It's not clear what you are trying to achieve here. Publishing Engine is usually access via its web API. To perform a transformation on multiple XML files simply means making a number of API calls, one for each XML file.
Or do you mean a transformation which merges information from a number of XML files into one output?
This can be achieved outside of PE pretty simply using XSLT processor and the document() function. I think the issue you will have with PE is that the composition pipeline expects only one XML input.
You can reference other XML documents in your XSLT, but if they are a local filesystem reference and not a URL then they cannot be resolved when PE is transforming the document. This is because the only other thing that is sent "over the wire" to PE besides the primary XML content is any related graphic files.
We have achieved something similar before by implementing a custom composition pipeline or custom PE function. In that case you package up the XML and its dependencies as a ZIP file, then have PE work on the ZIP file.
thank you for your feedback. Our developer will look into it.
On a side note, our developers was wondering where he can find the jar file mentioned in the Arbortext Content Pipeline Guide.