Active Vision for Robot Manipulation

Follow this topic Ask about this topic

Active vision for robot manipulation studies how a robot can change its viewpoint or interact with objects to see what it needs for grasping and task execution, especially under occlusion or uncertainty. Methods include next-best-view planning, gaze control, and information-seeking actions, learned through reinforcement learning, imitation, or vision-language-action policies. Researchers evaluate these approaches in cluttered scenes, occluded-object tasks, and settings where viewpoints or environments change, often using simulation and real-robot trials. A central tension is that active sensing can improve robustness, but its benefits vary by task and must be balanced against extra movements, latency, data needs, and hardware complexity.

Latest papers

37 papers16 papers in the last four weeks, up 167% on the four weeks before. 0.1% of all new papers.

Active Stereo-Camera Outperforms Multi-Sensor Setup in ACT Imitation Learning for Humanoid Manipulation
Mar 30, 2026Robin Kühn, Moritz Schappler, Thomas Seel +1Robot Imitation Learning Robot Learning
The authors are with the Institute of Mechatronic Systems, Leibniz University Hannover, 30167 Hannover, Germany.
PiCo: Active Manifold Canonicalization for Robust Robotic Visual Anomaly Detection
Mar 24, 2026Teng Yan, Binkai Liu, Shuai Liu +2Robotic Anomaly Detection Robotic Perception
The Hong Kong University of Science and Technology (Guangzhou)
Cortical Policy: A Dual-Stream View Transformer for Robotic Manipulation
Mar 22, 2026Xuening Zhang, Qi Lv, Xiang Deng +3Vision-Based Robot Control Robotic Manipulation
Harbin Institute of Technology (Shenzhen), Shenzhen, Guangdong 518055, China · Shandong Jianzhu University, Jinan, Shandong 250101, China
SaPaVe: Towards Active Perception and Manipulation in Vision-Language-Action Models for Robotics
Mar 12, 2026Mengzhen Liu, Enshen Zhou, Cheng Chi +6Active Vision for Robot Manipulation Robotic Manipulation
State Key Laboratory of Multimedia Information Processing, School of Computer Science, Peking University · Beijing Academy of Artificial Intelligence · School of Software, Beihang University
EgoAVFlow: Robot Policy Learning with Active Vision from Human Egocentric Videos via 3D Flow
Feb 25, 2026Daesol Cho, Youngseok Jang, Danfei Xu +1Scene Flow Estimation Active Vision for Robot Manipulation
School of Interactive Computing, Georgia Institute of Technology, Georgia, USA · InnoCORE AI-Transformed Aerospace Research Center, KAIST, Daejeon, Republic of Korea
Selective Perception for Robot: Task-Aware Attention in Multimodal VLA
Feb 17, 2026Young-Chae Son, Jung-Woo Lee, Yoon-Ji Choi +2Efficient VLA Models Vision-Language-Action Models
Dongguk University
Affordance-Aware Interactive Decision-Making and Execution for Ambiguous Instructions
Feb 5, 2026Hengxuan Xu, Fengbo Lan, Zhixin Zhao +5Robotics Robot Task Planning
Department of Automation, Tsinghua University, Beijing, China · Shanghai Artificial Intelligence Laboratory, Shanghai, China · Institute for Interdisciplinary Information Sciences, Tsinghua University, Beijing, China +1

Mar 30, 2026J/K move · Enter open · S save

Active Stereo-Camera Outperforms Multi-Sensor Setup in ACT Imitation Learning for Humanoid Manipulation

Robin Kühn, Moritz Schappler, Thomas Seel, Dennis Bank

The authors are with the Institute of Mechatronic Systems, Leibniz University Hannover, 30167 Hannover, Germany.

Robot Imitation Learning Robot Learning Active Vision for Robot Manipulation Multimodal Sensor Fusion Data-Efficient Robot Learning

Open paper AskPDF arXiv

The complexity of teaching humanoid robots new tasks is one of the major reasons hindering their widespread adoption in the industry. While Imitation Learning (IL), particularly Action Chunking with Transformers (ACT), enables rapid task acquisition, there is no consensus yet on the optimal sensory hardware required for manipulation tasks. This paper benchmarks 14 sensor combinations on the Unitree G1 humanoid robot equipped with three-finger hands for two manipulation tasks. We explicitly evaluate the integration of tactile and proprioceptive modalities alongside active vision. Our analysis demonstrates that strategic sensor selection can outperform complex configurations in data-limited regimes while reducing computational overhead. We develop an open-source Unified Ablation Framework that utilizes sensor masking on a comprehensive master dataset. Results indicate that additional modalities often degrade performance for IL with limited data. A minimal active stereo-camera setup outperformed complex multi-sensor configurations, achieving 87.5% success in a spatial generalization task and 94.4% in a structured manipulation task. Conversely, adding pressure sensors to this setup reduced success to 67.3% in the latter task due to a low signal-to-noise ratio. We conclude that in data-limited regimes, active vision offers a superior trade-off between robustness and complexity. While tactile modalities may require larger datasets to be effective, our findings validate that strategic sensor selection is critical for designing an efficient learning process.