Hand-gesture-control-based Navigation Using Wearable Armband with Surface Electromyography and Inertial Measurement Unit Sensor Data for Autonomous Guided Vehicles with Robot Operation System-based Simultaneous Localization and Mapping Navigation in Smart Manufacturing

摘要

Autonomous guided vehicles (AGVs) with a robot operation system (ROS)-based platform have been widely used in automation-assisted manufacturing. AGV robots in smart manufacturing are mainly used to handle materials. Such AGV robots with an ROS are generally capable of simultaneous localization and mapping (SLAM) and can therefore perform autonomous navigation (the well-known ROS-based SLAM navigation). From the viewpoint of "smart" manufacturing, which is expected to have richer artificial intelligence and human-robot interaction (HRI), an AGV robot with only ROS-based SLAM autonomous navigation is extremely restricted in functions and human-robot interactions. In this work, to increase HRIs and the flexibility of usage of AGVs with only ROS-based SLAM autonomous navigation, a hand-gesture-control-based navigation approach using a wearable armband with sensor data from both surface electromyography (SEMG) and an inertial measurement unit (IMU) is presented. The developed hand-gesture-control-based navigation with artificial neural network (ANN) hand gesture command recognition can be incorporated into a typical AGV operation with SLAM autonomous navigation. The hand-gesture-control-based navigation for AGVs proposed in this study mainly consists of two calculation phases: the detection of the significant hand gesture for the corresponding gesture operation command by the analysis of eight-axis SEMG data, and the recognition of hand gesture commands from the operator using an ANN with nine-axis IMU data. To appropriately combine the detection and recognition of hand gestures, two strategies were developed for the navigation control of an AGV in a certain continuous time period: ANN recognition by the IMU in a fixed decision window with an SEMG system wake-up, and ANN recognition by the IMU in a variable decision window with both a system wake-up and end. A series of online test experiments on AGV navigation by hand gesture control demonstrated that the presented approach has a competitive performance, particularly for short-path navigation.