I am generally referring a programmable chip to drive the gates of the FETs.
"A gating for each FET"
??? I am saying that in a P/N FET pair, they both behave in predictable ways relative to one another. WHen one does one thing, the other one always does the opposite (ie. when one is one, the other must be off. you don't want both on at the same time or it makes a short circuit). THis means you ideally can control them with a single signal (you might need an inverter along the way or use a P-type and N-type device so they behave oppositely to the same gate drive signal.
But, in real life FETs aren't identical and turn off slower than they turn on. THis means that if you apply a signal for one to turn off and other to turn on at the same time, there will be a time period when they are both on = short circuit. YOu either need circuitry to deal with this, or use a programmable chip (a microcontroller) that basically controls the gates independently and adds in a delay between switchings to make sure both are never on at the same time.
BTW, when I talk I speak of just one phase. Since controlling all the other phases is pretty much the same thing, you just delay everything in the other phases by 120 degrees.
YOu might as well use a microcontroller for this because if you don't use one, you still need to find a way to generate a sinusoid and a triangle wave to feed into a comparator to produce a PWM gate drive signal that has a sinusoidal fundamental (to produce your AC output). Actually producing those two signals isn't nearly as easy or as flexible as just having it as a big table of lookup values in software.
If you don't understand how a sinusoidal PWM signal is generated, then look at this:
**broken link removed**
Notice that the triangle wave has the same frequency as the PWM signal (it's the carrier) and that whenever it is above the desired sinusoidal signal (stored as a lookup table or equation in memory) then the PWM is HI and when it is below, then it is LO.