Adaptive Teleoperation Mimicry (ATOM)

ATOM is a ROS 2-based teleoperation framework for real-time human-to-robot motion imitation. It maps human wrist and head motion to robot joint trajectories using adaptive inverse kinematics (IK).

---

Overview

• Real-time dual-arm and neck teleoperation
• Adaptive IK with retries, perturbations, and reach projection
• Optional motion scaling and trajectory smoothing
• Compatible with any robot publishing a valid URDF and TF tree

---

Dependencies

ROS 2 (Humble or newer) Python packages:

• rclpy
• tf2_ros
• kdl_parser_py
• PyKDL
• numpy

---

Installation

cd ~/ros2_ws/src
git clone https://github.com/yourusername/adaptive-teleoperation-mimicry.git
cd ~/ros2_ws
colcon build
source install/setup.bash

---

Usage

Launch your robot description and TF publishers, then run:

ros2 run teleop_controller tf_listener
ros2 launch openarm_bringup openarm.bimanual.launch.py use_fake_hardware:=true
ros2 run controller_manager spawner neck_joint_trajectory_controller
ros2 run teleop_controller neck_tracker

ATOM will:

• Retrieve /robot_description

• Build KDL chains for arms and neck

• Start tracking TF frames:

  • left_wrist_pose_meta
  • right_wrist_pose_meta
  • head_stable_meta

---

Parameters

Name	Default	Description
control_rate	20.0	Control loop rate (Hz)
trajectory_duration	0.1	Trajectory duration (s)
enable_smoothing	True	Apply EMA smoothing
smoothing_alpha	0.7	EMA weighting factor
reach_tolerance	1.2	Reach tolerance multiplier
ik_max_attempts	10	IK retries
ik_perturbation_range	1.0	Perturbation range (rad)
ik_eps	0.01	IK solver epsilon
ik_maxiter	2000	IK solver iterations
ik_publish_on_failure	True	Publish best IK on failure
ik_failure_tol	0.05	Max IK error to publish (m)
project_out_of_reach	True	Project unreachable targets
projection_factor	0.95	Reach projection factor
use_relative_scaling	True	Enable motion scaling
motion_scale	0.5	Relative motion scale

---

Example

ATOM expects TF frames representing tracked human motion. These frames are converted to robot joint trajectories published to:

• /left_joint_trajectory_controller/joint_trajectory
• /right_joint_trajectory_controller/joint_trajectory
• /neck_joint_trajectory_controller/joint_trajectory

---

License

MIT License