In the next generation of mine hunting sonars, in particular on Autonomous Underwater Vehicles (AUVs), Synthetic Aperture Sonar (SAS) will play an important role. The benefit of SAS is to increase resolution and signal-tonoise ratio by coherent processing of successive pings. A challenge in SAS is the computational load. This is caused primarily by the actual transformation of motion compensated acoustic data into an image, i.e. beamforming. The major issue for the fastest method, wavenumber frequency processing, is the assumption that the synthetic aperture is a uniform linear anay. Platform deviations from the straight track degrade performance. This paper describes two methods to compensate for these deviations. Theoretical performance of the methods is presented for both in-plane and out-of-plane path errors. They are tested for a wide range of system conhgurations, including experimental navigation data from the HUGIN AUV. The result is a proof of operational validity of the proposed methods.