Visual Detection-Based Blockage Prediction for Beyond 5G Wireless Systems

This paper presents a novel visual detection-based approach for predicting imminent blockages of wireless radio links in industrial ultra reliable low-latency communication (URLLC) systems. The solution is based on detecting the bounding box of the UE-object and deducing from the temporal behavior of the bounding box whether the object is about to be blocked by an obstacle. It is shown that such detection approach allows one to predict the blockage as the bounding box starts to become smaller much before the full blockage occurs. Simulation results indicate that the proposed visual blockage prediction scheme can significantly reduce the latency of URLLC networks by preventing packet errors due to blockage and the associated reduction in link quality. In macro cell scenarios, the latency reduction is as much as 5 ms. In addition, we also show that the detection range of the proposed solution is sufficient for most industrial applications; with a 4K camera, reasonably sized objects can be detected from as far as 130 m away.