The Research Of Exacting And Fusion Feature Information From Perception System To Achieve The Whole Robot Soft-Grasping Process

The soft-grasping of robot is to fully sense the external experiments during the process of grasping, which would provide multi-dimension and whole process information for interaction while it grasp the objects. There will be great significance of the soft-grasping in robot area, which could be helpful for the service-type robots in the future to accomplish a secure interaction with human.This study is inspired from the realization of the whole-process monitor for mutual information, and major in designing multi-homogenesis, heterogeneity sensor in whole-process perception system for realizing the soft-grasping of robots. This research aims at the entire process of grasping which had not been considered in the process of robot grasping, and the difficulties in slip sensor. This thesis is based on the theory of material science, sense technology and artificial intelligence, we aim at realizing soft-grasping, and build the full-directional slip-tactile sensing unit with microstructure. The amount and direction of the force has been quantized, and the monitor model based on theory of decoupling the information of three dimensional fore and Mindlin's contact theoretics has also been established. Furthermore, we build the new multi-information compound sense array structure based on the previous work.In order to deal with the key scientific problems as the relevance of information from different sensors, and the extraction of the characteristics in our robot skin system. We research the intelligentialize extraction and combination route of composite information based on strategy of multi-sensor information composition. The fusion algorithm, SVM combined with PCA, optimize coupling redundancy and complementary information from multi-sensor, then assess states of the grasped-object and surrounding environment, achieve coordinated expression of multi-dimensional information and increase the intelligence of robot finally.The experiment results of manipulator grasping indicate the effective reliability of whole-process sense skin system and the information fusion algorithm, which could realize the highly precise judgment and prediction among proximity, quasi-contact, contact-grasp and pre-slide states, and would improve the level of interaction application of robots.