Files
pyrobolearn/examples/manipulability/com_manipulability_tracking_with_balance.py

396 lines
17 KiB
Python

#!/usr/bin/env python
# -*- coding: utf-8 -*-
"""Center of mass manipulability tracking
Track the velocity manipulability ellipsoid of the center of mass of a particular robot while keeping its balance.
See Also:
- `com_manipulability_tracking.py`: in this example, the base is fixed.
- `com_dynamic_manipulability_tracking_with_balance.py`: in this example, the dynamic manipulability ellipsoid is
tracked instead of the velocity one.
References:
[1] "Robotics: Modelling, Planning and Control" (section 3.9), Siciliano et al., 2010
[2] "Geometry-aware Tracking of Manipulability Ellipsoids", Jaquier et al., R:SS, 2018
"""
import time
# from itertools import count
import numpy as np
from scipy.linalg import block_diag
import matplotlib.pyplot as plt
import argparse
from pyrobolearn.simulators import Bullet
from pyrobolearn.worlds import BasicWorld
from pyrobolearn.robots import Nao, Centauro
# create parser
parser = argparse.ArgumentParser()
parser.add_argument('-r', '--robot', help='the robot to track the velocity manipulability ellipsoid', type=str,
choices=['nao', 'centauro'], default='nao')
parser.add_argument('-n', '--init_manipulability', help='which desired velocity manipulability we want to track',
type=int, default=4)
parser.add_argument('-i', '--init_config', help='the initial configuration for the robot', type=int,
choices=range(1, 5), default=1)
args = parser.parse_args()
# program variable
robot_name = args.robot
dt = 0.01 # Sampling time
num_des_manip = args.init_manipulability # Number of desired manipulability (Useful for tests)
init_qs = args.init_config # Number of initial configuration for the robots (Useful for tests)
# Create simulator and world
sim = Bullet()
world = BasicWorld(sim)
# load robot
if robot_name == 'nao':
robot = Nao(sim, fixed_base=False)
elif robot_name == 'centauro':
robot = Centauro(sim, fixed_base=False)
else:
raise NotImplementedError("The given robot has not been implemented")
# Loop for setting stable initial conditions
for i in range(50):
world.step(sleep_dt=0.1)
# Define velocity manipulability for CoM, proportional gain, and initial configurations
if robot.name == 'nao':
Km = 100 * np.eye(6) # Proportional gain for Nao for manip. tracking
# desired Velocity Manipulability for CoM (Nao)
des_manips = np.array([[[1.539e-03, 6.653e-04, 0.833e-04],
[6.653e-04, 2.080e-03, -5.843e-05],
[0.833e-04, -5.843e-05, 8.601e-04]],
[[2.580e-03, 1.653e-03, 4.833e-04],
[1.653e-03, 1.539e-03, -5.843e-05],
[4.833e-04, -5.843e-05, 9.601e-04]],
[[1.580e-03, .653e-03, 4.833e-04],
[.653e-03, 1.539e-03, -5.843e-05],
[4.833e-04, -5.843e-05, 9.601e-04]],
[[1e-04, 0.0, 0.0],
[0.0, 5e-04, 0.0],
[0.0, 0.0, 1e-04]]])
des_vel_manip = des_manips[num_des_manip - 1, :, :]
# Get ids for feet (used for kinematics function)
left_foot_id = robot.get_link_ids('l_ankle')
right_foot_id = robot.get_link_ids('r_ankle')
