17  Sensor Fusion Architectures

analytics-ml
data
fusion

17.1 Start With the Story

Picture an IoT team using the ideas in Sensor Fusion Architectures during a live operations review. A device has produced messy evidence, an analytic step is about to change an alert or control decision, and someone has to explain why the result should be trusted.

Read this page as that path from sensor evidence to accountable action. Start with what the system observes, keep the model or data treatment visible, and finish with the check that would convince an operator, maintainer, or auditor to act.

17.2 Where Fusion Evidence Combines

A sensor-fusion architecture defines where readings become a shared estimate. In a centralised architecture, sensors send evidence to one fusion process. In a hierarchical architecture, local groups produce intermediate estimates that flow upward. In a distributed architecture, nodes exchange estimates with peers and converge without a single coordinator.

The best architecture depends on latency, bandwidth, failure isolation, time alignment, uncertainty, and review needs. A single central process may give the cleanest global view when the network is reliable. Hierarchical fusion reduces bandwidth and isolates local failures. Distributed fusion is useful when no coordinator can be assumed, but it requires careful handling of duplicated evidence and peer disagreement.

If the architecture cannot explain which sensors contributed, when they were observed, and how uncertainty was combined, the fused estimate is difficult to trust even when the algorithm is sound.

A useful architecture review produces a contract, not only a diagram. It names the fusion owner, input level, clock authority, uncertainty format, duplicate-evidence rule, degraded-mode behavior, and audit record that travels with the output. That contract is what lets a downstream dashboard, controller, or incident review distinguish a fresh global estimate from a local summary or a decision-level vote with hidden disagreement.

The placement decision also controls where bad evidence is stopped. A central service can quarantine raw readings in one place, but it depends on transport and storage capacity. A room gateway can reject stale local sensors before sending a compact state upward, but the building service must still see the rejected-input count and uncertainty. A distributed peer can keep operating during an outage, but it must publish estimate age and conflict state so later convergence is explainable.

Multi-sensor data fusion architecture from sensor layer through preprocessing, feature extraction, fusion algorithms, decision output, feedback loop, and JDL data fusion levels.
A defensible fusion architecture connects placement, abstraction level, feedback, and evidence state so fused decisions can be replayed and audited.

Centralised

All raw or lightly processed evidence reaches one fusion service with a global state view.

Hierarchical

Local fusion nodes summarise nearby sensors, then send estimates upward by room, floor, zone, or asset.

Distributed

Peers share estimates or beliefs and update locally without depending on one central service.

Decision-Level

Independent detectors or classifiers vote, arbitrate, or escalate when raw evidence cannot be shared.

Overview Knowledge Check

17.3 Match Fusion Pattern to Bottleneck

Start with the bottleneck. If the bottleneck is estimator quality and the network is dependable, centralised fusion may be simplest. If the bottleneck is bandwidth or local resilience, use hierarchical fusion and send state estimates upward. If the bottleneck is intermittent connectivity or organisational boundaries, use distributed or decision-level fusion and treat disagreement as explicit evidence.

Worked example: room-to-building hierarchical fusion
rooms: 50
sensors per room: 4
reading interval: every 10 seconds
raw payload: 32 bytes per sensor reading

centralised raw stream:
50 rooms * 4 sensors * 32 bytes / 10 s = 640 bytes/s

room-level fused state:
one 64-byte room state every 30 seconds
50 rooms * 64 bytes / 30 s = 106.7 bytes/s

building-level summary:
one 128-byte building state every 30 seconds
128 bytes / 30 s = 4.3 bytes/s

hierarchical upstream rate:
106.7 + 4.3 = 111.0 bytes/s

reduction:
1 - (111.0 / 640.0) = 0.827, about 83% less upstream traffic

Design consequence:
The bandwidth saving is useful only if each room state also carries time span,
sensor health, uncertainty, and rejected-input evidence.
Pattern
Best Fit
Evidence to Preserve
Watch For
Centralised
Small or well-connected systems where one service can receive enough raw evidence.
All sensor timestamps, raw values, calibration state, covariance, and rejection reason.
Single point of failure, network dependence, and high raw-data transport.
Hierarchical
Buildings, factories, campuses, fleets, or zones with local gateways.
Local state, uncertainty, contributing sensor ids, coverage, and aggregation window.
Losing detail when local summaries omit rejected inputs or uncertainty.
Distributed
Meshes, mobile nodes, or sites where no stable coordinator exists.
Peer version, estimate age, update rule, conflict state, and convergence status.
Double-counting evidence, slow convergence, and inconsistent peer clocks.
Decision-level
Privacy boundaries, vendor systems, or classifiers that can share decisions but not raw data.
Detector id, confidence, label definition, voting rule, abstain reason, and escalation path.
Voting over incompatible labels or hiding uncertainty behind a final decision.

