Abstract: In order to improve the intelligence and work efficiency of an in-tank wall-climbing robot, a robot localization system using passive acoustic source localization technique is designed. The system receives audio signals emitted from robot with microphone array, and processes the signals with an improved localization method based on time delay estimation, thus the localization of robot is completed. Audio signal processing methods related to the system are represented and the localization data is processed with the Kalman filter algorithm. The system is proved that the positioning error is no more than 12cm within 15m.