This paper studies secure and resilient scheduling for cloud–edge flexible job shops, where equipment failures require rapid recovery but machine capability data must remain private. We propose an edge-native framework that combines hierarchical coordination, topology-bounded recovery, authenticated failover with fresh re-authentication, and 2SS-TopK for private backup-machine selection. Exact machine state and candidate ranking remain at the edge, while the cloud receives cluster-level surrogates and restricted recovery outcomes. The evaluation combines controlled placement and recovery-scope studies with a five-policy online FJSP comparison, one-factor component ablations, authentication-response injection, architecture sensitivity, and a selective-disclosure utility test on Brandimarte instances. The completed five-policy stress test produced metric-dependent results: MWKR has the lowest displayed mean makespan on all three high-failure instances, whereas Ours/2S-ECT has the lowest displayed mean tardiness and the lowest displayed mean propagated delay per fault on all three high-failure instances. These are descriptive rankings of displayed means; mixed makespan, recovery latency, escalation, and downstream outcomes do not support universal solver dominance or global optimality. The one-factor matrix shows that cloud placement adds about 105ms per fault to the same secure recovery path, while removing the within-zone candidate bound reduces escalation by 4.0–11.9 percentage points at a 1.6–4.7ms per-fault latency cost. A paired 16,000-task control rejects all seven injected invalid-response classes with fresh re-authentication, whereas the no-reauthentication arm transfers all of them. Selective disclosure creates a measurable privacy–performance trade-off. These are simulator and functional-gate results under the stated threat model, not solver-superiority, production-hardware, or complete-security claims. Accordingly, the contribution is framed as a systems co-design rather than a new global optimizer, cryptographic primitive, complete zero-trust architecture, or production-ready implementation.
Visual bridge inspection is a knowledge-intensive task in which inspectors coordinate visual search, spatial navigation, structural reasoning, and defect identification and documentation. It is a central maintenance task for bridges and a key basis for safety assessments, yet its results are susceptible to individual subjectivity. While eye-tracking-based behavioral studies quantify underlying processes, existing research often imposes restrictive simplifications to reduce environmental complexity, thereby reducing the realism and representativeness of the inspection task. This study proposes an automated data analytics framework for converting multimodal inspection data into an inspection mode time series. Screen-based 3D gaze, head-movement, drone navigation, and scene geometry data are segmented into temporal windows and classified into three functional modes: global scanning, local inspection, and navigation. The resulting temporal representation enables the extraction of interpretable behavioral descriptors, including transition probabilities, dwell times, transition entropy, fixation measures, and spatial revisit metrics. A proof-of-concept study using a virtual bridge inspection platform demonstrates the technical feasibility of converting these multimodal data into interpretable representations of participant behavior. The framework serves as a research tool for studying visual attention, navigation, and behavioral strategies during simulated bridge inspections.
