Structural control improves sensor data recovery during long outages
CTRL-FM: Inference-Time Structural Control for Flow-Based Virtual Sensing under Long-Horizon Node Outages
Machine Learning
Summary
Sensors sometimes go offline for a long time, making it hard to guess what their readings should have been. Existing methods assume a fixed pattern in how sensor data behaves, which can cause mistakes when conditions change suddenly. The authors developed CTRL-FM, a method that adapts its approach depending on which sensors are offline, using information from working sensors to decide how best to fill in missing data. This flexible approach keeps available sensor readings unchanged and picks the best way to guess missing ones. Tests show this method works better than others, even when many sensors fail at the same time.
What this means in practice
- •For infrastructure monitoring teams: Recover accurate sensor data during extended sensor downtime in traffic and air quality monitoring systems.
- •For industrial maintenance teams: Maintain reliable virtual sensing in complex industrial sensor networks experiencing multiple long-term sensor failures.
Authors
Boshra Khalili, Andrew Smyth
Abstract
Long-horizon node outages are harder than conventional missing-value imputation because complete sensor trajectories remain unobserved for extended periods, while the usefulness of spatial and temporal dependencies varies across outage locations and operating regimes. Existing graph, diffusion, and flow-based methods typically impose a fixed structural bias, which can oversmooth abrupt changes or propagate errors from unreliable neighbors. We introduce CTRL-FM, an observation-preserving flow-matching framework that treats structural guidance as an outage-conditioned inference-time decision. A missing-only flow-matching backbone learns a conditional velocity field while keeping observed measurements exactly fixed. For each realized outage, CTRL-FM constructs self-supervised proxy outages from surviving measurements and uses their known values to select among uncontrolled, graph-dominant, temporal-dominant, and balanced control policies. The selected policy is locally modulated by graph and temporal reliability estimated exclusively from surviving observations, while boundary-anchored trends provide structural guidance. Control is applied only to missing entries and is rejected when proxy evidence does not sufficiently improve over uncontrolled sampling. Across PeMS08, METR-LA, Air-36, and AQI, CTRL-FM achieves the strongest overall performance under 30% and 60% node outages. Stress tests with up to 75% sensor loss further show graceful degradation and consistent gains over training-time and fixed structural control. These results demonstrate that validating, selecting, and locally adapting structural guidance for each realized outage is more effective than imposing a single global structural prior.