Abstract: As a barrier between the external environment and its internal components,human skin plays a vital role in biological system.With the development of intelligent agriculture,software robots will have a wide application prospect in fruit and vegetable picking,grafting,weeding,pest control and so on.Researchers have developed many interesting forms of activation that more closely mimic functions and abilities in nature.To develop biologically inspired tactile sensing software robots that can safely interact and explore with their environment.The current work often focuses on the design and manufacture of software robots,and explores how machine learning can improve the perception ability of software robots.The safety and particularity of soft robot have attracted the attention of many scientists.Using deformable materials,soft robot can shape the powerful characteristics of the evolutionary behavior of various animals and plants,and can control the operation through physical contact.However,as the development of software and stretchable materials into robot systems,sensor integration faces numerous challenges,including stretchable multimodal sensors,the embedding of high-resolution large array sensors,as well as the increase of data volume and sensor fusion.This paper discusses the emerging intersection of e-skin and machine learning,focusing on how experts in the field of robotics develop their interdisciplinary,build autonomous and deployable software robots,and combine information tactile and perceptual ability to deal with the challenges of complex conditions in the real environment.The skin layer is embedded with a dense network of mechanical,chemical,vibration,temperature and pain receptors,so that the skin can have tactile and temperature perception.Solving many practical problems in many special scenes is of great significance for the development of the field of special robots.