def show_run(self, params):
model_file = '/Users/Siebe/Dropbox/Thesis/Scratches/model.xml'
perturbations = []
render = True
logging = False
evaluate(model_file, self.closed_loop, params, perturbations, render,
logging)
def verify_experiment(ex, params=None):
std_dev_percent = get_std_dev_percent(ex)
print('Verifying ' + str(std_dev_percent) + '% training noise')
if params is None:
params = get_best_params(ex)
test_percentages = [0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20]
# test_percentages = [22, 24, 26, 28, 30]
results = {}
for test_percentage in test_percentages:
print('Testing on ' + str(test_percentage) + '% test noise')
successful = []
fails = []
variation_paths = generate_variations(ex['default_morphology'], test_percentage)
test_id = 0
for path in variation_paths:
printProgressBar(test_id, 100, prefix = 'Progress:', suffix = 'Complete', length = 50)
succes, st, d, ec, ah, sh = sim.evaluate(path, False, params, [], False, False)
if succes and d > 0:
successful.append((d/10, ec/10))
else:
fails.append((d/10, ec/10, st))
test_id += 1
printProgressBar(test_id, 100, prefix = 'Progress:', suffix = 'Complete', length = 50)
print()
results[test_percentage] = {'successful': successful, 'fails': fails}
return (std_dev_percent, results)
def run_simulation(self, render, logging):
cpg_params = self.simulation['cpg_params']
self.model_file = None
if not self.experiment['default_morphology']:
self.model_file = '/Users/Siebe/Dropbox/Thesis/Scratches/model.xml'
elif self.experiment['variation_params']:
model_files, variation_dicts = generate_model_variations(
self.experiment['default_morphology'],
self.experiment['variation_params'],
num=1)
print(variation_dicts)
self.model_file = model_files[0]
elif not self.experiment['delta_dicts']:
self.model_file = generate_temp_model_file(
self.experiment['default_morphology'])
elif not self.model_file:
model_config = dict_elementwise_operator(
self.experiment['default_morphology'],
self.experiment['delta_dicts'][
self.simulation['variation_index']])
self.model_file = generate_temp_model_file(model_config)
perturbations = self.simulation.get('perturbation', [])
return evaluate(self.model_file, self.closed_loop, cpg_params,
perturbations, render, logging)
def eval_wrapper(variables): model_file = variables['model_file'] closed_loop = variables['closed_loop'] params = variables['params'] perturbations = variables['perturbations'] render = variables['render'] logging = variables['logging'] return sim.evaluate(model_file, closed_loop, params, perturbations, render, logging)
def eval_wrapper(variables):
model_files = variables['model_files']
closed_loop = variables['closed_loop']
params = variables['params']
perturbations = variables['perturbations']
render = variables['render']
logging = variables['logging']
# print(model_files)
results = []
for model_file in model_files:
r = sim.evaluate(model_file, closed_loop, params.tolist(),
perturbations, render, logging)
results.append(r)
# print(results)
print("I'm done")
return results
def perturbation_test(self):
perturbation_params = self.experiment.get('perturbation_params', [])
num_tests = 20
if perturbation_params:
test_perturbations = []
perturb_cov = np.diag(perturbation_params['perturb_variances'])
for _ in range(num_tests):
occurences = np.random.geometric(
p=1 / perturbation_params['expected_occurences']
) - 1 # numpy uses shifted geometric
perturbations = []
for i in range(occurences):
perturb_time = np.random.random() * 14
force_torque = np.random.multivariate_normal(
perturbation_params['perturb_means'], perturb_cov)
perturbations.append([perturb_time, list(force_torque)])
test_perturbations.append(perturbations)
# Test in simulation
cpg_params = self.experiment['results']['simulations'][
self.experiment['results']['best_id']]['cpg_params']
model_file = None
if not self.experiment['default_morphology']:
model_file = '/Users/Siebe/Dropbox/Thesis/Scratches/model.xml'
elif not self.experiment['delta_dicts']:
model_file = generate_temp_model_file(
self.experiment['default_morphology'])
elif not self.model_file:
model_config = dict_elementwise_operator(
self.experiment['default_morphology'],
self.experiment['delta_dicts'][
self.simulation['variation_index']])
model_file = generate_temp_model_file(model_config)
rewards = []
for perturbation in test_perturbations:
succes, simulated_time, distance, energy_consumed, action_history, sensor_history = evaluate(
model_file, self.closed_loop, cpg_params, perturbation,
False, False)
reward = 0 if distance < 0 or not succes else (
10 - 0.01 *
(energy_consumed - self.experiment['E0'])**2) * (distance)
rewards.append(reward)
print('Minimum: ' + str(min(rewards)))
print('Mean: ' + str(sum(rewards) / len(rewards)))
print('Maximum: ' + str(max(rewards)))
print('Standard deviation: ' + str(np.std(rewards)))