I just learned about this recently during the Udacity SDC! They go over an algorithm called A* hybrid which was super neat too. Instead of the discrete decision

I just learned about this recently during the Udacity SDC! They go over an algorithm called A* hybrid which was super neat too. Instead of the discrete decisions of up, down, left ,and right, you define a set of state space equation (e.g. x(t+1) = x(t) + vΔt). This way, your results ends up being continuous and smooth which a car can actually follow. I think this was invented by the Sebastian Thurn's team during DARPA challenge.