A Cloud-based Quadruped Service Robot with Multi-Scene Adaptability and various forms of Human-Robot Interaction

Abstract

Abstract With the development of CPS technology, more robots are endowed with intelligence to interact and communicate with humans such as service robot AIBO and AMWELL. These intelligent pets show great human-machine interactive ability by voice and face recognition, storytelling. However, they are limited only in family scenes and fail to provide cross-scene services. Therefore, we design a quadruped service robot called Cloud Sphinx - whose motion, voice, and vision system are driven by artificial intelligent models in the cloud. Our work pushes the interaction between CPS and humans. First, an open-source 12-DoF robotic cat is rapidly prototyped with a 3D printing technique, and its versatile gestures and gaits are coordinated by advanced servo control algorithms. Second, by merging motion control, NLP, image processing deep learning and cloud computing, Cloud Sphinx can provide services as emotion classification, face recognition, and garbage classification through human-robot interaction models. Third, an RaaS (Robot as a service) architecture is developed to fulfill the incrementally learning from the human-robot interaction processes. Sufficient simulations and on-board tests verify the human-robot interaction functionality and cross-scene service performances. Applications that we designed show the A.I. performance of service robots can be effectively enhanced by model training during the human-robot interactions and knowledge accumulation in cloud database.