Spatio-temporal and semantic enhanced context-aware navigation for indoor robots

Indoor robot navigation demands that robots not only map their environment, but also understand the context within it. Traditional geometric, topological, and semantic models offer specialized capabilities, but often lack the integration of contextual knowledge critical for navigating unfamiliar spaces. To address this limitation, we introduce the Context-Aware Rapidly-exploring Random Tree (CA-RRT) algorithm, which integrates semantic awareness into the exploration process. The CA-RRT algorithm improves navigational efficiency by combining metric, topological, and semantic models to construct a complete representation of contextual knowledge. This hybrid approach, augmented by spatio-temporal data, allows robots to interpret their environment and infer the relations more effectively. The proposed CA-RRT algorithm, implemented within the robot operating system, leverages ontology design patterns, such as n-ary semantic relations, to encode context-aware constraints directly into the exploration process. Experimental results indicate that CA-RRT substantially improves search efficiency and semantic accuracy compared to the baseline A ROS Multi Ontology References (ARMOR) framework and traditional Rapidly exploring Random Tree (RRT) algorithm, especially when dealing with complex environments where semantic constraints are vital. This novel approach holds promise for the advancement of the capabilities of autonomous robots, enabling them to navigate more intelligently and efficiently in dynamically changing indoor spaces. The proposed approach achieved minimum distance, execution time, and higher performance in terms of accuracy, precision, and coverage area compared to the baseline ARMOR framework and other exploration strategies used in the existing work.