Data-driven Context-Aware Traffic Prediction and Modeling for Tactile Internet

We propose a data-driven, context-aware approach for traffic modeling and a traffic predictor to support resource reservation in the tactile internet for bursty traffic patterns caused by Deadband Perceptual Data Reduction (DPDR). The traffic state is defined as the number of arriving packets in the next time window, which is modeled and predicted based on historical context information, which refers to user motion commands or haptic feedback. Since traffic prediction in this context is traditionally challenging and often limited to binary state consideration, the proposed method provides a more general framework for multi-state applications. By using a manually collected dataset from a self-developed visuo-haptic experiment setup, the proposed method provides a superior performance in terms of traffic modeling and prediction, compared to its own benchmark.