Sensor Fusion
Sensor Fusion (State-of-the-Art Review) · Also known as: Multisensor Data Fusion, Multi-Sensor Integration, Information Fusion, Sensör Füzyonu
Sensor fusion is a computational process that combines data from multiple heterogeneous sensors to produce an estimate of the environment that is more accurate, complete, and reliable than any single source alone. Systematized as a formal field by Khaleghi, Khamis, Karray, and Razavi in their 2013 state-of-the-art review in Information Fusion, the discipline addresses imperfections such as noise, incompleteness, temporal misalignment, and conflicting readings that arise whenever multiple sensing modalities operate in parallel.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Apply sensor fusion whenever a single sensor is insufficient in accuracy, coverage, or robustness for your task, and multiple complementary sensors are available or can be deployed. Classic use cases include autonomous navigation, structural health monitoring, medical imaging, and environmental surveillance. Key assumptions are that sensors share a common observable state and that their noise characteristics are at least approximately known. The method is inappropriate when sensors measure fundamentally incompatible quantities or when fusion overhead exceeds the accuracy benefit in latency-critical systems.
Strengths & limitations
- Improves estimation accuracy beyond any single sensor's limit by exploiting complementary and redundant modalities.
- Increases system robustness: the fused estimate degrades gracefully when one sensor fails or is temporarily occluded.
- Formally quantifies residual uncertainty via a joint covariance, enabling principled downstream decision-making.
- Scalable architecture — centralized, distributed, or hierarchical topologies suit different bandwidth and latency budgets.
- Requires accurate, often hard-to-obtain sensor noise models and calibration parameters; miscalibration propagates into the fused result.
- Conflicting or highly correlated sensors can degrade rather than improve estimates if not handled with appropriate association methods.
- Computational and communication costs grow with the number of sensors, posing challenges in real-time embedded systems.
- Data association across sensors — matching which observation corresponds to which target — is NP-hard in the general multi-target case.
Frequently asked
What is the difference between sensor fusion and data fusion?
Sensor fusion specifically refers to combining outputs from physical sensing devices (cameras, radar, IMUs). Data fusion is the broader umbrella term encompassing any integration of heterogeneous information sources, including databases, human reports, and simulated data. Khaleghi et al. (2013) use 'multisensor data fusion' to emphasize both the physical sensor origin and the data-level processing involved.
At which level should fusion be performed — signal, feature, or decision?
Fusion level is a design trade-off. Signal-level fusion preserves maximum information but demands tight synchronization and high bandwidth. Feature-level fusion balances richness and efficiency, working on extracted descriptors. Decision-level fusion is most tolerant of sensor heterogeneity and partial failures but discards raw data early. The optimal level depends on available bandwidth, latency requirements, and how much raw-data processing is feasible on each sensor node.
How does sensor fusion handle conflicting sensor readings?
Conflicting readings arise from sensor faults, adversarial conditions, or genuine environmental ambiguity. Probabilistic approaches down-weight low-likelihood observations via the measurement likelihood p(z_i | x). Dempster-Shafer theory can explicitly represent and combine belief masses under conflict. Robust fusion methods such as M-estimation or consensus algorithms actively detect and exclude outlier sensors before combining the remaining consistent evidence.
Sources
- Khaleghi, B., Khamis, A., Karray, F. O., & Razavi, S. N. (2013). Multisensor data fusion: A review of the state-of-the-art. Information Fusion, 14(1), 28–44. DOI: 10.1016/j.inffus.2011.08.001 ↗
How to cite this page
ScholarGate. (2026, June 2). Sensor Fusion (State-of-the-Art Review). ScholarGate. https://scholargate.app/en/data-fusion/sensor-fusion
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Data FusionData Fusion↔ compare
- Ensemble Kalman FilterData Fusion↔ compare
- Kalman Filter (Finance)Finance↔ compare