Practitioner Knowledge Check

17.4 Fusion Level Sets Audit Trail

Fusion can happen at several abstraction levels. Data-level fusion combines raw or calibrated measurements. Feature-level fusion combines extracted features such as residuals, FFT bands, occupancy counts, or localisation features. State-level fusion combines estimates and uncertainty, such as Kalman filter tracks. Decision-level fusion combines labels, votes, or incident decisions.

Each level loses and gains information. Raw data supports the richest reprocessing but costs bandwidth and storage. Feature and state estimates reduce transport, but they must preserve uncertainty, time span, calibration, and feature version. Decision-level fusion is compact and sometimes necessary across organisational boundaries, but it is the easiest place to hide incompatible labels or weak confidence.

Worked example: track-to-track covariance check
tracker A position: x = 10.2 m, sigma = 0.4 m
tracker B position: x = 10.4 m, sigma = 0.4 m

If the estimates are independent:
precision for each tracker = 1 / 0.4^2 = 6.25
combined position = (10.2*6.25 + 10.4*6.25) / 12.5 = 10.3 m
combined sigma = sqrt(1 / 12.5) = 0.28 m

If both trackers used the same upstream UWB anchor observation, that 0.28 m
confidence is optimistic because the same evidence was counted twice. The
architecture should mark common sources, correlation assumptions, and any
covariance inflation rule before publishing the fused state.

This is why architecture diagrams should show common sources and summary boundaries, not only arrows. A state-level fusion service needs to know whether two tracks came from independent sensors, from two gateways that shared one anchor, or from two classifiers trained on the same labeled incident set. When independence is unknown, a conservative design keeps the larger uncertainty, flags the common-source path, or routes the fused output for review before it drives actuation.

Data Level

Combines measurements before feature extraction, often after calibration and time alignment.

Feature Level

Combines derived features while preserving feature schema, window, and sensor coverage.

State Level

Combines estimates, tracks, covariance, and confidence from local filters or trackers.

Decision Level

Combines alarms, classes, votes, or abstentions when lower-level evidence is unavailable.

Level
Example
Required Metadata
Failure Mode
Data
Two temperature probes fused after calibration and time alignment.
Units, calibration id, event time, freshness, raw value, and accepted/rejected flag.
Combining measurements that describe different times or units.
Feature
IMU and magnetometer features fused for motion-state classification.
Feature schema, window, normalisation, sensor coverage, and feature version.
Comparing features from different windows or preprocessing versions.
State
Track-to-track fusion from two local Kalman filters.
State vector, covariance, process model, time step, and correlation assumptions.
Overconfident fusion because shared evidence is counted twice.
Decision
Gateway alarm and cloud classifier vote on the same incident.
Label definition, confidence, abstain rule, detector version, and voting policy.
Voting over incompatible labels or suppressing low-confidence disagreement.

Under-the-Hood Knowledge Check

17.5 Summary

Sensor-fusion architecture determines where data, features, states, or decisions are combined. Centralised fusion gives one service a global view but depends on network and coordinator reliability. Hierarchical fusion reduces upstream traffic and isolates local failures when summaries carry time, uncertainty, and health evidence. Distributed fusion avoids a single coordinator but must handle peer disagreement, clock differences, and duplicated evidence. Decision-level fusion is compact, but it requires explicit label definitions, confidence, abstain rules, and voting policy.

Key Takeaway

Choose the fusion architecture by the bottleneck: estimator quality, bandwidth, latency, coordinator reliability, privacy, or auditability. Every fused output must preserve timing, uncertainty, contributing evidence, and rejection metadata.

17.6 See Also

Data Fusion Introduction

Review why multiple sensors are combined and what evidence must travel with the fused value.

Kalman Filters

Use state and covariance when fusing estimates over time.

Complementary IMU Fusion

Place fast and slow sensor corrections in a practical embedded fusion design.

Fusion Best Practices

Keep calibration, time alignment, uncertainty, and fallback behavior reviewable.