Of all the more modern techniques of synthesis that I have seen, it's physical modelling that I feel holds the most promise for futher development. Not physical modelling of analogue synthesizers which I am not that crazy about, but physical modelling of instruments. Of all the synth makers out there, its Yamaha that has been the greatest advocate of physical modelling with Korg following with the Korg OASYS (although no new physical models have come out of the OASYS. On the software side we have Applied Acoustic Systems with Tasman and String Studio and an analogue modeller. Native Instruments Reaktor also has some physical modelling synthesizers.
I guess the reason that you don't see a lot of physical modelling synthesizers is that they take a lot of work and complex mathematics. The benefit is that you have a natural sounding and responding instrument that can be changed as easy as turning a knob which in some ways, is like having a whole warehouse full of instruments from more traditional sounds to other exotic sounds.
Of course, the actual feel of the instrument itself is lost. Controllers are getting better but I think we are a long way from a controller that is as expressive as the real thing. Controllers like the Eigenharp, while a good start, don't impress me as they are little more than MIDI triggering devices. Eigenharp does not even support OSC. ON the percussive side, the Korg Wavedrum seems to do a little better and for violins, the K-Bow is also an improvement.
So at NAMM 2010 we find the CP series of keyboards. I like these and if I had a lot of money and space I would get one. Clearly, there is some degree of physical modelliing going on here but how much is sampled and how much is modelled is not clear. Nice, but not really groundbreaking.
Korg also has the wavedrum which for percussion is clearly a step in the right direction considering this this drum responds differently depending on if you use brushes, sticks, mallets, ect.
But a quick look at the physical modelling offerings at winter NAMM 2010 clearly shows a reluctance on the part of developers to develop physical modelling synths soft or hard beyond what is already out there last year or beyond that.