# Gain matrices
Kcom = np.diag((50, 40, 0)) # Proportional gain for CoM position control
Klf = np.diag((20, 20, 20)) # Proportional gain for foot position control
Krf = np.diag((20, 20, 20)) # Proportional gain for foot position control
# Setting initial configuration of the robot
if init_qs == 1:
q0 = [1.55, 0.135, -1.05, -0.36, 1.55, -0.135, 1.05, 0.36]
q0id = ['LShoulderPitch', 'LShoulderRoll', 'LElbowYaw', 'LElbowRoll',
'RShoulderPitch', 'RShoulderRoll', 'RElbowYaw', 'RElbowRoll']
elif init_qs == 2:
q0 = [1.55, 0.135, -1.05, -0.36, 1.55, -0.135, 1.05, 0.36, 0.1, -0.1, 0.12, 0.12, -0.12, -0.12]
q0id = ['LShoulderPitch', 'LShoulderRoll', 'LElbowYaw', 'LElbowRoll', 'RShoulderPitch', 'RShoulderRoll',
'RElbowYaw', 'RElbowRoll', 'LHipRoll', 'RHipRoll', 'LKneePitch', 'RKneePitch', 'LAnklePitch',
'RAnklePitch']
elif init_qs == 3:
q0 = [2.0, 0.4, -1.3, -1.2, 2.0, -0.4, 1.3, 1.2, 0.05, -0.05, 0.2, 0.2, -0.2, -0.2]
q0id = ['LShoulderPitch', 'LShoulderRoll', 'LElbowYaw', 'LElbowRoll', 'RShoulderPitch', 'RShoulderRoll',
'RElbowYaw', 'RElbowRoll', 'LHipRoll', 'RHipRoll', 'LKneePitch', 'RKneePitch', 'LAnklePitch',
'RAnklePitch']
else:
q0 = 0
q0id = 0
elif robot.name == 'centauro':
Km = 50 * np.eye(6) # Proportional gain for Centauro
# desired Velocity Manipulability for CoM (Centauro)
des_manips = np.array([[[0.0207, 0.008, 0.0],
[0.008, 0.01, -0.005],
[0.0, -0.005, 0.006]],
[[5.173e-03, -2.733e-03, 1.920e-03],
[-2.733e-03, 2.038e-02, 7.185e-04],
[1.920e-03, 7.185e-04, 2.107e-03]],
[[2.2e-02, 0.0, -0.01],
[0.0, 2.1e-02, 0.0],
[-0.01, 0.0, 5.e-03]],
[[5e-04, 0.0, 0.0],
[0.0, .1, 0.0],
[0.0, 0.0, 5e-04]]])
des_vel_manip = des_manips[num_des_manip - 1, :, :]
print("desired velocity manipulability: {}".format(des_vel_manip))
# Get ids for "feet" (used for kinematics function)
left_foot1_id = robot.get_link_ids('wheel_1')
right_foot1_id = robot.get_link_ids('wheel_2')
left_foot2_id = robot.get_link_ids('wheel_3')
right_foot2_id = robot.get_link_ids('wheel_4')
# Gain matrices
Kcom = np.diag((250.0, 250.0, 0.0)) # Proportional gain for CoM position control
Kl1f = np.diag((180, 180, 180)) # Proportional gain for foot position control
Kr1f = np.diag((180, 180, 180)) # Proportional gain for foot position control
Kl2f = np.diag((180, 180, 180)) # Proportional gain for foot position control
Kr2f = np.diag((180, 180, 180)) # Proportional gain for foot position control
# Setting initial configuration of the robot
if init_qs == 1:
q0 = [-.7, -.65, .61, -.5, .71, .64, -.68, .7]
q0id = ['j_arm1_1', 'j_arm1_2', 'j_arm1_3', 'j_arm1_4', 'j_arm2_1', 'j_arm2_2', 'j_arm2_3', 'j_arm2_4']
elif init_qs == 2:
q0 = [-.7, -.65, .61, -.5, .71, .64, -.68, .7, -0.42, -0.96, -0.59,
0.42, 0.96, 0.59, 0.42, 0.96, 0.59, -0.42, -0.96, -0.59]
q0id = ['j_arm1_1', 'j_arm1_2', 'j_arm1_3', 'j_arm1_4', 'j_arm2_1', 'j_arm2_2', 'j_arm2_3', 'j_arm2_4',
'hip_pitch_1', 'knee_pitch_1', 'ankle_pitch_1', 'hip_pitch_2', 'knee_pitch_2', 'ankle_pitch_2',
'hip_pitch_3', 'knee_pitch_3', 'ankle_pitch_3', 'hip_pitch_4', 'knee_pitch_4', 'ankle_pitch_4']
elif init_qs == 3:
q0 = [-.3, -1.3, .61, -.5, .3, 1.3, -.68, .7, -0.42, -0.96, -0.59,
0.42, 0.96, 0.59, 0.42, 0.96, 0.59, -0.42, -0.96, -0.59]
q0id = ['j_arm1_1', 'j_arm1_2', 'j_arm1_3', 'j_arm1_4', 'j_arm2_1', 'j_arm2_2', 'j_arm2_3', 'j_arm2_4',
'hip_pitch_1', 'knee_pitch_1', 'ankle_pitch_1', 'hip_pitch_2', 'knee_pitch_2', 'ankle_pitch_2',
'hip_pitch_3', 'knee_pitch_3', 'ankle_pitch_3', 'hip_pitch_4', 'knee_pitch_4', 'ankle_pitch_4']
else:
q0 = 0
q0id = 0
else:
left_foot_id, right_foot_id = 0, 0
# Loop need to set the robot initial posture
if not (isinstance(q0, int) and q0 == 0 and isinstance(q0id, int) and q0id == 0):
for n in range(15):
robot.set_joint_positions(np.asarray(q0), robot.get_joint_ids(np.asarray(q0id)))
world.step()
# Augmented gain matrix for balancing controller
if robot.name == 'centauro':
Kbal = block_diag(Kl1f, Kr1f, Kl2f, Kr2f, Kcom)
else:
Kbal = block_diag(Klf, Krf, Kcom)
print("Kbal: {}".format(Kbal))
# Initial conditions
time.sleep(2.0)
num_dofs = robot.num_dofs - 6
CoMr = robot.get_center_of_mass_position() # Desired CoM
print("CoMr: {}".format(CoMr))
if robot.name == 'centauro':
xref_l1f = robot.get_link_world_frame_positions(left_foot1_id) # Desired position for left foot
xref_r1f = robot.get_link_world_frame_positions(right_foot1_id) # Desired position for right foot
xref_l2f = robot.get_link_world_frame_positions(left_foot2_id) # Desired position for left foot
xref_r2f = robot.get_link_world_frame_positions(right_foot2_id) # Desired position for right foot
else:
xref_lf = robot.get_link_world_frame_positions(left_foot_id) # Desired position for left foot
# Qref_lf = robot.get_link_world_frame_orientations(leftFootId)
xref_rf = robot.get_link_world_frame_positions(right_foot_id) # Desired position for right foot
# Qref_rf = robot.get_link_world_frame_orientations(rightFootId)
# Display initial and desired manipulability ellipsoid
robot = world.load_robot(robot)
world.step()
q0 = robot.get_joint_positions()
print("q0: {}".format(q0))
Jcom0 = robot.get_center_of_mass_jacobian(q0)
if not robot.has_fixed_base():
vel_manip = robot.compute_velocity_manipulability_ellipsoid(Jcom0[:, 6:])
else:
vel_manip = robot.compute_velocity_manipulability_ellipsoid(Jcom0)
print("Mv0: {}".format(vel_manip[0:3, 0:3]))
base_pos = robot.get_base_position()
robot.draw_velocity_manipulability_ellipsoid(link_id=-1, JJT=10 * des_vel_manip, color=(0.1, 0.75, 0.1, 0.6))
ellipsoid_id = robot.draw_velocity_manipulability_ellipsoid(link_id=-1, JJT=10 * vel_manip[0:3, 0:3],
color=(0.75, 0.1, 0.1, 0.6))
# Logging variables
# Format: [q minEigvalue(Jbal) minEigvalue(Jman) balanceError CurrentManip(1x9) SPDdistance]
log_array = np.zeros((400, num_dofs + 2 + Kbal.shape[0] + vel_manip[0:3, 0:3].size + 1))
# Run simulator
# for i in count():
for i in range(400):
# Update current robot state
qt = robot.get_joint_positions()
CoMt = robot.get_center_of_mass_position() # Current CoM
robot.draw_com_position(0.03)
if robot.name == 'centauro':
xt_l1f = robot.get_link_world_frame_positions(left_foot1_id) # Current position for left foot
