DexTele 双臂灵巧遥操作系统：基于运动重定向与自适应力控制

Jul 16, 2026

Upload by Miles

Dual-arm dexterous teleoperation requires cross-platform generalization of motion retargeting and compliant grasping interactivity, yet heterogeneous robot architectures and diverse graspable objects make precise retargeting and safe force regulation difficult. DexTele is proposed using motion retargeting and adaptive force control. A vision-based retargeting module generates initial robot motions from human images via a motion-graph encoder and latent optimization. An adaptive grasping module fuses a vision-language model with MPC and gradient-based online optimization to predict and apply target grasping force, enabling experiments to show precise retargeting and compliant grasping across multiple robot platforms.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages8

Reading Time20 min

Introduction

Motion retargeting across robot platforms

Adaptive grasping for diverse objects

DexTele: A Dual-Arm Dexterous Teleoperation System Based on Motion Retargeting and Adaptive Force Control Yuanchuan Lai 1 , Qing Gao 1 ,∗ , Ziyan Liang 1 , Xianfeng Cheng 1 , Junjie Hu2 , Zhaojie Ju3 arXiv :2607 .05883v 1 [ cs .RO] 7 Jul 2026 Abstract—In dual-arm dexterous teleoperation, crossplatform generalization of motion retargeting and interactivity of grasping are crucial. However, the heterogeneity of robotic architectures and the wide variety of grasping objects pose significant challenges to achieving precise motion retargeting and compliant grasping in dual-arm dexterous teleoperation. To address these challenges, a dual-arm dexterous teleoperation system (DexTele) is proposed based on motion retargeting and adaptive force control. First, a vision-based motion retargeting module is designed to generate preliminary robot motions from human images. In this module, a motion-graph encoder and latent optimization are proposed for precise and convenient crossplatform motion retargeting. Second, an adaptive grasping module is designed to achieve compliant grasping. This module combines a vision-language model (VLM) with model predictive control (MPC), allowing the system to predict the required grasping force for a target object and perform gradient-based online optimization. Finally, extensive experiments demonstrate that the DexTele achieves precise motion retargeting and compliant grasping with generalization across multiple robot platforms. Project can be found at: [https://github.io/DexTele](https://github.io/DexTele). I. INTRODUCTION Robotic teleoperation enables human operators to control robots remotely for complex tasks. A critical component is motion retargeting, which maps human movements to robots for natural and precise reproduction. Existing systems often implement motion retargeting through direct mappings, which perform adequately for simple tasks [1],[2] . However, the limitations of these approaches become evident when extending them to multiple robotic platforms. They are typically designed for a single platform and lack crossplatform generalization, which can easily lead to motion retargeting inaccuracies and compromise the naturalness and reliability of operations on robots with different architectures. In dexterous hand teleoperation, hand motions must not only replicate human gestures but also adapt to object interaction characteristics to ensure precise and safe manipulation [3],[4] . In this context, adaptive grasping is particularly important, as it enables the robot to adjust its actions based This work was supported in part by the Guangdong Basic and Applied Basic Research Foundation under Grant 2025A1515011954, 2023A1515110074, in part by the Shenzhen Science and Technology Program under Grant ZDCY20250901100201002 . 1Yuanchuan Lai, Qing Gao, Ziyan Liang and Xianfeng Cheng are with the School of Electronics and Communica-tion Engineering, Sun Yat-sen University, Shenzhen 518107, China.([email:laiych25@mail2.sysu.edu.cn](email:laiych25@mail2.sysu.edu.cn), [gaoqing2@mail.sysu.edu.cn](gaoqing2@mail.sysu.edu.cn)) 2Junjie Hu is with the School of Artificial Intelligence, The Chinese University of Hong Kong, Shenzhen, Shenzhen 518172, China.(email: [hujunjie@cuhk.edu.cn](hujunjie@cuhk.edu.cn) ) 3Zhaojie Ju is with the School of Computing, University of Portsmouth, Portsmouth PO1 3HE, UK.(email: [Zhaojie.Ju@port.ac.uk](Zhaojie.Ju@port.ac.uk)) ∗ Corresponding Author:Qing Gao, [gaoqing2@mail.sysu.edu.cn](gaoqing2@mail.sysu.edu.cn). Fig. 1. Schematic of the teleoperation system. Part (a) illustrates previous work on motion retargeting, part (b) illustrates previous work on force control, and part (c) presents the proposed pipeline for dexterous teleoperation. on object properties, thereby ensuring stability and safety during manipulation. Overall, cross-platform generalization, retargeting accuracy, and adaptive grasping jointly determine the practicality and reliability of tele

FAQs

DexTele主要解决双臂灵巧遥操作中的哪些关键问题？

它关注跨平台运动重定向的泛化能力以及抓取交互的顺应性与准确性，同时应对机器人架构异构与抓取物体多样带来的挑战。

DexTele的运动重定向模块如何从人的图像生成机器人初始动作？

系统使用基于视觉的重定向模块，通过运动图编码器与潜变量优化来完成跨平台的精确、便捷重定向。

DexTele的自适应抓取模块如何实现顺应性抓取？

它将视觉语言模型与模型预测控制（MPC）结合，预测目标物体所需抓取力，并利用基于梯度的在线优化进行实时调整，从而提升稳定性与安全性。

Jul 16, 2026

Upload by Miles