cs.RO（2026-02-03）

📊 共 24 篇论文 | 🔗 2 篇有代码

🎯 兴趣领域导航

支柱一：机器人控制 (Robot Control) (18 🔗2) 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) (1)

🔬 支柱一：机器人控制 (Robot Control) (18 篇)

#	题目	一句话要点	标签	🔗	⭐
1	CMR: Contractive Mapping Embeddings for Robust Humanoid Locomotion on Unstructured Terrains	提出CMR框架，通过Contractive Mapping提升人形机器人非结构化地形鲁棒运动能力	humanoid humanoid locomotion locomotion
2	RDT2: Exploring the Scaling Limit of UMI Data Towards Zero-Shot Cross-Embodiment Generalization	RDT2：探索UMI数据规模极限，实现机器人零样本跨具身泛化	manipulation flow matching distillation	✅
3	RPL: Learning Robust Humanoid Perceptive Locomotion on Challenging Terrains	RPL：学习在复杂地形上稳健的人形机器人感知运动	humanoid locomotion manipulation
4	HUSKY: Humanoid Skateboarding System via Physics-Aware Whole-Body Control	提出HUSKY框架，实现类人机器人在滑板上进行物理感知全身控制	humanoid whole-body control Unitree	✅
5	AffordanceGrasp-R1:Leveraging Reasoning-Based Affordance Segmentation with Reinforcement Learning for Robotic Grasping	AffordanceGrasp-R1：结合推理和强化学习提升机器人抓取的可供性分割	manipulation reinforcement learning affordance
6	Training and Simulation of Quadrupedal Robot in Adaptive Stair Climbing for Indoor Firefighting: An End-to-End Reinforcement Learning Approach	提出双阶段端到端强化学习方法，提升四足机器人在室内火灾场景下自适应爬楼梯能力	quadruped locomotion Unitree
7	Enhancing Navigation Efficiency of Quadruped Robots via Leveraging Personal Transportation Platforms	提出基于强化学习的主动式运输平台骑行方法，提升四足机器人导航效率。	quadruped legged locomotion locomotion
8	BridgeV2W: Bridging Video Generation Models to Embodied World Models via Embodiment Masks	BridgeV2W：通过具身掩码桥接视频生成模型与具身世界模型，提升机器人操作性能。	dual-arm world model cross-embodiment
9	MVP-LAM: Learning Action-Centric Latent Action via Cross-Viewpoint Reconstruction	MVP-LAM：通过跨视角重建学习动作中心化的潜在动作，用于VLA模型预训练。	manipulation vision-language-action VLA
10	Self-supervised Physics-Informed Manipulation of Deformable Linear Objects with Non-negligible Dynamics	SPiD：基于物理信息的自监督学习框架，用于动态操作可变形线性物体	manipulation sim-to-real
11	Manipulation via Force Distribution at Contact	提出基于力分布线接触模型的操作方法，提升接触操作的效率和鲁棒性	manipulation trajectory optimization
12	Learning-based Initialization of Trajectory Optimization for Path-following Problems of Redundant Manipulators	提出基于学习的轨迹优化初始化方法，加速冗余机械臂路径跟随	trajectory optimization reinforcement learning
13	Learning-based Adaptive Control of Quadruped Robots for Active Stabilization on Moving Platforms	提出LAS-MP，用于四足机器人在移动平台上的主动稳定控制	quadruped
14	Estimation of Ground Reaction Forces from Kinematic Data during Locomotion	提出一种仅依赖运动学数据的步态地面反作用力估计方法	locomotion
15	When Should Agents Coordinate in Differentiable Sequential Decision Problems?	提出基于二阶信息的协调决策方法，解决可微序列决策问题中的多智能体通信时机选择问题	motion planning
16	Variance-Reduced Model Predictive Path Integral via Quadratic Model Approximation	提出基于二次模型近似的方差缩减MPPI方法，提升采样效率。	manipulation
17	Human-in-the-Loop Failure Recovery with Adaptive Task Allocation	提出基于自适应任务分配的人在环故障恢复方法，提升人机协作效率	humanoid
18	Model-based Optimal Control for Rigid-Soft Underactuated Systems	针对刚柔混合欠驱动系统，提出基于模型的优化控制策略，实现动态摆动任务。	model predictive control

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

#	题目	一句话要点	标签	🔗	⭐
19	A Scene Graph Backed Approach to Open Set Semantic Mapping	提出基于场景图的开放集语义地图构建方法，提升机器人环境感知能力	semantic mapping semantic map large language model
20	Depth Completion in Unseen Field Robotics Environments Using Extremely Sparse Depth Measurements	提出基于极稀疏深度信息的深度补全模型，用于未知环境下的机器人导航。	depth estimation monocular depth metric depth
21	Collision Detection with Analytical Derivatives of Contact Kinematics	iDCOL：通过解析导数进行碰撞检测，解决梯度优化中接触动力学不光滑问题	implicit representation

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

#	题目	一句话要点	标签	🔗	⭐
22	ProAct: A Benchmark and Multimodal Framework for Structure-Aware Proactive Response	ProAct：提出一个结构感知的多模态主动响应基准和框架，用于辅助、维护和安全监控等任务。	large language model multimodal
23	When Attention Betrays: Erasing Backdoor Attacks in Robotic Policies by Reconstructing Visual Tokens	提出Bera框架，通过重构视觉tokens擦除机器人策略中的后门攻击。	vision-language-action VLA multimodal

🔬 支柱二：RL算法与架构 (RL & Architecture) (1 篇)

#	题目	一句话要点	标签	🔗	⭐
24	Hierarchical Proportion Models for Motion Generation via Integration of Motion Primitives	提出一种分层比例模型，通过融合运动原语生成机器人运动	imitation learning motion synthesis motion generation

⬅️ 返回 cs.RO 首页 · 🏠 返回主页