I assume you're using the 1.x version of 3DMLW engine.
In this case you could take any kind of shape, make it collidable (and optionally invisible) and move it. Then the camera could reference the shape's (finalx,finaly,finalz) for it's (x,y,z).
You should try out our version 2.0, which is a little more flexible and mature in those areas. Unfortunately there's no documentation available for the 2.0, but if you're interested in moving to the next version early on then PM me.

Check out the nightly builds for your platform:
http://3dmlw.com/download/nightly_builds/http://3dmlw.com/download/nightly_builds_linux/