Abstract-Recent research on the array geometrical configuration shows that the two L-shaped array (TLSA) has a lower CramerRao Low-Bound (CRLB) of two-dimensional (2-D) directions-of-arrival (DOAs) estimation than other array configurations. However, in this array configuration, there are some problems to note: i) three electric angles are independently obtained from three uniformly linear subarrays on three axes, so they must be matched before solving elevation and azimuth angles from them; ii) Similar to other array geometries, the effect of mutual coupling in the TLSA on the estimation performance cannot be ignored; and iii) the conventional elevation estimators may encounter estimation failure.In this paper, we develop a new TLSA-based 2-D DOAs estimation algorithm. The key points of this paper are: i) using some particularly selecting matrices, a trilinear model is constructed to compensate the effect of mutual coupling on three subarrays. In addition, the steering vector is restored using the trilinear alternating least square method; ii) 2-D DOAs are estimated from the properly chosen elements of the restored steering vector to avoid pairing parameters and the severe performance degradation resulted from the failure in pairing; and iii) a new elevation estimator is designed to avoid estimation failure. Simulation results are presented to validate the performance of the proposed method.