Semantic Map Construction Technology And System For Home Service Robot

In order to solve the problem of adaptive scene recognition in the home environment of service robot,an adaptive scene recognition algorithm based on Places205-Alex Net was proposed,which used recursive Bayesian filtering algorithm to realize adaptive annotation of scene semantics.Aiming at the detection problem of robot picking up objects in the scene,a target entity detection algorithm based on YOLO V3 was proposed to obtain semantic information and location information of objects in the environment,which improved the speed and accuracy of entity detection in the process of semantic map construction.Aiming at the relationship between the entities and the scene in the map,the semantic knowledge base of the entities with the purpose of semantic navigation was established,and the semantic navigation reasoning mechanism reflecting the mapping relationship between the semantic map and the scene entity information was established.Finally,the adaptive semantic map construction of the robot was realized.At the same time,the service robot platform was built,and the proposed algorithm was integrated tested and verified by experiments.The main work of this paper includes the following four aspects:1)An adaptive recognition algorithm for home environment based on semantic map construction is proposedFirstly,according to the characteristics of different scenes in home indoor environment,a scene recognition algorithm based on Places205-Alex Net was proposed.Then,aiming at the problem of the robot's perception of important objects in the home,a target entity detection algorithm based on YOLO V3 network was proposed.The integration of the service robot's adaptive scene recognition function can realize the detection and location calculation of objects in the home environment.The experimental results show that the use of deep learning technology combined with laser SLAM can effectively realize the adaptive semantic labeling of the robot.2)This paper proposes a semantic navigation entity search algorithm for entity searchFor the purpose of the service robot to complete the entity search task in the home indoor environment,firstly,this paper added attachment attributes,search priority and other related attributes to the object and map,and then combined with the geometric information of the laser grid map to form the semantic knowledge base of the service robot in the indoor home environment.Then,based on the semantic knowledge base of the robot,the inference mechanism of the robot when it performs the entity search task is designed,and a semantic navigation entity search scheme based on semantic map and semantic knowledge base is formed.3)A semantic map system for home service robot is designed and implementedAiming at the proposed semantic map construction scheme,a home service robot platform was built,and the semantic map construction system test on the service robot platform was realized by integrating the proposed algorithm.Through experiments,it is verified that the semantic map scheme using deep learning technology combined with laser SLAM can effectively realize the robot's semantic adaptive labeling and can effectively carry out the entity target search task.