def TransformAngularVelocityToLocalFrame(self, angular_velocity,
orientation):
"""Transform the angular velocity from world frame to robot's frame.
Args:
angular_velocity: Angular velocity of the robot in world frame.
orientation: Orientation of the robot represented as a quaternion.
Returns:
angular velocity of based on the given orientation.
"""
# Treat angular velocity as a position vector, then transform based on the
# orientation given by dividing (or multiplying with inverse).
# Get inverse quaternion assuming the vector is at 0,0,0 origin.
tr = dp.TinySpatialTransform()
tr.translation = dp.Vector3(1, 2, 3)
tr.translation.set_zero()
orn = dp.Quaternion(orientation[0], orientation[1], orientation[2],
orientation[3])
tr.rotation = dp.Matrix3(orn)
tr_inv = tr.get_inverse()
rel_vel = tr.apply_inverse(
dp.Vector3(angular_velocity[0], angular_velocity[1],
angular_velocity[2]))
#tr.print("tds trans")
#print("tds rel_vel=",rel_vel)
#print(tr)
return vec3_to_np(rel_vel)
def main(argv):
if len(argv) > 1:
raise app.UsageError("Too many command-line arguments.")
w = pydiffphys.TinyWorld()
w.gravity = pydiffphys.Vector3(0, 0, -9.81)
g = w.gravity
print("g=", g.z)
plane = pydiffphys.TinyPlane()
sphere = pydiffphys.TinySphere(5.0)
d = w.get_collision_dispatcher()
mass = 0.0
b1 = pydiffphys.TinyRigidBody(mass, plane)
mass = 1.0
b2 = pydiffphys.TinyRigidBody(mass, sphere)
print("b.linear_velocity=", b1.linear_velocity)
pos_a = pydiffphys.Vector3(0, 0, 0)
orn_a = pydiffphys.Quaternion(0, 0, 0, 1)
pos_b = pydiffphys.Vector3(0, 0, 50)
orn_b = pydiffphys.Quaternion(0, 0, 0, 1)
pose_a = pydiffphys.TinyPose(pos_a, orn_a)
pose_b = pydiffphys.TinyPose(pos_b, orn_b)
b1.world_pose = pose_a
b2.world_pose = pose_b
steps = 1000
for step in range(steps):
b2.apply_gravity(g)
dt = 1. / 240.
b2.apply_force_impulse(dt)
b2.clear_forces()
bodies = [b1, b2]
rb_contacts = w.compute_contacts_rigid_body(bodies, d)
num_iter = 50
solver = pydiffphys.TinyConstraintSolver()
for _ in range(num_iter):
for contact in rb_contacts:
solver.resolve_collision(contact, dt)
for body in bodies:
body.integrate(dt)
print("step=", step, ": b2.world_pose.position=",
b2.world_pose.position, ", b2.linear_velocity=",
b2.linear_velocity)
def create_multi_body(mass, collision_shapes, is_floating, world):
urdf = pd.TinyUrdfStructures()
base_link = pd.TinyUrdfLink()
base_link.link_name = "plane_link"
inertial = pd.TinyUrdfInertial()
inertial.mass = mass
inertial.inertia_xxyyzz = pd.Vector3(mass, mass, mass)
base_link.urdf_inertial = inertial
base_link.urdf_collision_shapes = collision_shapes
urdf.base_links = [base_link]
mb = pd.TinyMultiBody(is_floating)
urdf2mb = pd.UrdfToMultiBody2()
res = urdf2mb.convert2(urdf, world, mb)
return mb
def create_multi_body(link_name, mass, collision_shapes, visual_shapes,
is_floating, mc, world):
urdf = pd.TinyUrdfStructures()
base_link = pd.TinyUrdfLink()
base_link.link_name = link_name
inertial = pd.TinyUrdfInertial()
inertial.mass = mass
inertial.inertia_xxyyzz = pd.Vector3(mass, mass, mass)
base_link.urdf_inertial = inertial
base_link.urdf_collision_shapes = collision_shapes
base_link.urdf_visual_shapes = visual_shapes
urdf.base_links = [base_link]
mb = pd.TinyMultiBody(is_floating)
urdf2mb = pd.UrdfToMultiBody2()
res = urdf2mb.convert2(urdf, world, mb)
b2vis = meshcat_utils_dp.convert_visuals(urdf, None, vis)
return mb, b2vis
def ApplyAction(self, motor_commands, motor_control_mode):
"""Apply the motor commands using the motor model.
