Unmanned Underwater Vehicles (UUVs) are of great importance in ocean robotics, because of the wide industrial and military application range they span. Tracking a undersea moving object is frequently encountered in the UUV applications, such as tracking a halobios. To achieve this, the UUV needs to follow the target in certain distance based on the mother ship's sensor information. The traditional tracking methods perform poor in controlling the distance, the value of which depends on the random sensor observation noise. Based on the noise model of the sensor observation, this paper proposes a tracking method to minimize the mean squared-error (MSE) of the distance, which reduces the effect of the sensor observation noise on the distance. As the optimal tracking method may be too computationally intensive to implement in practice, we derived a suboptimal tracking method, which has a closed form. Simulation results verify the improvement of tracking precision of the optimal tracking method. In addition, the suboptimal tracking method, which is practical and easy-to-implement, performs very close to the optimal tracking method.