Differentially constrained motion replanning using state lattices with graduated fidelity