Args:
motor_commands: np.array. Can be motor angles, torques, hybrid commands
motor_control_mode: A MotorControlMode enum.
"""
dp.forward_kinematics(self.mb, self.mb.q, self.mb.qd)
dp.forward_dynamics(self.mb, dp.Vector3(0., 0., -10.))
mb_q = vecx_to_np(self.mb.q)[7:]
mb_qd = vecx_to_np(self.mb.qd)[6:]
actual_torque, observed_torque = self._motor_model.convert_to_torque(
motor_commands,
mb_q,
mb_qd,
mb_qd,
motor_control_mode=motor_control_mode)
#print("actual_torque=",actual_torque)
for i in range(len(actual_torque)):
self.mb.tau[i] = actual_torque[i]
#now convert urdf structs to diffphys urdf structs
med0.convert_urdf_structures()
print("conversion to TinyUrdfStructures done!")
texture_path = os.path.join(pd.getDataPath(), "laikago/laikago_tex.jpg")
laikago_vis = meshcat_utils_dp.convert_visuals(med0.urdf_structs, texture_path,
vis)
print("convert_visuals done")
urdf2mb = dp.UrdfToMultiBody2()
is_floating = True
laikago_mb = dp.TinyMultiBody(is_floating)
res = urdf2mb.convert2(med0.urdf_structs, world, laikago_mb)
laikago_mb.set_base_position(dp.Vector3(3., 2., 0.7))
laikago_mb.set_base_orientation(
dp.Quaternion(0.8042817254804792, 0.08692563458095628,
-0.12155529396079404, 0.5751514153863713)
) #0.2474039592545229, 0.0, 0.0, 0.9689124217106448))
mb_solver = dp.TinyMultiBodyConstraintSolver()
q = dp.VectorX(2)
q[0] = -0.53167
q[1] = 0.30707
qd = dp.VectorX(2)
qd[0] = -1
qd[1] = 0
qd_new = laikago_mb.qd
#qd_new[2] = 1
def convert_vec(self, vec):
vec = dp.Vector3(vec[0], vec[1], vec[2])
return vec
urdf_parser = dp.TinyUrdfParser()
plane_urdf_data = urdf_parser.load_urdf("../../data/plane_implicit.urdf")
plane2vis = meshcat_utils_dp.convert_visuals(plane_urdf_data, "../../data/checker_purple.png", vis, "../../data/")
plane_mb = dp.TinyMultiBody(False)
plane2mb = dp.UrdfToMultiBody2()
res = plane2mb.convert2(plane_urdf_data, world, plane_mb)
urdf_data = urdf_parser.load_urdf("../../data/laikago/laikago_toes_zup.urdf")
print("robot_name=",urdf_data.robot_name)
b2vis = meshcat_utils_dp.convert_visuals(urdf_data, "../../data/laikago/laikago_tex.jpg", vis, "../../data/laikago/")
is_floating=True
mb = dp.TinyMultiBody(is_floating)
urdf2mb = dp.UrdfToMultiBody2()
res = urdf2mb.convert2(urdf_data, world, mb)
mb.set_base_position(dp.Vector3(0,0,0.6))
mb.set_base_orientation(dp.Quaternion(0.0, 0.0, 0.706825181105366, 0.7073882691671998))
knee_angle = -0.5
abduction_angle = 0.2
initial_poses = [abduction_angle, 0., knee_angle, abduction_angle, 0., knee_angle,
abduction_angle, 0., knee_angle, abduction_angle, 0., knee_angle]
qcopy = mb.q
print("mb.q=",mb.q)
print("qcopy=",qcopy)
for q_index in range (12):
qcopy[q_index+7]=initial_poses[q_index]
print("qcopy=",qcopy)
inertial.inertia_xxyyzz = pd.Vector3(mass, mass, mass)
base_link.urdf_inertial = inertial
base_link.urdf_collision_shapes = collision_shapes
urdf.base_links = [base_link]
mb = pd.TinyMultiBody(is_floating)
urdf2mb = pd.UrdfToMultiBody2()
res = urdf2mb.convert2(urdf, world, mb)
return mb
world = pd.TinyWorld()
collision_shape = pd.TinyUrdfCollision()
collision_shape.geometry.geom_type = pd.PLANE_TYPE
collision_shape.origin_xyz = pd.Vector3(0., 0., 2.)
