This paper deals with the local optimization of the periodic or aperiodic ambiguity functions of a set of complex sequences by using a gradient method. This optimization aims to locally minimize the cross-ambiguity and the auto-ambiguity functions of a family of sequences for Multi-Static Primary Surveillance Radar (MSPSR) systems. The optimization of the ambiguity functions is done by an optimized gradient. An extension is also presented for controlling both the spectrum and the ambiguity functions.