def __init__(self, BOdict, PolicyDict):
self.PID_Object = PID_Objective(BOdict, PolicyDict)
self.PIDMODE = BOdict['POLICYMODE']
evals = BOdict['ITERATIONS']
zeros = [0, 0, 0]
maximums = [300, 150, 20]
if self.PIDMODE == 'EULER':
self.n_parameters = 9
elif self.PIDMODE == 'HYBRID':
self.n_parameters = 12
elif self.PIDMODE == 'RATE' or self.PIDMODE == 'ALL':
self.n_parameters = 18
else:
print("Invalid PID mode selected")
sys.exit(0)
self.task = OptTask(f=self.PID_Object,
n_parameters=self.n_parameters,
n_objectives=1,
bounds=bounds(
min=zeros * int(self.n_parameters / 3),
max=maximums * int(self.n_parameters / 3)),
task={'minimize'},
vectorized=False)
# labels_param = ['KP_pitch','KI_pitch','KD_pitch', 'KP_roll' 'KI_roll', 'KD_roll', 'KP_yaw', 'KI_yaw', 'KD_yaw', 'KP_pitchRate', 'KI_pitchRate', 'KD_pitchRate', 'KP_rollRate',
# 'KI_rollRate', 'KD_rollRate', "KP_yawRate", "KI_yawRate", "KD_yawRate"])
self.Stop = StopCriteria(maxEvals=evals)
self.sim = BOdict['sim']
def __init__(self, cfg, opt_function):
# self.Objective = SimulationOptimizer(bo_cfg, policy_cfg)
self.b_cfg = cfg.bo
self.p_cfg = cfg.policy
self.cfg = cfg
self.t_c = self.p_cfg.pid.params.terminal_cost
self.l_c = self.p_cfg.pid.params.living_cost
self.norm_cost = 1
self.policy = PidPolicy(cfg)
evals = cfg.bo.iterations
param_min = [0] * len(list(cfg.pid.params.min_values))
param_max = [1] * len(list(cfg.pid.params.max_values))
self.n_parameters = self.policy.numParameters
self.n_pids = self.policy.numpids
params_per_pid = self.n_parameters / self.n_pids
assert params_per_pid % 1 == 0
params_per_pid = int(params_per_pid)
self.task = OptTask(f=opt_function,
n_parameters=self.n_parameters,
n_objectives=1,
bounds=bounds(min=param_min * self.n_pids,
max=param_max * self.n_pids),
task={'minimize'},
vectorized=False)
# labels_param = ['KP_pitch','KI_pitch','KD_pitch', 'KP_roll' 'KI_roll', 'KD_roll', 'KP_yaw', 'KI_yaw', 'KD_yaw', 'KP_pitchRate', 'KI_pitchRate', 'KD_pitchRate', 'KP_rollRate',
# 'KI_rollRate', 'KD_rollRate', "KP_yawRate", "KI_yawRate", "KD_yawRate"])
self.Stop = StopCriteria(maxEvals=evals)
self.sim = cfg.bo.sim
from opto.opto.classes import StopCriteria
from opto.utils import bounds
from opto.opto.acq_func import EI
from opto import regression
from objective_functions import *
import numpy as np
import matplotlib.pyplot as plt
from dotmap import DotMap
init_vrep()
load_scene('scenes/normal.ttt')
obj_f = generate_f(parameter_mode='normal', objective_mode='single', steps=400)
task = OptTask(f=obj_f, n_parameters=4, n_objectives=1, \
bounds=bounds(min=[1, -np.pi, 0, 0], max=[60, np.pi, 1, 1]), \
vectorized=False)
stopCriteria = StopCriteria(maxEvals=50)
p = DotMap()
p.verbosity = 1
p.acq_func = EI(model=None, logs=None)
p.optimizer = opto.CMAES
p.model = regression.GP
opt = opto.BO(parameters=p, task=task, stopCriteria=stopCriteria)
opt.optimize()
logs = opt.get_logs()
print("Parameters: " + str(logs.get_parameters()))
print("Objectives: " + str(logs.get_objectives()))
exit_vrep()