HUMAN ROBOT INTERACTION USING KNOWLEDGE BASE APPROACH

This research investigated Human Robot Interaction modalities. The performance of a robotic prosthetic hand (RPH) was used as a test bed robot. A user wearable glove was fitted with sensors to provide tele-control of the RPH. An open hardware control and easily expandable research platform based on LabVIEW software and myRIO control hardware was developed. LabVIEW graphical programming platform provides the tools for the development to customized interfaces for visualization purposes which is desired in research. The research platform was used to calibrate and control the operation of the artificial hand using various modalities such as open loop, glove master-slave setup and knowledge base interaction. Mapping algorithms between the motion of the master glove and slave RPH were developed. The knowledge based modality was based on artificial neural networks (ANN), where supervised learning identified appropriate grasping patterns for a set of objects based on a training data set. The training data set was developed using manual and glove control of the RPH and consists of the object geometric features and object location relative to the RPH. The training data set is then processed using the LabVIEW ANN toolkit to identify in real-time, the grasping patterns for other similar objects that include the desired motion for each RPH finger. The developed research platform and tools have been demonstrated through manual, glove and ANN control of the RPH and display of system information on the LabVIEW GUI in real-time.