teleopro_core

Teleoperate any robot with any input device, in pure Python (numpy only, no ROS). It turns the motion of a device (XR headset, tracker, mocap, ...) into a target pose for the robot, with clutch, speed limits and input-loss handling.

ROS users: use the teleopro_ros package, which wraps this library with TF.

Install

pip install "git+https://github.com/ShifeiDuan/teleopro#subdirectory=teleopro_core"

Usage

import time
from teleopro_core import Teleop, as_pos_quat

teleop = Teleop(translation_scale=0.5, max_lin_vel=0.1)

teleop.engage()                       # e.g. when a button is pressed
while True:
    device_pose = my_device.get_pose()       # ((x, y, z), (qx, qy, qz, qw)) or 4x4, None if not tracked
    robot_pose = my_robot.get_tcp_pose()     # same formats, in the robot base frame
    target = teleop.update(time.monotonic(), device_pose, robot_pose)
    if target is not None:
        my_robot.move_to(*as_pos_quat(target))
    time.sleep(0.01)

• On engage() the target starts at the current robot pose, then follows the device's motion since that moment. disengage() holds it still (clutch).
• If the device is not tracked (None) or its input_stamp is older than input_timeout, the target holds; when tracking returns it continues from where it stopped, without jumping.
• Before the first engage, update() returns the robot pose latched once.

Velocity devices (keyboard, SpaceMouse) use update_twist() instead:

teleop = Teleop(twist_frame='tool')      # or 'base'
target = teleop.update_twist(time.monotonic(), ((vx, vy, vz), (wx, wy, wz)), robot_pose)

Parameters

License

Apache-2.0