Robot

The RobotArm class holds all information about the robot kinematics, its operating limits, its link and end-effector geometries. It allows to calculate the forward or inverse kinematics between joint and Cartesian taks space. A list of all supported robots is given at the end of this document.

Limits

By default, Jacobi will use robot limits specified by the manufacturer - or reasonable values in case noen are known publicly. You can optionally overwrite the joint limits via

PythonC++

robot.max_position = [2.97, 2.53, 3.72, 3.32, 2.18, 6.28]  # [rad]
robot.min_position = [-2.97, -1.75, -1.22, -3.32, -2.18, -6.28]  # [rad]

robot.max_velocity = [6.46, 5.41, 7.16, 9.60, 9.51, 17.45]  # [rad/s]
robot.max_acceleration = [5.0, 5.0, 5.0, 10.0, 10.0, 12.0]  # [rad/s^2]
robot.max_jerk = [1e3, 1e3, 1e3, 1e3, 1e3, 1e3]  # [rad/s^3]

to restrict or slow down calculated motions for your application. Jacobi also supports optional minimum velocity and acceleration limits (min_velocity and min_acceleration) for asymmetric constraints.

Frames

Geometric poses and transformations are represented by the Frame class. It can be constructed in several ways, including from translation only, from Euler angles, from Quaternions, or from affine transformation matrices.

PythonC++

grasp_pose = Frame(x=0.5, y=0.2, z=0.75)  # [m]

# Euler angles (intrinsic ZYX convention)
grasp_pose = Frame.from_euler(x=0.5, y=0.2, z=0.75, a=0.0, b=0.0, c=0.0)  # [m] [rad]

# Quaternion (w, x, y, z convention)
grasp_pose = Frame.from_quaternion(x=0.5, y=0.2, z=0.75, qw=1.0, qx=0.0, qy=0.0, qz=0.0)

# Either flattened 16 or 4x4 affine matrix
grasp_pose = Frame.from_matrix([0.8656, ..., 1.0])

Frames support a range of features that you’d expect: They can be multiplied or inversed via .inverse(), you can get the translational or rotational part via .translation or .rotation, and you can interpolate between two frames via a.interpolate(b, t=0.5).

End-Effector

The end-effector (e.g. a gripper or other custom tooling) is attached to the robot’s flange. It can be specified by an obstacle for collision checking and the transformation between the robot’s flange and the new tool center point (TCP):

PythonC++

robot.end_effector_obstacle = Convex.load_from_file(project_path / 'ee_hull.obj')
robot.flange_to_tcp = Frame.from_euler(0, 0, 0, np.pi, 0, np.pi/2)  # [m, rad]

Note that an end-effector is optional. By default, the flange is used interchangbly for the robot’s TCP.

Item Obstacle

An item obstacle (e.g. for a grasped item) can be attached easily at the robot’s TCP. For both the item_obstacle, as well as the end_effector_obstacle, we can either set an Obstacle

PythonC++

robot.item_obstacle = Obstacle(object=Box(0.2, 0.2, 0.1), origin=Frame(z=0.05))  # [m]

or an Object without any offset.

Forward & Inverse Kinematics

In case the robot’s base is not located at the environment origin, please update the pose first by setting

PythonC++

robot.base = Frame.from_translation(0.0, 0.0, 0.5)  # [m]

The forward kinematics maps the robot’s joint position to the Cartesian end-effector pose. In this regard, the tool center point (TCP) of the robot can be calculated via

PythonC++

tcp_frame = robot.calculate_tcp([1.0, 2.0, 0.2, 0.1, 0.0, 0.0])  # [rad]

Note that the TCP is equalivalent to the robot’s flange in case no end-effector was specified.

The inverse kinematics (IK) maps the Cartesian end-effector to a joint position and is, in general, not unique and computationally much more complex. Jacobi supports an analytical IK for all 6-DoF robot arms adn a numerical IK for all redundant robots (e.g. the Kuka iiwa7 or the Franka Emika robot). Without any additional information, Jacobi’s inverse kinematics will select the best solution (satisfying the joint position limits) in case multiple exists and will return an empty solution if none exist.

PythonC++

grasp_pose = Frame.from_euler(0.5, 0.0, 0.75, 0.393, 0.785, 0.0)  # [m] [rad]
new_position = robot.inverse_kinematics(grasp_pose)  # Might return None

You can specify an initial joint position to change the solution (in case multiple exist) via robot.inverse_kinematics(grasp_pose, reference_config). The analytical IK will return the nearest solution (according to the joint space distance), while the numerical IK will use it as its initial solution for the iterative update step.

Custom Link Objects

Jacobi includes collision models of all supported robots. If you want to specify custom collision models for the robot links (so beside the end-effector, e.g. for cable housing), you can overwrite a default link via

PythonC++

robot.link_obstacles[2] = Convex.load_from_file(project_path / 'upper_with_backpack.obj')

Please make sure that your link objects have their base in the corresponding link frame e.g. link_frames[2].

Supported Robots

Jacobi currently supports the following list of robots. Reach out to us on https://jacobirobotics.com/get-started to get your robot supported!

Manufacturer	Robot Model	Driver
ABB	IRB 1200-5/90	ABBDriver
	IRB 1300-7/1.4	ABBDriver
	IRB 1600-6/1.2	ABBDriver
	IRB 2600-12/185	ABBDriver
	IRB 4600-60/205	ABBDriver
	IRB 6700-150/320	ABBDriver
	IRB 14000 YuMi	ABBDriver
FANUC	CRX-30iA
	LR 10-iA/10
	LRMate 200iD/7L
	M 20-iB/25
	M 710-iC/45M
Franka	Panda
Kinova	Gen3 7DoF
KUKA	KR6 R700 sixx
	KR70 R2100
	LBR iiwa 7
Mecademic	Meca500
UFACTORY	xArm 7
Universal	UR5e	UniversalDriver
	UR10	UniversalDriver
	UR10e	UniversalDriver
	UR20	UniversalDriver
Yaskawa	GP12	YaskawaDriver
	GP180	YaskawaDriver
	GP180-120	YaskawaDriver
	GP50	YaskawaDriver
	HC10	YaskawaDriver
	HC20	YaskawaDriver

Loading Custom Robots

Jacobi is robot agnostic. So if you do not see your desired robot in our robot library, you can use your own URDF (Unified Robot Description Format) file that represents the robot. You must provide the name of the base link of your kinematic chain as well as the end link of your robot kinematic chain. Note that custom robots are currently in beta preview with rudimentary support. Please reach out to us if you encounter any issues.

Custom robot arms are loaded using the CustomRobot class, which extends the RobotArm class. It provides all methods and properties we described above, like forward and inverse kinematics, specifying robot limits and adding end-effector and item obstacles.

PythonC++

Copy code

robot = CustomRobot.load_from_urdf_file(
    file='path/to/your_robot.urdf',
    base_link='base_link_name',
    end_link='flange_link_name'
)

There are two requirements for the URDF file you want to load: all links require a single mesh file for collision checking, provided as a .obj file, and all links need to be connected via joints from the base link to the end link.