Robot Paper Digest历史日报

2026-11-05 机器人论文日报

候选论文 300 篇,精选 5 篇。

#1 · 2611.06649 · 相关度 9/10

SCRUB: Contact-Based Surface-Frame Estimation for Force-Motion Control with an Armless Quadruped

Sourav Selvaraj, Juan Pablo Espinosa Chessal, Mahdi Agheli

中文摘要

提出SCRUB,一种基于接触的表面坐标系估计方法,使无臂四足机器人能在未知表面上以受控力滑动工具。控制器利用卡尔曼滤波融合力线索与速度线索估计表面法向,并减去指令按压速率、按步态周期平均以补偿行走干扰;在Unitree Go1上完成全部爬墙与绕柱任务,消融实验验证两线索、步态平均和在线估计均不可缺。

English summary

SCRUB estimates the surface frame from contact so an armless quadruped can slide a body-mounted tool on unknown near-vertical surfaces. A predictive controller presses and slides using a Kalman-filter normal estimate fusing a friction-robust force cue and a gait-averaged velocity cue. On a Unitree Go1 it completes every wall traverse and column orbit; ablations removing cues, gait averaging, or online estimation all fail.

与你的研究相关性

面向四足机器人本体-环境接触力控与状态估计,直接关联四足运动与loco-manipulation。

阅读 arXiv 原文 →

#2 · 2611.05507 · 相关度 8.5/10

Ctrl-F: Force as an Interface for Humanoid Loco-Manipulation

Mengjie Zhao, Yuxin Chen, Yiyang Shao, Chuye Hong, Jiakang Chen, Koushil Sreenath, Masayoshi Tomizuka

中文摘要

提出Ctrl-F分层运动-力框架,将运动跟踪扩展为带显式接触力指令的控制:底层统一运动-力控制器跟踪参考运动并向世界施加指令力,高层力规划器依据物体状态与本体感知在线生成力指令。仿真与硬件完成手部拖拽、推、举及足部踢击,可推动达20kg箱体并在负载变化时在线调整。

English summary

Ctrl-F augments humanoid motion tracking with explicit 3-D contact force commands at designated bodies. A unified motion-force controller tracks reference motion while applying commanded forces, and a high-level planner issues forces online from object state and proprioception. Simulation and hardware show hand dragging, pushing, lifting, and foot kicking of boxes up to 20 kg with online load adaptation.

与你的研究相关性

涉及足式机器人全身控制与接触力接口,对四足loco-manipulation的WBC设计有借鉴意义。

阅读 arXiv 原文 →

#3 · 2611.05585 · 相关度 8/10

Contact-Implicit Whole-Body Trajectory Optimization for Dynamic Humanoid Locomotion on Parameterizable Uneven Terrain

Roberto Mauceri, Stefano Dafarra, Giulio Romualdi, Marco Gabiccini, Daniele Pucci

中文摘要

提出面向人形机器人动态运动的接触隐式全身轨迹优化规划器,无需预定义接触序列,同时考虑全身动力学与运动学。用光滑曲面模型近似台阶、斜坡等地面几何,并以新的接触约束将适用范围从水平面扩展到任意光滑面;通过非线性规划在多种地形上验证,并在ErgoCub实物上完成初步可行性测试。

English summary

This letter presents a contact-implicit whole-body trajectory planner for humanoid locomotion on parameterizable uneven terrain without predefined contact sequences. A smooth surface model and novel contact constraints extend applicability from flat ground to arbitrary smooth surfaces. Nonlinear programming yields dynamic motions validated in simulation and on the real ErgoCub robot.

与你的研究相关性

接触隐式优化与全身动力学规划直接关联足式运动与MPC/WBC轨迹生成。

阅读 arXiv 原文 →

#4 · 2611.06564 · 相关度 7.5/10

Playing Soccer Like Humans Do: Learning Versatile and Agile Humanoid Soccer via Latent Motion Priors

Haoyu Shao, Yunshen Wang, Jingze Zhang, Shaohang Zhu, Yong-Lu Li, Tianzhu Zhang, Baoxiong Jia, Siyuan Huang

中文摘要

提出soccerbot分层强化学习框架,从大规模人类足球动作捕捉数据蒸馏结构化潜在运动先验,将策略学习约束在动力学可行的运动流形内,使单一技能条件策略统一学习移动、带球、停球和踢球并产生类人运动与流畅切换;经鲁棒性仿真设计后零样本迁移至Unitree G1,实现敏捷自然的完整足球技能。

English summary

soccerbot is a hierarchical RL framework that distills a structured latent motion prior from large-scale human soccer mocap, constraining policy learning to a dynamically plausible motion manifold. One skill-conditioned policy learns moving, dribbling, trapping, and kicking with human-like transitions, transferring zero-shot to the Unitree G1 humanoid for agile, robust real-world soccer.

与你的研究相关性

涉及足式机器人的运动先验、分层RL与sim-to-real,方法与四足运动学习相通。

阅读 arXiv 原文 →

#5 · 2611.06196 · 相关度 7/10

Beyond Terminal Constraints: Reliable Robot Control with Learned Safe Sets

Gianni Lunardi, Elias Fontanari, Matteo Saveriano, Andrea Del Prete

中文摘要

指出安全性不必仅依赖控制不变集,引入比经典控制不变集更广的几类安全集,并分析通过在预测时域不同节点施加约束而非仅终端状态来保证递归可行性的MPC方案;当安全集为控制不变集的内近似时,即使无递归可行性也可借安全中止策略保证安全。在6关节机械臂仿真中约束违反少于对比方法,并完成硬件实验。

English summary

This work introduces new classes of safe sets broader than control-invariant sets and analyzes MPC schemes that constrain several prediction-horizon nodes rather than only the terminal state to guarantee recursive feasibility. A safe-abort strategy ensures safety without recursive feasibility. Simulations on a 6-joint manipulator show fewer constraint violations with comparable tracking and hardware practicality.

与你的研究相关性

为足式机器人MPC与RL控制提供安全约束设计思路,关联WBC中的安全性保障。

阅读 arXiv 原文 →