We show how additional constraints, restricting the spectrum of the optimized pulse or
confining the system dynamics, can be used to steer optimization in quantum control 
towards distinct solutions. Our examples are multi-photon excitation in atoms and 
vibrational population transfer in molecules. We show that a spectral constraint is 
most effective in enforcing non-resonant two-photon absorption pathways in atoms and 
avoiding unnecessarily broad spectra in Raman transitions in molecules. While a 
constraint restricting the system to stay in an allowed subspace is also capable of 
identifying non-resonant excitation pathways, it does not avoid spurious peaks in the 
pulse spectrum. Both constraints are compatible with monotonic convergence but imply 
different additional numerical costs.