cs.RO（2026-05-11）

📊 共 21 篇论文 | 🔗 1 篇有代码

🎯 兴趣领域导航

支柱一：机器人控制 (Robot Control) (14 🔗1) 支柱三：空间感知与语义 (Perception & Semantics) (3) 支柱九：具身大模型 (Embodied Foundation Models) (2) 支柱二：RL算法与架构 (RL & Architecture) (2)

🔬 支柱一：机器人控制 (Robot Control) (14 篇)

#题目一句话要点标签🔗⭐
1 Explicit Stair Geometry Conditioning for Robust Humanoid Locomotion 提出显式楼梯几何条件化框架，提升人形机器人楼梯攀爬的鲁棒性 humanoid humanoid locomotion locomotion
2 VEGA: Visual Encoder Grounding Alignment for Spatially-Aware Vision-Language-Action Models 提出VEGA框架，通过视觉编码器对齐提升视觉-语言-动作（VLA）模型的空间感知能力 manipulation 3D gaussian splatting gaussian splatting
3 HarmoWAM: Harmonizing Generalizable and Precise Manipulation via Adaptive World Action Models 提出HarmoWAM框架，通过自适应世界动作模型实现机器人通用性与操作精度的统一 manipulation world model world models
4 StereoPolicy: Improving Robotic Manipulation Policies via Stereo Perception 提出StereoPolicy框架，通过立体视觉感知提升机器人操作策略的几何推理能力 manipulation mobile manipulation bi-manual
5 Unified Noise Steering for Efficient Human-Guided VLA Adaptation 提出UniSteer框架，通过动作到噪声的反演实现高效的人机协作VLA模型适应 manipulation reinforcement learning flow matching
6 ALAM: Algebraically Consistent Latent Transitions for Vision-Language-Action Models 提出代数一致性潜在动作模型（ALAM），通过结构化视频先验提升视觉-语言-动作（VLA）模型的泛化能力。 manipulation flow matching vision-language-action
7 Retrieve-then-Steer: Online Success Memory for Test-Time Adaptation of Generative VLAs 提出Retrieve-then-Steer框架，通过在线成功记忆实现生成式VLA模型的测试时自适应 manipulation flow matching vision-language-action
8 PriorVLA: Prior-Preserving Adaptation for Vision-Language-Action Models 提出PriorVLA框架，通过先验保持机制实现视觉-语言-动作模型的高效适应 manipulation vision-language-action VLA
9 EFGCL: Learning Dynamic Motion through Spotting-Inspired External Force Guided Curriculum Learning 提出外力引导课程学习（EFGCL）框架，通过物理辅助加速四足机器人动态全身运动学习。 quadruped legged robot reinforcement learning
10 HeteroGenManip: Generalizable Manipulation For Heterogeneous Object Interactions 提出HeteroGenManip框架，通过解耦抓取与交互规划实现异构物体的泛化操作 manipulation diffusion policy foundation model
11 JODA: Composable Joint Dynamics for Articulated Objects 提出JODA框架：通过可组合的关节动力学建模实现铰接物体的精细化物理仿真 manipulation differentiable simulation embodied AI
12 Decentralized Contingency MPC based on Safe Sets for Nonlinear Multi-agent Collision Avoidance 提出一种基于安全集的去中心化应急模型预测控制（MPC）框架，解决非线性多智能体无通信避障问题。 MPC
13 Muninn: Your Trajectory Diffusion Model But Faster 提出Muninn：一种无需重训练的轨迹扩散模型加速框架，实现实时机器人规划 manipulation multimodal ✅
14 VRA: Grounding Discrete-Time Joint Acceleration in Voltage-Constrained Actuation 提出电压可实现加速度（VRA）框架，解决电压受限执行器下的运动控制失真问题。 quadruped

🔬 支柱三：空间感知与语义 (Perception & Semantics) (3 篇)

#题目一句话要点标签🔗⭐
15 MAGS-SLAM: Monocular Multi-Agent Gaussian Splatting SLAM for Geometrically and Photometrically Consistent Reconstruction 提出MAGS-SLAM：首个基于单目视觉的多智能体3D高斯溅射SLAM框架 metric depth 3D gaussian splatting 3DGS
16 ConsistNav: Closing the Action Consistency Gap in Zero-Shot Object Navigation with Semantic Executive Control 提出ConsistNav框架，通过语义执行控制解决零样本物体导航中的动作一致性鸿沟问题 open-vocabulary open vocabulary
17 ObjView-Bench: Rethinking Difficulty and Deployment for Object-Centric View Planning 提出ObjView-Bench评估框架，通过解耦难度维度与部署约束重构以物体为中心的视角规划评估体系。 3D reconstruction

🔬 支柱九：具身大模型 (Embodied Foundation Models) (2 篇)

#题目一句话要点标签🔗⭐
18 Embodied AI in Action: Insights from SAE World Congress 2026 on Safety, Trust, Robotics, and Real-World Deployment SAE 2026白皮书：构建具身智能系统的安全、信任与全生命周期治理框架 embodied AI
19 RoboMemArena: A Comprehensive and Challenging Robotic Memory Benchmark 提出RoboMemArena基准与PrediMem架构，解决机器人长时记忆与复杂任务规划难题 VLA multimodal

🔬 支柱二：RL算法与架构 (RL & Architecture) (2 篇)

#题目一句话要点标签🔗⭐
20 Data-Asymmetric Latent Imagination and Reranking for 3D Robotic Imitation Learning 提出DALI-R框架，通过数据非对称潜空间想象与重排序提升3D机器人模仿学习性能 imitation learning flow matching world model
21 Network-Efficient World Model Token Streaming 提出一种基于自适应关键帧与增量更新的离散世界模型状态流式传输算法 world model world models

⬅️ 返回 cs.RO 首页 · 🏠 返回主页