Development and Implementation of an AI-Embedded and ROS-Compatible Smart Glove System in Human-Robot Interaction

Robotics technology is being widely used for an array of tasks in today's evolving markets. Human-robot collaboration is inevitable which leads to the need for safe, untroublesome, and easy-to-produce products. A smart glove has capabilities to collect data concerning its wearer's movements by the use of sensors. Motivated by this, in this study, we develop an AI-embedded and ROS-compatible smart glove system to realize real-time human-robot interaction in collaborative tasks. To allow the robot to intelligently learn and predict new human intentions for human-robot interaction, we propose an Extreme Learning Machine (ELM)-based human gesture understanding approach using the data from a set of strip and force sensors embedded in the smart glove and effectively run it through ROS. Three typical baseline gestures are conjured for ELM training purposes and fed into the algorithm with an appended label and corresponding sensor data. The developed system and proposed approach are validated in real-world human-robot collaborative tasks with efficiency and success. This work can also serve as a catalyst for the implementation of many important robot-supported applications such as healthcare and daily assistance for senior groups. Future work of this study is also discussed.