**6. Conclusions**

Choosing a topology is one of the main challenges in information fusion system design. Associativity, consistency, and redundancy play key roles in the performance of a topology. In this article, we detailed and discussed a data-driven design approach resulting in a two-layer topology inspired by MCS fusion. Due to the associativity of fusion rules in the first layer nodes, the topology can be extended to multiple layers without affecting the fusion results.

The basic design approach relies on the consistency of information items to find MCS nodes. The resulting consistency-based topology was susceptible to unexpected behaviour from information sources caused by unrepresentative training data or defective sources. We proposed adaptations to the basic design comprising the inclusion of a redundancy metric, the automated discounting of defective sources, and the application of outlier robust estimation fusion.

In the evaluation, we demonstrated that the redundancy-enhanced design resulted in more robust topologies in the case of epistemic uncertainty. Furthermore, evaluation showed that discounting defective sources and estimation fusion reduced the effects of defective sources. Estimation fusion outperformed the discounting approach in this regard mainly because, in certain situations, the discounting approach incorrectly identified sources as defective. Further work is required to improve this.

While the consistency-based approach found MCS in linear time regarding the number of sources and number of data instances, the redundancy-enhanced version searched the power set of all MCS. Although <sup>∀</sup>*<sup>k</sup>* : <sup>|</sup>**S**MCS−*<sup>α</sup>* (*k*) |≤|**S**| and although, in practical applications, it is reasonable to assume <sup>∀</sup>*<sup>k</sup>* : <sup>|</sup>**S**MCS−*<sup>α</sup>* (*k*) ||**S**|, the scalability of the redundancy-based approach needs to be improved in further works. Another topic that should be addressed in further works is to adapt the design approaches so that they are able to update a topology on streamed data. With new dates becoming available, the epistemic uncertainty is reduced. Updating a topology has the potential to improve the fusion results continuously in small steps.

**Author Contributions:** C.-A.H. conceptualised the methodology, conducted the research, and wrote the article. V.L. supervised the research activity and revised the article. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was partly funded by the German Federal Ministry of Education and Research (BMBF) within the project ITS.ML (grant no. 01IS18041D) and the Ministry of Economic Affairs, Innovation, Digitalisation and Energy of the State of North Rhine-Westphalia (MWIDE) within the project ML4Pro2 (grant no. 005-1807-0090).

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.
