Enhancing Object Search in Indoor Spaces via Personalized Object-factored Ontologies

2025-06-17

AI Summary

To address the challenge of long-term autonomous indoor multi-object search for service robots, this paper proposes a semantic-aware approach integrating personalized object factorized ontologies with dynamic belief updating. Methodologically, we introduce the first learnable and updatable personalized ontology generation framework, explicitly modeling user preferences and environmental evolution as hierarchical semantic structures; this is coupled with multi-layer semantic mapping and dynamic Bayesian inference for online belief optimization. Key contributions include: (1) a personalized schema learning mechanism enabling adaptive ontology construction from user feedback and environmental changes; and (2) an adaptive ontology reasoning strategy supporting real-time belief refinement. Experiments in realistic indoor environments demonstrate significant improvements—+23.6% in search success rate and −31.4% reduction in average search time—while maintaining plug-and-play compatibility to enhance multiple state-of-the-art baseline methods.

Technology Category

• Search and Optimization: Learning to Search
• Intelligent Robots: Learning & Optimization for ROB
• Constraint Satisfaction and Optimization: Search

Application Category

• Semantics and Knowledge: Applications of semantic technologies for improving search, browsing, recommendation, personalization
• User Modeling, Personalization and Recommendation: ML for personalized search and recommendations
• Search and Retrieval-Augmented AI: Personalized, context-aware and across-device search

Abstract

Personalization is critical for the advancement of service robots. Robots need to develop tailored understandings of the environments they are put in. Moreover, they need to be aware of changes in the environment to facilitate long-term deployment. Long-term understanding as well as personalization is necessary to execute complex tasks like prepare dinner table or tidy my room. A precursor to such tasks is that of Object Search. Consequently, this paper focuses on locating and searching multiple objects in indoor environments. In this paper, we propose two crucial novelties. Firstly, we propose a novel framework that can enable robots to deduce Personalized Ontologies of indoor environments. Our framework consists of a personalization schema that enables the robot to tune its understanding of ontologies. Secondly, we propose an Adaptive Inferencing strategy. We integrate Dynamic Belief Updates into our approach which improves performance in multi-object search tasks. The cumulative effect of personalization and adaptive inferencing is an improved capability in long-term object search. This framework is implemented on top of a multi-layered semantic map. We conduct experiments in real environments and compare our results against various state-of-the-art (SOTA) methods to demonstrate the effectiveness of our approach. Additionally, we show that personalization can act as a catalyst to enhance the performance of SOTAs. Video Link: https://bit.ly/3WHk9i9

Problem

• Enhancing object search in indoor spaces via personalized ontologies
• Developing adaptive inferencing for multi-object search tasks
• Improving long-term object search with dynamic belief updates

Innovation

• Personalized Ontologies for indoor environments
• Adaptive Inferencing with Dynamic Belief Updates
• Multi-layered semantic map integration