AHSWN Home · Issue Contents · Forthcoming Papers

Information-Driven Sensor Scheduling with Transient QoS for Blueprint-Compliant Digital Twin Maintenance
Zhixian Xu, Guangjian Mo, Xiaowei Zhu, Weijian Guan, Jiaji Li, Zhuolang Ao, Anji Chen, Si Shi and Zhen Tian

Wireless sensor networks (WSNs) that support blueprint-driven digital twin (DT) maintenance in power substations often waste energy on uniform telemetry, yet still fail to deliver the packets that are most decisive for compliance diagnosis. We propose Evidence-Aware Telemetry (EAT), a cross-layer scheduling framework that allocates sensing and communication budgets to maximize expected information gain (EIG) on rule-specific hypotheses under a practical expected transmission count (ETX) and energy cost model. EAT: (i) binds blueprint rules to rule-relevant candidate sensors via a blueprint knowledge graph, achieving 74% reduction in search space; (ii) performs budgeted EIG selection using a greedy strategy with lazy gain updates; and (iii) temporarily lifts quality-of-service (QoS) for selected evidence flows by adjusting per-hop retries, queue precedence, and low-power and lossy network (LLN) resources, with hysteresis and budget guards to protect network stability. In Python-based simulations on synthetic substation topologies with 50 nodes over 50 episodes, EAT achieves 91% higher information-per-joule and 40% lower energy consumption than uniform polling, while reducing P95 and P99 tail delivery delays by 37% and 36% respectively. Compared to information-only selection (MI-only), EAT reduces energy by 36% while maintaining comparable diagnostic quality through cost-aware tradeoffs. These results demonstrate that mapping evidential value to transient network service can improve diagnostic efficiency and tail latency without increasing overall traffic load.

Keywords: ad hoc & sensor networks, low-power and lossy networks (LLNs), QoS, routing and scheduling, expected information gain, blueprint knowledge graph, digital twin, power substations