mirror of
https://github.com/WallabyLester/RBF-aPID-Controller.git
synced 2026-05-10 04:58:14 +08:00
Andru Liu
58 lines
2.2 KiB
Python
58 lines
2.2 KiB
Python
import unittest
|
|
import numpy as np
|
|
import tensorflow as tf
|
|
from RBF_tf import RBFLayer, RBFAdaptiveModel, train_rbf_adaptive
|
|
|
|
class TestRBFLayer(unittest.TestCase):
|
|
def setUp(self):
|
|
""" Set up a RBFLayer class instance."""
|
|
self.n_centers = 5
|
|
self.input_dim = 3
|
|
self.rbf_layer = RBFLayer(self.n_centers, self.input_dim)
|
|
|
|
def test_initialization(self):
|
|
""" Test the initialization of the layer."""
|
|
self.assertEqual(self.rbf_layer.n_centers, self.n_centers)
|
|
self.assertEqual(self.rbf_layer.centers.shape, (self.n_centers, self.input_dim))
|
|
self.assertEqual(self.rbf_layer.sigmas.shape, (self.n_centers,))
|
|
|
|
def test_call(self):
|
|
""" Test the call method."""
|
|
inputs = tf.random.normal((3, self.input_dim))
|
|
output = self.rbf_layer(inputs)
|
|
|
|
self.assertEqual(output.shape, (3, self.n_centers))
|
|
self.assertTrue(tf.reduce_all(output >= 0))
|
|
|
|
class TestRBFAdaptiveModel(unittest.TestCase):
|
|
def setUp(self):
|
|
""" Set up a RBFAdaptiveModel class instance."""
|
|
self.n_centers = 5
|
|
self.input_dim = 3
|
|
self.model = RBFAdaptiveModel(self.n_centers, self.input_dim)
|
|
|
|
def test_initialization(self):
|
|
""" Test the initialization of the model."""
|
|
self.assertIsInstance(self.model.rbf_layer, RBFLayer)
|
|
self.assertEqual(self.model.output_layer.units, 1)
|
|
|
|
def test_call_output_shape(self):
|
|
""" Test the call method for output shape."""
|
|
inputs = tf.random.normal((3, self.input_dim))
|
|
output = self.model(inputs)
|
|
|
|
self.assertEqual(output.shape, (3, 1))
|
|
|
|
def test_train(self):
|
|
""" Test the train method."""
|
|
initial_weights = self.model.rbf_layer.centers.numpy().copy()
|
|
errors = np.random.normal(size=(100, self.input_dim)) # create 100 samples of input errors
|
|
control_signals = np.random.normal(size=(100, 1)) # create 100 matching control signals
|
|
train_rbf_adaptive(self.model, errors, control_signals, 10)
|
|
|
|
new_weights = self.model.rbf_layer.centers.numpy()
|
|
self.assertFalse(np.array_equal(initial_weights, new_weights))
|
|
|
|
if __name__ == '__main__':
|
|
unittest.main()
|