*3.2. Search Strategy*

An exhaustive search of the articles regarding EI was performed in February 2023 by two of the authors on the following digital libraries: ScienceDirect, Scopus, IEEEXplore, ACM Digital Library, and Web of Science. The search focused on retrieving scientific publications proposing solutions (i.e., models, techniques, approaches, architectures) that aim to understand how to bring intelligence as close as possible to the data sources. The keyword search string was defined according to three key concepts: (i) edge intelligence; (ii) embedded intelligence; (iii) on-device intelligence. Considering such key terms and their synonyms, the following search string was identified:

$$("\text{\textquotedblleft edge\textquotedblright OR\textquotedblright}\text{\textquotedblleft OR\textquotedblright OR\textquotedblright}\text{\textquotedblright}\text{\textquotedblleft on\textquotedblleft}\text{\textquotedblleft AND\textquotedblright}\text{\textquotedblleft OR\textquotedblleft All\textquotedblright})\tag{1}$$

To find relevant results, we applied the search string to articles' titles, and we forced a distance of a maximum of two words between the key terms. Later on, since the object of our work was specifically on EI's secondary studies, we applied the following search string to manuscripts' abstracts:

$$(survey^\* \text{ OR } reviw^\* \text{ OR } literatur \text{ OR } round \text{ map } OR \text{ discuss\*}) \tag{2}$$

It is worth noting that the search string (2) incorporates the wildcard symbol "\*", and this allows for the stemming task over key terms such as "survey", "review", and "discussion".

## *3.3. Eligibility Criteria*

The articles were eligible for selection if they met all of the following inclusion criteria:


The articles were excluded from the selection if they met one of the following exclusion criteria:

