pykin 0.2.7

Robotics Kinematics Library

pykin

Python Interface for the Robot Kinematics Library pykin

This library has been created simply by referring to ikpy.

Features

• Pure python library
• Support only URDF file
• Compute Forward, Inverse Kinematics and Jacobian
• There are two ways to find the IK solution, referring to the Introduction to Humanoid Robotics book.
• Compute Collision checkinkg
• Plot Robot Kinematic Chain and Robot Mesh (STL file)

Installation

Requirements

You need a python-fcl package to do object collision checking.

• For Ubuntu, using apt

  sudo apt install liboctomap-dev

  sudo apt install libfcl-dev

• For Mac, First, Download the source and build it.

  • octomap

    git clone https://github.com/OctoMap/octomap.git

    $ cd octomap
    $ mkdir build
    $ cd build
    $ cmake ..
    $ make
    $ make install
    

  • fcl

    git clone https://github.com/flexible-collision-library/fcl.git

    Since python-fcl uses version 0.5.0 of fcl, checkout with tag 0.5.0

    $ cd fcl
    $ git checkout 0.5.0
    $ mkdir build
    $ cd build
    $ cmake ..
    $ make
    $ make install
    

If the above installation is complete

pip install python-fcl

Install Pykin

pip install pykin

When git clone, use the --recurse-submodules option.

The download may take a long time due to the large urdf file size.

git clone --recurse-submodules https://github.com/jdj2261/pykin.git

Quick Start

• Robot Info

  You can see 4 example robot information.

  baxter, iiwa14, panda, and sawyer

  Code

  import sys
  from pykin.robot import Robot
  
  file_path = '../asset/urdf/baxter/baxter.urdf'
  if len(sys.argv) > 1:
      robot_name = sys.argv[1]
      file_path = '../asset/urdf/' + robot_name + '/' + robot_name + '.urdf'
  
  robot = Robot(file_path)
  robot.show_robot_info()
  

• Forward Kinematics

  Code

  from pykin.robot import Robot
  from pykin.kinematics.transform import Transform
  from pykin.utils.kin_utils import ShellColors as sc
  
  # baxter_example
  file_path = '../asset/urdf/baxter/baxter.urdf'
  robot = Robot(file_path, Transform(rot=[0.0, 0.0, 0.0], pos=[0, 0, 0]))
  
  # set input joints 
  head_thetas = [0.0]
  right_arm_thetas = [0, 0, 0, 0, 0, 0, 0]
  left_arm_thetas = [0, 0, 0, 0, 0, 0, 0]
  thetas = head_thetas + right_arm_thetas + left_arm_thetas
  
  # compute FK
  fk = robot.kin.forward_kinematics(thetas)
  for link, transform in fk.items():
      print(f"{sc.HEADER}{link}{sc.ENDC}, {transform.rot}, {transform.pos}")
  

• Jacobian

  Code

  from pykin.kinematics import transform as tf
  from pykin.robot import Robot
  
  # import jacobian
  from pykin.kinematics import jacobian as jac
  
  file_path = '../asset/urdf/baxter/baxter.urdf'
  robot = Robot(file_path, tf.Transform(rot=[0.0, 0.0, 0.0], pos=[0, 0, 0]))
  
  left_arm_thetas = [0, 0, 0, 0, 0, 0, 0]
  
  # Before compute Jacobian, you must set from start link to end link
  robot.set_desired_frame("base", "left_wrist")
  fk = robot.kin.forward_kinematics(left_arm_thetas)
  
  # If you want to get Jacobian, use calc_jacobian function
  J = jac.calc_jacobian(robot.desired_frames, fk, len(left_arm_thetas))
  print(J)
  
  right_arm_thetas = [0, 0, 0, 0, 0, 0, 0]
  robot.set_desired_frame("base", "right_wrist")
  fk = robot.kin.forward_kinematics(right_arm_thetas)
  J = jac.calc_jacobian(robot.desired_frames, fk, len(right_arm_thetas))
  print(J)
  

• Inverse Kinematics

  Code

  import numpy as np
  from pykin.robot import Robot
  from pykin.kinematics.transform import Transform
  
  # baxter_example
  file_path = '../asset/urdf/baxter/baxter.urdf'
  robot = Robot(file_path, Transform(rot=[0.0, 0.0, 0.0], pos=[0, 0, 0]))
  
  # set joints for targe pose
  right_arm_thetas = np.random.randn(7)
  
  # set init joints
  init_right_thetas = np.random.randn(7)
  
  # Before compute IK, you must set from start link to end link
  robot.set_desired_frame("base", "right_wrist")
  
  # Compute FK for target pose
  target_fk = robot.kin.forward_kinematics(right_arm_thetas)
  
  # get target pose
  target_r_pose = np.hstack((target_fk["right_wrist"].pos, target_fk["right_wrist"].rot))
  
  # Compute IK Solution using LM(Levenberg-Marquardt) or NR(Newton-Raphson) method
  ik_right_result, _ = robot.kin.inverse_kinematics(init_right_thetas, target_r_pose, method="LM")
  
  # Compare error btween Target pose and IK pose
  result_fk = robot.kin.forward_kinematics(ik_right_result)
  error = robot.compute_pose_error(
      target_fk["right_wrist"].homogeneous_matrix,
      result_fk["right_wrist"].homogeneous_matrix)
  print(error)
  

• Self-Collision Check

  Code

  import numpy as np
  
  from pykin.kinematics.transform import Transform
  from pykin.robot import Robot
  
  """
  If you want to check robot's collision, install python-fcl 
  And then, import FclManager in fcl_utils package
  """
  from pykin.utils.fcl_utils import FclManager
  from pykin.utils.kin_utils import get_robot_geom
  from pykin.utils import plot_utils as plt
  
  file_path = '../asset/urdf/baxter/baxter.urdf'
  
  robot = Robot(file_path, Transform(rot=[0.0, 0.0, 0.0], pos=[0, 0, 0]))
  
  head_thetas = np.zeros(1)
  right_arm_thetas = np.array([np.pi, 0, 0, 0, 0, 0, 0])
  left_arm_thetas = np.array([-np.pi, 0, 0, 0, 0, 0, 0])
  
  thetas = np.hstack((head_thetas, right_arm_thetas, left_arm_thetas))
  transformations = robot.kin.forward_kinematics(thetas)
  
  # call FclManager class
  fcl_manager = FclManager()
  for link, transformation in transformations.items():
      # get robot link's name and geometry info 
      name, gtype, gparam = get_robot_geom(robot.links[link])
      # get 4x4 size homogeneous transform matrix
      transform = transformation.homogeneous_matrix
      # add link name, geometry info, transform matrix to fcl_manager 
      fcl_manager.add_object(name, gtype, gparam, transform)
  
  # you can get collision result, contacted object name, fcl contatct_data
  result, objs_in_collision, contact_data = fcl_manager.collision_check(return_names=True, return_data=True)
  
  print(result, objs_in_collision, contact_data)
  
  """
  If you want to check collision check after transform, 
  add the link name and transform matrix to the set_transform function.
  """
  left_arm_thetas = np.array([0, 0, 0, 0, 0, 0, 0])
  thetas = np.hstack((head_thetas, right_arm_thetas, left_arm_thetas))
  transformations = robot.kin.forward_kinematics(thetas)
  
  for link, transformation in transformations.items():
      name, _, _ = get_robot_geom(robot.links[link])
      transform = transformation.homogeneous_matrix
      fcl_manager.set_transform(name=name, transform=transform)
  
  result, objs_in_collision, contact_data = fcl_manager.collision_check(return_names=True, return_data=True)
  print(result, objs_in_collision, contact_data)
  

Visualization

• urdf

  You can see visualization using matplotlib.

baxter	sawyer	iiwa14	panda

Code

import sys

from pykin.robot import Robot
from pykin.utils import plot_utils as plt

file_path = '../../asset/urdf/sawyer/sawyer.urdf'

if len(sys.argv) > 1:
    robot_name = sys.argv[1]
    file_path = '../../asset/urdf/' + robot_name + '/' + robot_name + '.urdf'
robot = Robot(file_path)

fig, ax = plt.init_3d_figure("URDF")

# For Baxter robots, the name argument to the plot_robot function must be baxter.
plt.plot_robot(robot, 
               transformations=robot.transformations,
               ax=ax, 
               name=robot.robot_name,
               visible_visual=False, 
               visible_collision=False, 
               mesh_path='../../asset/urdf/baxter/')
ax.legend()
plt.show_figure()

• collision

  You can see collision defined in collision/geometry tags in urdf.

baxter	sawyer

Code

import sys

from pykin.robot import Robot
from pykin.utils import plot_utils as plt

file_path = '../../asset/urdf/baxter/baxter.urdf'

if len(sys.argv) > 1:
    robot_name = sys.argv[1]
    file_path = '../../asset/urdf/' + robot_name + '/' + robot_name + '.urdf'
robot = Robot(file_path)

fig, ax = plt.init_3d_figure("URDF")

"""
Only baxter and sawyer robots can see collisions.
It is not visible unless sphere, cylinder, and box are defined in collision/geometry tags in urdf.
"""
# If visible_collision is True, visualize collision
plt.plot_robot(robot, 
               transformations=robot.transformations,
               ax=ax, 
               name=robot.robot_name,
               visible_visual=False, 
               visible_collision=True, 
               mesh_path='../../asset/urdf/baxter/')
ax.legend()
plt.show_figure()

• mesh

  You can see mesh defined in visual/geometry tags in urdf.

baxter	sawyer	iiwa14	panda

Code

import sys

from pykin.robot import Robot
from pykin.utils import plot_utils as plt

file_path = '../../asset/urdf/baxter/baxter.urdf'

if len(sys.argv) > 1:
    robot_name = sys.argv[1]
    file_path = '../../asset/urdf/' + robot_name + '/' + robot_name + '.urdf'
robot = Robot(file_path)

fig, ax = plt.init_3d_figure("URDF")

"""
Only baxter and sawyer robots can see collisions.
It is not visible unless sphere, cylinder, and box are defined in collision/geometry tags in urdf.
"""
# If visible_visual is True, visualize mesh
# and you have to input mesh_path
plt.plot_robot(robot, 
               transformations=robot.transformations,
               ax=ax, 
               name=robot.robot_name,
               visible_visual=True, 
               visible_collision=False, 
               mesh_path='../../asset/urdf/'+robot.robot_name+'/')
"""
The mesh file doesn't use matplotlib, 
so it's okay to comment out the line below.
"""
# ax.legend()
# plt.show_figure()

• Planning