Enabling edge intelligence in remote robot control: Architecture, prediction mechanism and offloading

Abstract

With recent rise in the integration of robotics and the 5G Wireless Communication Networks (WCNs), Tactile Internet (TI) has gained a significant momentum as a real-time application. This demands a complete immersion, with optimal Quality of Service (QoS) and Quality of Experience (QoE). This paper proposes a haptics communication network with a Haptic Glove and a Robotic Hand communicating over a 5G network, with the multi-access edge computing (MEC) carrying out the prediction and offloading tasks. In a scenario to control remote robot, as per the movement from the HG, the robot will perform a task and adjusts the movement according to the haptic feedback. The prediction mechanism to estimate the reverse feedback and forward control signals, based on Long Short Term Memory (LSTM) model. As a use-case the reverse feedback and the missing values are estimated for transition from an arbitrary position to the goal pose, meeting the ultra-low latency requirement for remote robot control. The proposed system leveraging edge enhances the overall QoE by predicting the missing values which are lost in the network (in forward direction) while returning the immediate feedback to the controller using prediction mechanism (in backward direction).