True:
glucose = pre_glucose
#update observation
if observation.CGM != glucose:
observation=Observation(CGM=glucose)
print(observation)
#get the action beased on policy and observation
# (1) random action
# action = env.action_space.sample()
# (2) PID or BB control action
ctrl_action = ctrller.policy(observation, reward, done, **info)
action = ctrl_action.basal + ctrl_action.bolus
# # (3) RL or DDPG action
# action,_ = policy.get_action(observation)# action = es.get_action(t, observation, policy) #algo.policy.get_action(observation)
# print(action)
# print(es.get_action(t, observation, policy))
#Fault injection Hook################
#rate:HOOK#
#hold the action value when fresh signal is false
if rate_refresh != True:
action = pre_rate
#take the action
observation, reward, done, info = env.step(action)
#update previous glucose and rate value
pre_glucose = glucose
############################################3#
#==========mitigation code####++==============
if Mitigation_Enable:
bg=glucose
insulinRate = action
bgTarget = 140
iob =cal_single_iob(insulinRate)
if t1 >1:
delBg = bg-pre_bg #data["CGM_glucose"][i] - data["CGM_glucose"][i-1]
delIob = iob - pre_iob #data["IOB"][i] - data["IOB"][i-1]
delInsulinRate = insulinRate-pre_insulinRate #data["rate"][i] - data["rate"][i-1]
if t1 >5:
sub_alert_msg = ""
if Monitor==0: #CAWT
sub_alert_flag = False
if bg > 180 and delBg > 5: #insufficient insulin
if iob <threshold[6] and insulinRate<0.1 :
sub_alert_flag = True
sub_alert_msg = "row_7" # rule14->7
if bg > bgTarget+10: #+10? >150
#if delBg >= -3:
# if bg>180 and iob < -0.120728641206 and insulinRate == 0: # row_37
# if bg>180 and
if delBg > 0.3:
# if bg>190:
if delBg>2.5 and delIob <= 0 and iob <threshold[4]:#-0.3: #row_10 IOB is falling
if delInsulinRate == 0 and insulinRate<0.1: #keep insulin
sub_alert_flag = True
sub_alert_msg = "row_5" # rule10 ->5
elif delIob < 0 and iob < 0:#threshold[0]:#0.145605040799: # row_1
if delInsulinRate < 0: #dec_insulin
sub_alert_flag = True
sub_alert_msg = "row_1"
elif delIob == 0 and iob < threshold[1]: # row_2
if delInsulinRate < 0: #dec_insulin
sub_alert_flag = True
sub_alert_msg = "row_2"
else: #BG<HBGT
if bg < threshold[8]:#bgLowerTh:
if insulinRate != 0 :#and iob>-0.5:#zero insulin
sub_alert_flag = True
sub_alert_msg = "row_9"
elif bg < bgTarget+10: #110
if delBg < -0.3:
if delIob >=0 and iob >threshold[5]:#0.3: # IOB is not falling
if delInsulinRate == 0 and insulinRate>0.05 :
sub_alert_flag = True
sub_alert_msg = "row_6" # rule12->6
# checking if BG is falling more than the threshold
#if delBg < thBgFall:
# if bg<80:
if delIob > 0 and iob > threshold[2]:#-0.199631233636: # row_7->3
if delInsulinRate > 0.05:
sub_alert_flag = True
sub_alert_msg = "row_3"
elif delIob == 0 and iob > threshold[3]: # row_8->4
if delInsulinRate > 0.05:
sub_alert_flag = True
sub_alert_msg = "row_4"
if sub_alert_flag:
if sub_alert_msg in ["row_3","row_4","row_6","row_9","row_13"]:#H1hazard
mitigate_H1_flag = True
elif sub_alert_msg in ["row_1","row_2","row_5","row_7"]:#H2hazard
mitigate_H2_flag = True
#==========mitigation code####++==============
elif Monitor ==1: #MPC
if glucose <70 :#H1hazard
# insulinRate = 0
# loaded_suggested_data["fault"] = "yes"
# loaded_suggested_data["fault_reason"] = sub_alert_msg+"_Mitigation"
# print("\n***************************************")
# print("********** Unsafe Action !!!!! *************")
sub_alert_msg = 'H1'
mitigate_H1_flag = True
elif glucose>180:#H2hazard
# insulinRate = 2.1
# loaded_suggested_data["fault"] = "yes"
# loaded_suggested_data["fault_reason"] = sub_alert_msg+"_Mitigation"
# print("\n***************************************")
# print("********** Unsafe Action !!!!! *************")
sub_alert_msg = 'H2'
mitigate_H2_flag = True
elif Monitor == 2: #DT
predict_proba=clf.predict_proba(np.array([glucose,iob,insulinRate]).reshape(1,-1))
if int(np.argmax(predict_proba,axis=1)):
if glucose < bgTarget:
sub_alert_msg = 'H1'
mitigate_H1_flag = True
else:#H2hazard
sub_alert_msg = 'H2'
mitigate_H2_flag = True
#########start to mitigate#########
if mitigate_H1_flag == True:
if insulinRate < pre_insulinRate or glucose>bgTarget+10:#if fault is removed stop mitigation
mitigate_H1_flag = False #reset hazard flag
insulinRate = 0
print("\n***************************************")
print(sub_alert_msg)
print("********** Unsafe Action !!!!! *************")
elif mitigate_H2_flag == True:
if insulinRate > pre_insulinRate or glucose<bgTarget+40:#if fault is removed stop mitigation
mitigate_H2_flag = False
insulinRate = 0.1 #0.1
print("\n***************************************")
print(sub_alert_msg)
print("********** Unsafe Action !!!!! *************")
pre_insulinRate = insulinRate
pre_iob = iob
pre_bg =bg
#######################33End of mitigation######################################
pre_rate = action
# if done:
# print("Episode finished after {} timesteps".format(t + 1))
# break
# print(env.show_history())
save_results('./simulation_data/',env.show_history(),Patient_list[patient_id])
env.close()
# if __name__ == "__main__":
# Run_simulation(argv[1],argv[2])