In context
As home service robots move from factories into daily life, achieving long-term autonomy remains a key challenge. Robots must continuously understand and adapt to dynamic household environments, yet existing semantic mapping and 3D reconstruction methods often struggle with real-time updates and comprehensive object-level information. This research addresses that gap by introducing a novel information representation and maintenance strategy.
What was reported
Researchers at Shandong University developed an online method for constructing and maintaining 'information entities'—a unified representation that integrates physical and semantic data for each object instance. The system uses EMSANet for instance-level semantic segmentation and RGB-D data to generate 3D point clouds, capturing position, shape, volume, and color. A confidence-based update strategy enables automatic creation, modification, and deletion of entities based on spatial and semantic matching, improving accuracy and reducing redundant mapping.
To handle dynamic changes, the method combines passive maintenance during task execution with active exploration during idle times. An improved RRT-based algorithm identifies semantic frontiers and uses information gain to select optimal observation points, ensuring comprehensive environment coverage with lower computational cost. Experiments in simulated and real environments demonstrated accurate and rich environment description, as well as efficient and complete information maintenance, meeting long-term autonomy requirements.
Why it mattered
This work advances service robot autonomy by shifting from whole-scene mapping to object-centric information management, enabling more robust and efficient updates in changing home environments. The approach could improve the reliability of robots performing tasks like object retrieval and navigation, supporting broader adoption in domestic and assistive applications.
"The experimental results demonstrate that the proposed method can accurately and comprehensively describe environmental information and achieve efficient and complete information maintenance, meeting the requirements of long-term autonomy."
Source: 《机器人》期刊 (robot.sia.cn) · Published 2025-03-15 · “基于在线信息体构建与维护的长期自治家庭服务机器人环境理解方法”