xt_r1f = robot.get_link_world_frame_positions(right_foot1_id) # Current position for right foot
xt_l2f = robot.get_link_world_frame_positions(left_foot2_id) # Current position for left foot
xt_r2f = robot.get_link_world_frame_positions(right_foot2_id) # Current position for right foot
else:
xt_lf = robot.get_link_world_frame_positions(left_foot_id) # Current left foot pos
# Qt_lf = robot.get_link_world_frame_orientations(leftFootId)
xt_rf = robot.get_link_world_frame_positions(right_foot_id) # Current right foot pos
# Simple balance control with IK kinematics for CoM and feet
# Get Jacobians: Jcom, Jlf, and Jrf
Jcom = robot.get_center_of_mass_jacobian(qt)
if robot.name == 'centauro':
Jl1f = robot.get_jacobian(left_foot1_id, qt)
Jr1f = robot.get_jacobian(right_foot1_id, qt)
Jl2f = robot.get_jacobian(left_foot2_id, qt)
Jr2f = robot.get_jacobian(right_foot2_id, qt)
else:
Jlf = robot.get_jacobian(left_foot_id, qt)
Jrf = robot.get_jacobian(right_foot_id, qt)
# Compose Jacobian and nullspace for balancing task
if robot.name == 'centauro':
Jbal = np.vstack((Jl1f[0:3, ], Jr1f[0:3, ], Jl2f[0:3, ], Jr2f[0:3, ], Jcom[0:3, ]))
else:
Jbal = np.vstack((Jlf[0:3, ], Jrf[0:3, ], Jcom[0:3, ]))
Ubal, Sbal, VhBal = np.linalg.svd(Jbal)
if np.min(Sbal) < 4.5E-2:
pJbal = robot.get_damped_least_squares_inverse(Jbal, 4.5E-2)
else:
pJbal = robot.get_damped_least_squares_inverse(Jbal, 1E-8)
Nbal = np.eye(Jbal.shape[1]) - np.dot(pJbal, Jbal)
# Compute balancing task errors
dx_com = CoMr - CoMt # CoM error
if robot.name == 'centauro':
dx_l1f = xref_l1f - xt_l1f # Left foot position error
dx_r1f = xref_r1f - xt_r1f # Right foot position error
dx_l2f = xref_l2f - xt_l2f # Left foot position error
dx_r2f = xref_r2f - xt_r2f # Right foot position error
dx_bal = np.vstack((dx_l1f.reshape(3, 1), dx_r1f.reshape(3, 1),
dx_l2f.reshape(3, 1), dx_r2f.reshape(3, 1), dx_com.reshape(3, 1))) # Augmented error vector
else:
dx_lf = xref_lf - xt_lf # Left foot position error
dx_rf = xref_rf - xt_rf # Right foot position error
dx_bal = np.vstack((dx_lf.reshape(3, 1), dx_rf.reshape(3, 1), dx_com.reshape(3, 1))) # Augmented error vector
# dx_bal = np.vstack((np.zeros((6, 1)), dx_com.reshape(3, 1))) # Augmented error vector
# Proportional controller for position
dxref_bal = np.dot(Kbal, dx_bal)
# Compute desired joint velocities for balancing
dq_bal = np.dot(pJbal, dxref_bal)
dq_bal = dq_bal.reshape((Jbal.shape[1],))
# Tracking of CoM velocity manipulability in nullspace
if not robot.has_fixed_base():
vel_manip = robot.compute_velocity_manipulability_ellipsoid(Jcom[:, 6:])
else:
vel_manip = robot.compute_velocity_manipulability_ellipsoid(Jcom)
# Plot current manipulability ellipsoid
if i % 40 == 0:
ellipsoid_id = robot.update_manipulability_ellipsoid(link_id=-1, ellipsoid_id=ellipsoid_id,
ellipsoid=10 * vel_manip[:3, :3],
color=(0.75, 0.1, 0.1, 0.6))
# Obtaining joint velocity command
if not robot.has_fixed_base():
dq_man, minSman, SPDdist = robot.calculate_inverse_differential_kinematics_velocity_manipulability(Jcom[:, 6:],
