Toward Proximity Surveillance and Data Collection in Industrial IoT: A Multi-Stage Statistical Optimization Design

摘要

In this letter, we consider a practical heterogeneous traffic scenario in industrial Internet of Things (IIoT), where a multi-functionality robot performs proximity surveillance task concurrent with data collection from multiple sensors. While ensuring video resolution constraint for surveillance streams, the robot also adaptively selects subset of sensors for data collection, in order to guarantee the queue stability of the sensors. In particular, the trajectory of the robot subjects to a no-go zone due to the geography restrictions in IIoT, which breaks the convex property of the feasible trajectory space. Also, resource scheduling policy is extended in temporal dimension, where the decisions made in prior slot have an impact to the ones in subsequent slots. To address such a multi-stage statistical optimization design, we propose a novel algorithm to jointly optimize the robot’s trajectory and resource allocation. The proposed algorithm decouples the original multi-stage statistical problem into a series of deterministic problems, while does not violate the causality of the knowledge of the queue status and channel information. Simulation results confirm the superiority of the proposed algorithm in both maintaining data queue stability and surveillance distance performance.