This study introduces a robust measurement update method for panel-based bathymetric simultaneous localization and mapping (SLAM). The bathymetric SLAM algorithm enables autonomous navigation in unknown underwater environments without using external telemetry systems. The bilinear panel model is used to represent the surface of the terrain, which allows SLAM with a single-beam acoustic altimeter. However, this approach may suffer from incorrect measurement update when the estimated location does not belong to the panel where the vehicle is actually located. The dynamic grid adaptation method determines the region in which bilinear interpolation is used based on the confidence interval, and the constrained extended Kalman filter is used for the estimation of inner grid points. Numerical simulations are presented to demonstrate the performance of the bathymetric SLAM with the proposed measurement update method compared with the standard panel-based bathymetric SLAM.