plane_mb = create_multi_body(0, [collision_shape], False, world)
collision_shape = pd.TinyUrdfCollision()
collision_shape.geometry.geom_type = pd.SPHERE_TYPE
collision_shape.geometry.sphere.radius = 2.0
sphere_mb = create_multi_body(2.0, [collision_shape], True, world)
sphere_mb.set_base_position(pd.Vector3(0., 0., 25.))
dt = 1. / 240.
mb_solver = pd.TinyMultiBodyConstraintSolver()
while 1:
pd.forward_dynamics(sphere_mb, pd.Vector3(0., 0., -10.))
#vis = meshcat.Visualizer(zmq_url='tcp://127.0.0.1:6000')
#vis.delete()
urdf_parser = dp.TinyUrdfParser()
plane_urdf_data = urdf_parser.load_urdf("../../data/bunny.urdf")
#plane2vis = meshcat_utils_dp.convert_visuals(plane_urdf_data, "../../data/checker_purple.png", vis, "../../data/")
plane_mb = dp.TinyMultiBody(False)
plane2mb = dp.UrdfToMultiBody2()
res = plane2mb.convert2(plane_urdf_data, world, plane_mb)
colliders = plane_urdf_data.base_links[0].urdf_collision_shapes
print("num_colliders=", len(colliders))
ray_from = []
ray_to = []
ray_from.append(dp.Vector3(-2, .1, 0))
ray_to.append(dp.Vector3(2, .1, 0))
caster = dp.TinyRaycast()
colliders = []
if 1:
collision_shape = dp.TinyUrdfCollision()
collision_shape.geometry.geom_type = dp.SPHERE_TYPE
collision_shape.geometry.sphere.radius = 1
colliders.append(collision_shape)
collision_shape = dp.TinyUrdfCollision()
collision_shape.geometry.geom_type = dp.SPHERE_TYPE
collision_shape.geometry.sphere.radius = 1.5
colliders.append(collision_shape)
def joint_angles_from_link_position(self,
robot,
link_position,
link_id,
joint_ids,
position_in_world_frame,
base_translation=(0, 0, 0),
base_rotation=(0, 0, 0, 1)):
"""Uses Inverse Kinematics to calculate joint angles.
Args:
robot: A robot instance.
link_position: The (x, y, z) of the link in the body or the world frame,
depending on whether the argument position_in_world_frame is true.
link_id: The link id as returned from loadURDF.
joint_ids: The positional index of the joints. This can be different from
the joint unique ids.
position_in_world_frame: Whether the input link_position is specified
in the world frame or the robot's base frame.
base_translation: Additional base translation.
base_rotation: Additional base rotation.
Returns:
A list of joint angles.
"""
if not position_in_world_frame:
#tds
base_world_tr = self.mb.get_world_transform(-1)
local_base_tr = dp.TinySpatialTransform()
local_base_tr.translation = dp.Vector3(base_translation[0],
base_translation[1],
base_translation[2])
local_base_tr.rotation = dp.Matrix3(
dp.Quaternion(base_rotation[0], base_rotation[1],
base_rotation[2], base_rotation[3]))
world_tr = base_world_tr * local_base_tr
local_link_tr = dp.TinySpatialTransform()
local_link_tr.rotation = dp.Matrix3(dp.Quaternion(0, 0, 0, 1))
local_link_tr.translation = dp.Vector3(link_position[0],
link_position[1],
link_position[2])
link_tr = world_tr * local_link_tr
world_link_pos = [
link_tr.translation[0], link_tr.translation[1],
link_tr.translation[2]
]
else:
world_link_pos = link_position
ik_solver = 0
target_world_pos = dp.Vector3(world_link_pos[0], world_link_pos[1],
world_link_pos[2])
#target_local_pos = self.mb.get_world_transform(-1).apply_inverse(target_world_pos)
all_joint_angles2 = dp.inverse_kinematics(self.mb, link_id,
target_world_pos)
#print("--")
#print("len(all_joint_angles)=",len(all_joint_angles))
#print("all_joint_angles=",all_joint_angles)
vec = []
for i in range(all_joint_angles2.size() - 7):
vec.append(all_joint_angles2[i + 7])
#print("len(all_joint_angles2)=",len(vec))
#print("all_joint_angles2=",vec)
vec = np.array(vec)
#diff = vec - all_joint_angles
#print("diff=",diff)
