Not just obvious, but prior art.
Just about any market does this; the change of price brings other players in, or causes them to leave.
I wrote code for a simulation in '95 or so that had the simulated merchants applying a quadratic equation to the amount that their sales missed the sell-out quantity. It was trivial to cause markets to clear, on just that one piece of information. (In fact, at one point, due to a coding error, the product was a "bad" rather than a "good"--and it still cleared at a negative price.
The algorithm for Uber would be trivial: once the wait time goes above or below its usual band, the price adjusts by some portion per time unit (e.g., 1%/minute) until the wait time is normal. Or include lagged time periods to damp oscillations.
This is just plain trivial. I, or any other computational economist, could sit around all day kicking out new algorithms for this.
It's really pretty simple: if you sell out to quickly, or can't service all your customers, raise your prices; if you have excess, lower them. Doing it by algorithm is nothing new; the trick to patentability would be to find an algorithm that not only hasn't been done before, but is actually better than the other trivially reachable algorithms.
I drove the demand in that model various ways, whether constant, sine waves, stochastic, saw tooth, and probably others I'm not recalling off-hand. A rather simple genetic algorithm rapidly converged in all cases. Mathematically, that method was probably mathematically equivalent to large classes, possibly all, other second order and lower and lower methods or solutions--and the method rather clearly could be extended to nth order . . . (second order methods tend to be sufficient for most things).
hawk