def test_inverse_dynamics_controller(self):
urdf_path = FindResourceOrThrow(
"drake/manipulation/models/" +
"iiwa_description/urdf/iiwa14_primitive_collision.urdf")
tree = RigidBodyTree(
urdf_path, floating_base_type=FloatingBaseType.kFixed)
kp = np.array([1., 2., 3., 4., 5., 6., 7.])
ki = np.array([0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7])
kd = np.array([.5, 1., 1.5, 2., 2.5, 3., 3.5])
controller = InverseDynamicsController(robot=tree,
kp=kp,
ki=ki,
kd=kd,
has_reference_acceleration=True)
context = controller.CreateDefaultContext()
output = controller.AllocateOutput()
estimated_state_port = 0
desired_state_port = 1
desired_acceleration_port = 2
control_port = 0
self.assertEqual(
controller.get_input_port(desired_acceleration_port).size(), 7)
self.assertEqual(
controller.get_input_port(estimated_state_port).size(), 14)
self.assertEqual(
controller.get_input_port(desired_state_port).size(), 14)
self.assertEqual(
controller.get_output_port(control_port).size(), 7)
# current state
q = np.array([-0.3, -0.2, -0.1, 0.0, 0.1, 0.2, 0.3])
v = np.array([-0.9, -0.6, -0.3, 0.0, 0.3, 0.6, 0.9])
x = np.concatenate([q, v])
# reference state and acceleration
q_r = q + 0.1*np.ones_like(q)
v_r = v + 0.1*np.ones_like(v)
x_r = np.concatenate([q_r, v_r])
vd_r = np.array([1., 2., 3., 4., 5., 6., 7.])
integral_term = np.array([-1., -2., -3., -4., -5., -6., -7.])
vd_d = vd_r + kp*(q_r-q) + kd*(v_r-v) + ki*integral_term
context.FixInputPort(estimated_state_port, BasicVector(x))
context.FixInputPort(desired_state_port, BasicVector(x_r))
context.FixInputPort(desired_acceleration_port, BasicVector(vd_r))
controller.set_integral_value(context, integral_term)
# compute the expected torque
cache = tree.doKinematics(q, v)
expected_torque = tree.massMatrix(cache).dot(vd_d) + \
tree.dynamicsBiasTerm(cache, {})
controller.CalcOutput(context, output)
self.assertTrue(np.allclose(output.get_vector_data(0).CopyToVector(),
expected_torque))
def test_inverse_dynamics_controller(self):
urdf_path = FindResourceOrThrow(
"drake/manipulation/models/" +
"iiwa_description/urdf/iiwa14_primitive_collision.urdf")
tree = RigidBodyTree(
urdf_path, floating_base_type=FloatingBaseType.kFixed)
kp = np.array([1., 2., 3., 4., 5., 6., 7.])
ki = np.array([0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7])
kd = np.array([.5, 1., 1.5, 2., 2.5, 3., 3.5])
controller = InverseDynamicsController(robot=tree,
kp=kp,
ki=ki,
kd=kd,
has_reference_acceleration=True)
context = controller.CreateDefaultContext()
output = controller.AllocateOutput()
estimated_state_port = 0
desired_state_port = 1
desired_acceleration_port = 2
control_port = 0
self.assertEqual(
controller.get_input_port(desired_acceleration_port).size(), 7)
self.assertEqual(
controller.get_input_port(estimated_state_port).size(), 14)
self.assertEqual(
controller.get_input_port(desired_state_port).size(), 14)
self.assertEqual(
controller.get_output_port(control_port).size(), 7)
# current state
q = np.array([-0.3, -0.2, -0.1, 0.0, 0.1, 0.2, 0.3])
v = np.array([-0.9, -0.6, -0.3, 0.0, 0.3, 0.6, 0.9])
x = np.concatenate([q, v])
# reference state and acceleration
q_r = q + 0.1*np.ones_like(q)
v_r = v + 0.1*np.ones_like(v)
x_r = np.concatenate([q_r, v_r])
vd_r = np.array([1., 2., 3., 4., 5., 6., 7.])
integral_term = np.array([-1., -2., -3., -4., -5., -6., -7.])
vd_d = vd_r + kp*(q_r-q) + kd*(v_r-v) + ki*integral_term
context.FixInputPort(estimated_state_port, BasicVector(x))
context.FixInputPort(desired_state_port, BasicVector(x_r))
context.FixInputPort(desired_acceleration_port, BasicVector(vd_r))
controller.set_integral_value(context, integral_term)
# compute the expected torque
cache = tree.doKinematics(q, v)
expected_torque = tree.massMatrix(cache).dot(vd_d) + \
tree.dynamicsBiasTerm(cache, {})
controller.CalcOutput(context, output)
self.assertTrue(np.allclose(output.get_vector_data(0).CopyToVector(),
expected_torque))
def test_inverse_dynamics_controller(self):
sdf_path = FindResourceOrThrow(
"drake/manipulation/models/"
"iiwa_description/sdf/iiwa14_no_collision.sdf")
plant = MultibodyPlant(time_step=0.01)
Parser(plant).AddModelFromFile(sdf_path)
plant.WeldFrames(plant.world_frame(),
plant.GetFrameByName("iiwa_link_0"))
plant.mutable_gravity_field().set_gravity_vector([0.0, 0.0, 0.0])
plant.Finalize()
# We verify the (known) size of the model.
kNumPositions = 7
kNumVelocities = 7
kNumActuators = 7
kStateSize = kNumPositions + kNumVelocities
self.assertEqual(plant.num_positions(), kNumPositions)
self.assertEqual(plant.num_velocities(), kNumVelocities)
self.assertEqual(plant.num_actuators(), kNumActuators)
kp = np.array([1., 2., 3., 4., 5., 6., 7.])
