This is a surprisingly difficult problem.

Here’s one take on the matter: A medium power laser is positioned on a hill overlooking the areas where the ice is located.  Either a natural path exists nearby or a rail is laid alongside.  Line of sight, the laser is aimed at the rover and heats the stirling engine up to operating capability.  The rover shoots its red guidance laser; the power station responds with its power laser. Whenever the power station can operate, the rover and ice extraction operations can be done.  This happens whenever solar rays hit the hill where the power station sits.

Microwave sintering is used to create a linear gear rail leading to the areas where ice mining will occur.  The rover rides this rail, its wheels riding a transmission wheel attached to the bucket.  The transmission is just a small gear moving a big gear, a 20:1 speed reduction.  This allows the lightly powered rover to move a bucket of ore, ice, or whatever up the side of the crater while riding the rail.

To explore the eternally dark crater, the power system on the peak simply must continually hit the heat absorption disk of the rover with its laser beam.  The rover would then use a flywheel for energy storage , allowing some limited movement without power.

After you have the ice, the fun is just beginning.  Perhaps by establishing temperature gradients – areas cut into the moon where the temperature is established and is the same always – the ice can be boiled off fractionally.  This creates an issue of storage of raw materials.  Perhaps the separated materials can be moved back into the crater, or perhaps the separation can be done in the crater itself.

Here is a camera shot of one such crater where the rim sees sunlight a fair amount of the time, but the interior never does.  This is 10 km across crater Erlander, near the North Pole.  A cog rail style railroad transport system slanting down to the floor or base would be one to four km in length.