UTTG_ A Universal Teleoperation Approach via Online Trajectory Generation

TL;DR

The paper tackles cross-platform teleoperation by addressing hardware heterogeneity and the mismatch between low-rate human inputs and high-rate robot controllers. It introduces UTTG, a URDF-driven framework that outputs high-frequency joint commands from low-frequency inputs via online trajectory generation based on cubic splines, with a min-stretch objective and a tunable parameter $$\mu\$ in $[0,1]$ to balance accuracy and smoothness. Two operation modes—precise and rapid—enable task-specific performance, and the method demonstrates improved trajectory quality and task success across Realman, Aloha, and Franka, including a $92\%$ MAV reduction and a $36\%$ rise in complex-task success. These results suggest practical impact for collecting expert demonstrations and building embodied intelligence datasets, with future work on human motion prediction and reactive obstacle avoidance networks.

Abstract

Teleoperation is crucial for hazardous environment operations and serves as a key tool for collecting expert demonstrations in robot learning. However, existing methods face robotic hardware dependency and control frequency mismatches between teleoperation devices and robotic platforms. Our approach automatically extracts kinematic parameters from unified robot description format (URDF) files, and enables pluggable deployment across diverse robots through uniform interfaces. The proposed interpolation algorithm bridges the frequency gap between low-rate human inputs and high-frequency robotic control commands through online continuous trajectory generation, \n{while requiring no access to the closed, bottom-level control loop}. To enhance trajectory smoothness, we introduce a minimum-stretch spline that optimizes the motion quality. The system further provides precision and rapid modes to accommodate different task requirements. Experiments across various robotic platforms including dual-arm ones demonstrate generality and smooth operation performance of our methods. The code is developed in C++ with python interface, and available at https://github.com/IRMV-Manipulation-Group/UTTG.

Figures (4)

• Figure 1: Illustrations of rapid mode and precise mode.
• Figure 2: UTTG framework: We utilize URDF of robot to automatically generate the required parameter files, decoupling human input from specific robot configurations. Our input can be either the end-effector pose or the joint angles. Using the generated files, we perform IK to compute the corresponding joint angles, and the output consists of high-frequency joint position commands for the robot.
• Figure 3: Snapshots in different platforms: (A) Realman with motion-capture suit, (B) Aloha using bimanual joint mapping, (C) Franka Research 3 Tasks. We refer readers to our video for better presentation.
• Figure 4: \ref{['draw_ex1']}a and \ref{['draw_ex1']}b present the comparison of generated trajectories between UTTG framework and Realman algorithm. Our framework interpolates joint positions before sending commands to the motor, and the Realman algorithm optimizes and sends positions in the internal control loop, which is unreachable thus cannot be migrated to other robot platform.