As part of a multi-national study to compare exact and heuristic techniques to solve a range of very hard combinatorial problems associated with military applications, we describe the application of genetic algorithms to the solution of the Radio Link Frequency Assignment Problem, (RLFAP). The RLFAP, as described here, is that of assigning frequencies, from a limited set of discrete frequency values, to a number of radio links in such a manner as to simultaneously satisfy a large number of constraints and use as few distinct frequencies as possible. This problem is known to be NP-complete. A range of new genetic operators is described. These were developed to overcome the high level of epistasis that occurs. Dynamically altering the priorities of the search also proved to be effective in improving the performance of the genetic algorithm, and two methods of effecting this are described. In addition, we comment on the ways and effectiveness with which ideas based on dedicated heuristics for this problem can be incorporated into the GA. Finally we describe a hybrid GA for the RLFAP and comment on the performance of the various approaches described. This work is being undertaken as part of the EUCLID (European Cooperation for the Long Term in Defence) CALMA Project — RTP 6.4 (Combinatorial Algorithms for Military Applications [Haje93]).