First, the existing audio-visual navigation dataset only has limited audio samples, making it difficult to simulate diverse multiple sound-source environments. Second, existing navigation frameworks ...