cs.RO（2025-09-29）

📊 共 28 篇论文 | 🔗 5 篇有代码

🎯 兴趣领域导航

支柱一：机器人控制 (Robot Control) (21 🔗4) 支柱九：具身大模型 (Embodied Foundation Models) (3 🔗1) 支柱三：空间感知与语义 (Perception & Semantics) (3) 支柱二：RL算法与架构 (RL & Architecture) (1)

🔬 支柱一：机器人控制 (Robot Control) (21 篇)

#题目一句话要点标签🔗⭐
1 World-Env: Leveraging World Model as a Virtual Environment for VLA Post-Training 提出World-Env，利用世界模型作为VLA模型后训练的虚拟环境，解决数据稀缺问题。 manipulation reinforcement learning imitation learning ✅
2 AIRoA MoMa Dataset: A Large-Scale Hierarchical Dataset for Mobile Manipulation AIRoA MoMa：用于移动操作的大规模分层数据集，助力通用机器人 manipulation mobile manipulation generalist agent ✅
3 IA-VLA: Input Augmentation for Vision-Language-Action models in settings with semantically complex tasks 提出IA-VLA框架，利用大型视觉语言模型增强VLA在语义复杂任务中的表现 manipulation vision-language-action VLA
4 Unlocking the Potential of Soft Actor-Critic for Imitation Learning 提出AMP+SAC模仿学习框架，提升四足机器人运动控制的数据效率与泛化性 quadruped PPO SAC
5 Stabilizing Humanoid Robot Trajectory Generation via Physics-Informed Learning and Control-Informed Steering 提出一种融合物理信息学习与控制引导的人形机器人轨迹生成方法，提升轨迹稳定性和物理可行性。 humanoid humanoid robot locomotion
6 JuggleRL: Mastering Ball Juggling with a Quadrotor via Deep Reinforcement Learning JuggleRL：基于深度强化学习的四旋翼飞行器空中杂耍控制 sim-to-real domain randomization reinforcement learning
7 CoTaP: Compliant Task Pipeline and Reinforcement Learning of Its Controller with Compliance Modulation 提出CoTaP框架，通过强化学习和柔顺控制实现人型机器人全身运动控制 humanoid humanoid robot locomotion
8 Learning to Sample: Reinforcement Learning-Guided Sampling for Autonomous Vehicle Motion Planning 提出基于强化学习引导采样的运动规划方法，提升自动驾驶在复杂城市环境中的决策效率。 motion planning reinforcement learning world model ✅
9 CEDex: Cross-Embodiment Dexterous Grasp Generation at Scale from Human-like Contact Representations CEDex：通过类人接触表示大规模生成跨具身灵巧抓取 manipulation cross-embodiment
10 Annotation-Free One-Shot Imitation Learning for Multi-Step Manipulation Tasks 提出一种无标注单样本模仿学习方法，用于多步操作任务 manipulation imitation learning
11 Parallel Heuristic Search as Inference for Actor-Critic Reinforcement Learning Models 提出PACHS算法，利用Actor-Critic模型进行高效并行启发式搜索，提升机器人操作任务性能。 manipulation motion planning reinforcement learning
12 SARM: Stage-Aware Reward Modeling for Long Horizon Robot Manipulation 提出SARM：用于长时程机器人操作的阶段感知奖励建模 manipulation imitation learning behavior cloning
13 MSG: Multi-Stream Generative Policies for Sample-Efficient Robotic Manipulation 提出多流生成策略MSG，提升机器人操作任务的样本效率和泛化能力。 manipulation policy learning flow matching
14 SRMP: Search-Based Robot Motion Planning Library SRMP：面向机器人操作的、基于搜索的运动规划库，提升轨迹一致性和可靠性。 manipulation motion planning
15 U-DiT Policy: U-shaped Diffusion Transformers for Robotic Manipulation 提出U-DiT Policy，结合U-Net和Transformer优势，提升机器人操作任务中Diffusion Policy的性能。 manipulation diffusion policy
16 PoseDiff: A Unified Diffusion Model Bridging Robot Pose Estimation and Video-to-Action Control PoseDiff：统一扩散模型桥接机器人姿态估计与视频到动作控制 manipulation world model embodied AI ✅
17 From Code to Action: Hierarchical Learning of Diffusion-VLM Policies 提出基于扩散-VLM策略的分层模仿学习框架，提升机器人操作的泛化性和数据效率 manipulation imitation learning diffusion policy
18 Crop Spirals: Re-thinking the field layout for future robotic agriculture 提出螺旋形农田布局，优化机器人导航，提升农业自动化效率 MPC model predictive control
19 APREBot: Active Perception System for Reflexive Evasion Robot APREBot：用于反射性避障机器人的主动感知系统 quadruped sim-to-real
20 Game Theory to Study Cooperation in Human-Robot Mixed Groups: Exploring the Potential of the Public Good Game 利用公共物品博弈研究人机混合群体中的合作行为 humanoid humanoid robot
21 Trajectory Prediction via Bayesian Intention Inference under Unknown Goals and Kinematics 提出一种自适应贝叶斯算法，用于未知目标和运动学条件下的实时轨迹预测。 quadruped

🔬 支柱九：具身大模型 (Embodied Foundation Models) (3 篇)

#题目一句话要点标签🔗⭐
22 DynaMIC: Dynamic Multimodal In-Context Learning Enabled Embodied Robot Counterfactual Resistance Ability DynaMIC：动态多模态上下文学习增强具身机器人反事实抵抗能力 multimodal
23 AdaNav: Adaptive Reasoning with Uncertainty for Vision-Language Navigation AdaNav：面向视觉-语言导航，基于不确定性的自适应推理框架 vision-language-navigation VLN ✅
24 LLM-Handover:Exploiting LLMs for Task-Oriented Robot-Human Handovers LLM-Handover：利用大语言模型实现面向任务的人机物体交接 large language model

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

#题目一句话要点标签🔗⭐
25 PhysiAgent: An Embodied Agent Framework in Physical World 提出PhysiAgent框架以解决VLA与VLM协作不足问题 scene understanding vision-language-action VLA
26 Safe Planning in Unknown Environments using Conformalized Semantic Maps 提出基于Conformalized Semantic Maps的安全规划方法，解决未知环境下的语义导航问题 semantic map
27 SONAR: Semantic-Object Navigation with Aggregated Reasoning through a Cross-Modal Inference Paradigm SONAR：融合跨模态推理的语义对象导航方法，提升未知环境适应性 semantic map

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

#题目一句话要点标签🔗⭐
28 Curriculum Imitation Learning of Distributed Multi-Robot Policies 提出基于课程学习的分布式多机器人策略模仿学习方法，提升长期协调能力。 imitation learning curriculum learning egocentric

⬅️ 返回 cs.RO 首页 · 🏠 返回主页