# Extract the relevant joint angles.
joint_angles = [vec[i] for i in joint_ids]
return joint_angles
def __init__(self, simulation_time_step):
self.MPC_BODY_MASS = MPC_BODY_MASS
self.MPC_BODY_INERTIA = MPC_BODY_INERTIA
self.MPC_BODY_HEIGHT = MPC_BODY_HEIGHT
self.time_step = simulation_time_step
self.num_legs = NUM_LEGS
self.num_motors = NUM_MOTORS
self.skip_sync = 0
self.world = dp.TinyWorld()
self.world.friction = 10.0
self.mb_solver = dp.TinyMultiBodyConstraintSolver()
self.vis = meshcat.Visualizer(zmq_url='tcp://127.0.0.1:6000')
self.vis.delete()
urdf_parser = dp.TinyUrdfParser()
plane_urdf_data = urdf_parser.load_urdf(
"../../../data/plane_implicit.urdf")
plane2vis = meshcat_utils_dp.convert_visuals(
plane_urdf_data, "../../../data/checker_purple.png", self.vis,
"../../../data/")
self.plane_mb = dp.TinyMultiBody(False)
plane2mb = dp.UrdfToMultiBody2()
res = plane2mb.convert2(plane_urdf_data, self.world, self.plane_mb)
self.urdf_data = urdf_parser.load_urdf(
"../../../data/laikago/laikago_toes_zup_joint_order.urdf")
print("robot_name=", self.urdf_data.robot_name)
self.b2vis = meshcat_utils_dp.convert_visuals(
self.urdf_data, "../../../data/laikago/laikago_tex.jpg", self.vis,
"../../../data/laikago/")
is_floating = True
self.mb = dp.TinyMultiBody(is_floating)
urdf2mb = dp.UrdfToMultiBody2()
self.res = urdf2mb.convert2(self.urdf_data, self.world, self.mb)
self.mb.set_base_position(dp.Vector3(0, 0, 0.6))
knee_angle = -0.5
abduction_angle = 0.2
self.initial_poses = [
abduction_angle, 0., knee_angle, abduction_angle, 0., knee_angle,
abduction_angle, 0., knee_angle, abduction_angle, 0., knee_angle
]
qcopy = self.mb.q
print("mb.q=", self.mb.q)
print("qcopy=", qcopy)
for q_index in range(12):
qcopy[q_index + 7] = self.initial_poses[q_index]
print("qcopy=", qcopy)
self.mb.set_q(qcopy)
print("2 mb.q=", self.mb.q)
print("self.mb.links=", self.mb.links)
print("TDS:")
for i in range(len(self.mb.links)):
print("i=", i)
l = self.mb.links[i]
#print("l.link_name=", l.link_name)
print("l.joint_name=", l.joint_name)
dp.forward_kinematics(self.mb, self.mb.q, self.mb.qd)
self._BuildJointNameToIdDict()
self._BuildUrdfIds()
self._BuildMotorIdList()
self.ResetPose()
self._motor_enabled_list = [True] * self.num_motors
self._step_counter = 0
self._state_action_counter = 0
self._motor_offset = np.array([0.] * 12)
self._motor_direction = np.array([1.] * 12)
self.ReceiveObservation()
self._kp = self.GetMotorPositionGains()
self._kd = self.GetMotorVelocityGains()
self._motor_model = LaikagoMotorModel(
kp=self._kp, kd=self._kd, motor_control_mode=MOTOR_CONTROL_HYBRID)
self.ReceiveObservation()
self.mb.set_base_position(
dp.Vector3(START_POS[0], START_POS[1], START_POS[2]))
self.mb.set_base_orientation(
dp.Quaternion(START_ORN[0], START_ORN[1], START_ORN[2],
START_ORN[3]))
dp.forward_kinematics(self.mb, self.mb.q, self.mb.qd)
meshcat_utils_dp.sync_visual_transforms(self.mb, self.b2vis, self.vis)
self._SettleDownForReset(reset_time=1.0)