Another entry in the field of farm (or rather garden) automation: FarmBot. This is a project to build and sell devices that automate gardening, from seeding through watering and weeding. They put special effort into open source and free access principles, so this is one to watch. I'm also pleased to see that all plastic parts are printable
Of particular interest is their table of yields. They seem to have taken some very conservative numbers for yield, which is appropriate for projecting the performance of a single-layer soil-based plot under natural light and weather conditions. For example, they estimate about 20 g/m² per day for potatoes while I use 65 g/m² per day; my source is a NASA life support paper where those yields were achieved under controlled atmosphere and lighting conditions. That suggests the usual rule of thumb (hydroponics doubles yields) still holds and that further gains (another 50-100%) can be achieved with precise climate control and artificial light.
If I had the time or money I'd contribute to the project, most probably under the crop data source OpenFarm. It looks like they could use some help to get off the ground.
For my purposes I'd prefer to see the system adapted to handling hydroponic/aeroponic trays on a conveyor system. Each tray would be ID tagged (RFID or visual indicators like QR codes). A growth plan for the tray would be entered, then the system would prepare and fill the appropriate media and seed the tray. Seeded trays would be stored in a germination rack, periodically removed to check for progress and proper moisture. As the plants mature their trays would typically spend 8-12 hours in a dark rack resting, go through a morning check for abnormalities, spend 12-16 hours in a lit rack producing, then go through an evening check. Plants that prefer soil-replacement media would be handled much like the existing system, with periodic spot application of water (frequent) and suppression of weeds (which should be extremely rare). Other types would be mostly self-sufficient in their lit racks, but a visual check twice a day for any deviation would still be most efficient with an automated check station handling trays as fast as it can. Trays would be managed by something similar to a warehouse automation system, minimizing wasted space
If a fairly advanced harvesting attachment were developed for leafy greens then the system could do daily pruning of outer leaves, maintaining plants at an optimal size (limiting shading) and yielding a steady stream of produce. Likewise, if the current X-Y bed were adapted into an X-Z bed (standing vertically) or a free-rolling unit (like a warehouse robot) then the system could be used to monitor, prune and harvest vine crops like tomato, pepper, cucumber and squash that are grown via commercial hydroponic techniques.
Something like this could be trialled on the ISS in a very limited form (though the conversion to microgravity operation may be nontrivial), then later in a larger facility. (private, public? NASA exploration gateway, Bigelow hotel or ESA moon village? Maybe even here on Earth at McMurdo or other arctic/antarctic bases?) The medium-term goal would be automating food production on Mars and other destinations of interest. In the long term, new units on Mars, Luna or in space would be built out of locally-sourced aluminum and plastic (via existing 3d printing technology) plus microcontrollers shipped from Earth.