Another, completely different solution is not to use pawn movement at all. No movement = nothing to interrupt 
- when a key is hit and direction established, trace until you hit something
- now you have start & end points, measure the distance to calculate timeline play ratio
- use the timeline to Lerp between start & end
This is quite efficient as it would require a minimalistic setup regarding collision, it also does not tick. It’s not the best choice if you’re planning on having moving / dynamic obstacles, though. On the other hand, a TL would give you plenty of control of the curve so you could setup something advanced quite rapidly.