ki = np.array([0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7])
kd = np.array([.5, 1., 1.5, 2., 2.5, 3., 3.5])
controller = InverseDynamicsController(robot=plant,
kp=kp,
ki=ki,
kd=kd,
has_reference_acceleration=True)
context = controller.CreateDefaultContext()
output = controller.AllocateOutput()
estimated_state_port = controller.get_input_port(0)
desired_state_port = controller.get_input_port(1)
desired_acceleration_port = controller.get_input_port(2)
control_port = controller.get_output_port(0)
self.assertEqual(desired_acceleration_port.size(), kNumVelocities)
self.assertEqual(estimated_state_port.size(), kStateSize)
self.assertEqual(desired_state_port.size(), kStateSize)
self.assertEqual(control_port.size(), kNumVelocities)
# Current state.
q = np.array([-0.3, -0.2, -0.1, 0.0, 0.1, 0.2, 0.3])
v = np.array([-0.9, -0.6, -0.3, 0.0, 0.3, 0.6, 0.9])
x = np.concatenate([q, v])
# Reference state and acceleration.
q_r = q + 0.1*np.ones_like(q)
v_r = v + 0.1*np.ones_like(v)
x_r = np.concatenate([q_r, v_r])
vd_r = np.array([1., 2., 3., 4., 5., 6., 7.])
integral_term = np.array([-1., -2., -3., -4., -5., -6., -7.])
vd_d = vd_r + kp*(q_r-q) + kd*(v_r-v) + ki*integral_term
estimated_state_port.FixValue(context, x)
desired_state_port.FixValue(context, x_r)
desired_acceleration_port.FixValue(context, vd_r)
controller.set_integral_value(context, integral_term)
# Set the plant's context.
plant_context = plant.CreateDefaultContext()
x_plant = plant.GetMutablePositionsAndVelocities(plant_context)
x_plant[:] = x
# Compute the expected value of the generalized forces using
# inverse dynamics.
tau_id = plant.CalcInverseDynamics(
plant_context, vd_d, MultibodyForces(plant))
# Verify the result.
controller.CalcOutput(context, output)
self.assertTrue(np.allclose(output.get_vector_data(0).CopyToVector(),
tau_id))
def test_inverse_dynamics_controller(self):
sdf_path = FindResourceOrThrow(
"drake/manipulation/models/" +
"iiwa_description/sdf/iiwa14_no_collision.sdf")
plant = MultibodyPlant(time_step=0.01)
Parser(plant).AddModelFromFile(sdf_path)
plant.WeldFrames(plant.world_frame(),
plant.GetFrameByName("iiwa_link_0"))
plant.Finalize()
# We verify the (known) size of the model.
kNumPositions = 7
kNumVelocities = 7
kNumActuators = 7
kStateSize = kNumPositions + kNumVelocities
self.assertEqual(plant.num_positions(), kNumPositions)
self.assertEqual(plant.num_velocities(), kNumVelocities)
self.assertEqual(plant.num_actuators(), kNumActuators)
kp = np.array([1., 2., 3., 4., 5., 6., 7.])
ki = np.array([0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7])
kd = np.array([.5, 1., 1.5, 2., 2.5, 3., 3.5])
controller = InverseDynamicsController(robot=plant,
kp=kp,
ki=ki,
kd=kd,
has_reference_acceleration=True)
context = controller.CreateDefaultContext()
output = controller.AllocateOutput()
estimated_state_port = 0
desired_state_port = 1
desired_acceleration_port = 2
control_port = 0
self.assertEqual(
controller.get_input_port(desired_acceleration_port).size(),
kNumVelocities)
self.assertEqual(
controller.get_input_port(estimated_state_port).size(), kStateSize)
self.assertEqual(
controller.get_input_port(desired_state_port).size(), kStateSize)
self.assertEqual(
controller.get_output_port(control_port).size(), kNumVelocities)
# Current state.
q = np.array([-0.3, -0.2, -0.1, 0.0, 0.1, 0.2, 0.3])
v = np.array([-0.9, -0.6, -0.3, 0.0, 0.3, 0.6, 0.9])
x = np.concatenate([q, v])
# Reference state and acceleration.
q_r = q + 0.1*np.ones_like(q)
v_r = v + 0.1*np.ones_like(v)
x_r = np.concatenate([q_r, v_r])
vd_r = np.array([1., 2., 3., 4., 5., 6., 7.])
integral_term = np.array([-1., -2., -3., -4., -5., -6., -7.])
vd_d = vd_r + kp*(q_r-q) + kd*(v_r-v) + ki*integral_term
context.FixInputPort(estimated_state_port, BasicVector(x))
context.FixInputPort(desired_state_port, BasicVector(x_r))
context.FixInputPort(desired_acceleration_port, BasicVector(vd_r))
controller.set_integral_value(context, integral_term)
# Set the plant's context.
plant_context = plant.CreateDefaultContext()
x_plant = plant.GetMutablePositionsAndVelocities(plant_context)
x_plant[:] = x
# Compute the expected value of the generalized forces using
# inverse dynamics.
tau_id = plant.CalcInverseDynamics(
plant_context, vd_d, MultibodyForces(plant))
# Verify the result.
controller.CalcOutput(context, output)
self.assertTrue(np.allclose(output.get_vector_data(0).CopyToVector(),
tau_id))