des_vel_manip,
Km)
# dq_man = np.vstack((np.zeros((6, 1)), dq_man.reshape(num_dofs, 1)))
else:
dq_man, minSman, SPDdist = robot.calculate_inverse_differential_kinematics_velocity_manipulability(Jcom, des_vel_manip, Km)
# Logging
# Format: [q minEigvalue(Jbal) minEigvalue(Jman) balanceError CurrentManip(1x9) SPDdistance]
log_array[i,] = np.hstack((qt.reshape(1, num_dofs), np.min(Sbal).reshape(1, 1),
minSman.reshape(1, 1), dx_bal.T, vel_manip[0:3, 0:3].reshape(1, vel_manip[0:3, 0:3].size),
SPDdist.reshape(1, 1)))
# Set joint position
if not robot.has_fixed_base():
dq_man = np.concatenate((np.zeros((6,)), dq_man))
dq = dq_bal + np.dot(Nbal, dq_man)
dq = dq[6:, ]
else:
dq = dq_bal + np.dot(Nbal, dq_man)
q = qt + (dq * dt)
robot.set_joint_positions(q)
world.step(sleep_dt=dt)
# Saving log data
# np.savetxt(robot.name + 'log_Man' + str(num_des_manip) + 'Pos' + str(init_qs) + '.csv', log_array, delimiter=',')
# Plotting logged data
fig1 = plt.figure(1, figsize=(14, 10))
# wspace: width reserved for blank space between subplots, hspace: height reserved for white space between subplots
fig1.subplots_adjust(left=0.09, bottom=0.05, right=0.99, wspace=0.2)
plt.suptitle('Robot joints')
plt.rcParams.update({'font.size': 8})
for i in range(num_dofs):
if robot.name == 'nao':
plt.subplot(6, 7, i+1) # NAO
elif robot.name == 'centauro':
plt.subplot(7, 7, i + 1) # Centauro
plt.ylabel(robot.get_joint_names(robot.get_joint_ids(i)))
plt.plot(log_array[:, i])
plt.ylim((-1.5, 1.5))
# fig1.savefig(robot.name + '_joints_Man' + str(num_des_manip) + 'Pos' + str(init_qs) + '.png',
# bbox_inches='tight', dpi=200)
fig2 = plt.figure(2, figsize=(14, 10))
fig2.subplots_adjust(left=0.09, bottom=0.05, right=0.99, wspace=0.2)
plt.rcParams.update({'font.size': 12})
plt.suptitle('MinEigenvalues, balance and manipulatility errors.')
plt.rcParams.update({'font.size': 8})
if robot.name == 'nao':
plt.subplot(4, 3, 1) # NAO
plt.ylim((0, 0.05)) # NAO
elif robot.name == 'centauro':
plt.subplot(6, 3, 1) # Centauro
plt.ylim((0, 0.02)) # Centauro
plt.ylabel('minEigvalue(Jbal)')
plt.plot(log_array[:, num_dofs])
if robot.name == 'nao':
plt.subplot(4, 3, 2) # NAO
plt.ylim((0., 0.0001)) # NAO
elif robot.name == 'centauro':
plt.subplot(6, 3, 2) # Centauro
plt.ylim((0., 0.001)) # Centauro
plt.ylabel('minEigvalue(Jman)')
plt.plot(log_array[:, num_dofs + 1])
if robot.name == 'nao':
plt.subplot(4, 3, 3) # NAO
plt.ylim((0., 4.0)) # NAO
elif robot.name == 'centauro':
plt.subplot(6, 3, 3) # Centauro
plt.ylim((0., 5.0)) # Centauro
plt.ylabel('SPDdist')
plt.plot(log_array[:, -1])
for i in range(dx_bal.shape[0]):
if robot.name == 'nao':
plt.subplot(4, 3, i+4) # NAO
elif robot.name == 'centauro':
plt.subplot(6, 3, i + 4) # Centauro
plt.ylabel('dx_bal'+str(i+1))
plt.plot(log_array[:, num_dofs + 2 + i])
plt.ylim((-.05, .05))
# fig2.savefig(robot.name + '_eigValsAndErrors_Man' + str(num_des_manip) + 'Pos' + str(init_qs) + '.png',
# bbox_inches='tight')
# plt.tight_layout()
plt.show()