Electronic – Nearby jail’s signal jammer affecting cellphone and devices using 2.4GHz band

What about a photodetector and transistor as the receiver - the transmitter could then be a torch which have to shine on a device to switch it on. This is the simplest solution I can think of.

If you are thinking higher tech communications the lowest power and I suspect lightest RF solution might be a the Texas instruments CC430 Microcontroller which has integrated RF. You could of course remove the packaging to reduce weight.

That module gives you low level access to the radio -- e.g. you're not going to be sending high level prototcol data. Yes -- it's trivial for you to send any arbitrary message to the radio and have your arduino interpret the message as you please (set pwm duty cycle etc). You shouldn't need to do a whole lot else to get proof of concept. If you need super long range 0 dBm might not cut it, but it's definitely OK for several meters.

With regards to your second question: there are a million ways to identify the different radios. The most expensive and worst way is to have a different transmitter for each receiver. It's obvious why this is a bad solution. All your other solutions will require you to implement some kind of higher level protocol. There are lots and lots and lots of options here (this is a whole area of academic research). I used a similar Nordic radio several years ago and it had different channels you could utilize. That can help you (e.g. put one receiver on one channel), but it obviously doesn't scale well (e.g. if you have more devices than channels you're back to square one. Equally importantly you might not want to use certain channels -- they might be noisy, have poor PRR, etc). One simple, but not particularly robust (from the networking perspective) is to invent a protocol wherein you blast a message and depending on the contents of that message one and only one receiver responds. Your arduino will have to implement whatever protocol you want so obviously you can't go crazy with the complexity (e.g. you probably don't want to implement posix sockets). For a few motors (less than 10) you'll probably be OK doing something like this provided that you don't have stringent latency requirements.