I think realistic character animation in FPS games is a tougher problem to solve than would intially appear.
In life, the expectation of changing direction results in the prepatory actions neccessary to see it through. If I am running forward and I plan to stop, I will slow and plant my heels accordingly so that I stop at the desired location or as soon as possible. The important thing to note is that I planned to stop at a particular point, and the stop actually occurred after my decision.
On a computer, since the controlled character is disjoined from your intent, it can’t animate in preparation for the intended action. If you’re running forward in Unreal Tournament, how does the computer know that you plan to stop before the piller and break right? Certainly in real life, knowing this course of action, you would shift your weight accordingly. If you didn’t know, well, as you might suspect hilarity would likely ensue…
With some sort of action/display lag, improved animation is possible. Across a network, it may also be possible to hide this lag whilst not seriously hindering gameplay. Another potential solution is a predictability model of motion. For example, a computer uses the model to determine preparatory animations to stop as a character approaches a wall.
It seems, for the time being, that graphics remain the focus of the technological advance in games, but I’m sure it’s only a matter of time before this level of detail is addressed.