<p>Traditional positioning methods are usually assumed that the measurement noise is independent and identically distributed Gaussian noise, while the noise variance in the actual underwater environment is not static. To this end, a distance-dependent noise model is established in this paper. The noise variance is modeled as a function of the distance between the sensors and the target. A robust underwater time difference of arrival (TDOA) localization method under the condition of distance-dependent noises is proposed to solve the problem of uneven distribution of sensors and noise interference in underwater sound source localization. By introducing a Boolean vector to characterize the selection state of the sensors, the minimum Cramer-Rao lower bound (CRLB) is taken as the core criterion, and the distance and angle characteristics between the sensors are considered. By using semidefinite relaxation (SDR), the non-convex problem is transformed into a convex semidefinite programming problem. Simulation results show that the proposed algorithm has higher positioning accuracy and stronger robustness.</p>