Hierarchical Semantic Knowledge-Based Object Search Method for Household Robots

Object search is a basic and necessary task for household robots to perform home services. However, due to the complexity and dynamics in home environment, some target objects can hardly be observed in the first place, reducing the efficiency of object search. For human beings, target objects can always be associated with relevant but obvious objects, or room categories, and these kinds of information may provide a guidance on helping robots locate target objects more quickly and accurately. In order to simulate this process, this article proposes a hierarchical semantic knowledge based object search method. Firstly, semantic knowledge is constructed by modelling the (room category, environmental object, dynamic object) relationship in natural languages related to home services. Secondly, to deal with information loss caused by the ambiguity in natural language, a priori knowledge identifying the relationship between room categories and environmental objects is constructed for further completing the semantic knowledge. After that, the rules for knowledge deployment and updating are presented to increase the practical value of the semantic knowledge. Finally, a heuristic object search strategy based on the semantic knowledge is proposed to guide the robot on locating the target object efficiently and quickly. In order to make the strategy stable and adaptable, the room layout and the distance between the robot and the candidate room are also taken into consideration. Our method is tested in both the simulated and real environments, and the results demonstrated that the proposed method can effectively assist the robots on locating the target object with less time cost and shorter path length.