Research On Exoskeleton Robot Cloud-Brain Architecture And Learning Algorithm

The intelligent integration of exoskeleton robot and wearer,especially combined with cloud platform,is the future development direction of exoskeleton robot.In this paper,the architecture of robot cloud brain and the learning algorithm of cloud machine are reviewed.The cloud brain architecture and learning algorithm for exoskeleton robots are proposed,and the actual running state of the exoskeleton robot is studied and applied to the man-machine security state detection and analysis of exoskeleton robots.The main innovative work is as follows.In the medical application scene of rehabilitative exoskeleton robot,aimimg at the problem of accessing the data of doctor-patient and the real-time sensor data of exoskeleton robot into the cloud,this paper proposed three stratus cloud brain robot architecture that is similar to lambda architecture from the dimension of the transmission speed of data information based on ECS service of Ali.It mainly includes: batch processing layer connecting doctor-patient data,fast processing layer connecting exoskeleton robot sensor data and forming virtual clones at the cloud platform,as well as a service layer for monitoring,verification and intelligent expansion.Using database technology in batch processing layer to interconnect PC and APP data of users.In the fast processing layer,we use the internet of things technology and real-time message middleware technology to restore the man-machine state of every user in rehabilitative exoskeleton robot training,and forming a real-time virtual clone in the cloud platform.Such an architecture not only solves the interconnection of user data in the scene,but also solves the real-time cloning of robot data,which provides technical guarantee for extending the intelligence and sociality of the exoskeleton robot in the rehabilitation medical field.Aiming at the problem of human-machine safety state detection of exoskeleton robots in practical application,this paper uses the different dimension sensor data of the cloud virtual clone to cloud machine learning,and sets up a personalized machine learning module from the wearer's angle based on the above proposed exoskeleton robot cloud brain architecture.The module mainly includes the man-machine state anomaly detection model based on the wearer's left and right sole pressure data and the Isolation Forest algorithm,and the man-machine state prediction model based on the left and right shoe sole pressure sensing data,the attitude instrument data and the Auto Encoder and Seq2 seq network.Based on left and right sole pressure sensing data,attitude data,left and right hip joint angle data and random forest algorithm,a human-machine state classification model is established.Aiming at the problem of human-machine safety state analysis of exoskeleton robots in practical application,this paper designs and implements a human-machine safety state analysis unit from the perspective of doctors or rehabilitation practitioners.The unit mainly includes: the state interpretability model based on human-machine state classification model and the Lasso algorithm,as well as the associated rule analysis model based on the Apriori algorithm.Therefore,the unit has explored the problem of humanmachine safety state analysis for exoskeleton robot,and has practical production and application value.