听觉系统是机器人感知周围环境信息的重要途径之一，精准有效地进行声源定位，可极大提高移动机器人的感知与决策能力。将声源定位应用于危险环境救援与巡检具有重要工程意义。随着深度学习的广泛应用，引入卷积神经网络（convolutional neural networks, CNNs）的声源定位效果显著改善。将移动机器人声源定位研究从网络架构与改进、声音特征类型、数据仿真与增强，以及多模态信息融合四个角度进行综合对比及分析，并对技术的应用提出思考与展望。

The auditory system is considered one of the crucial pathways for robots to perceive environmental information. The perception and decision-making capabilities of mobile robots are greatly enhanced by accurate and effective sound source localization, making it highly significant for applications in hazardous environment rescue and inspection. With the widespread adoption of deep learning, the effectiveness of sound source localization has been notably improved through the introduction of convolutional neural networks (CNNs). Sound source localization for mobile robots was comprehensively compared and analyzed from four perspectives: network architecture and improvements, types of sound features, data simulation and augmentation, as well as the fusion of multimodal information. Reflections and prospects on the application of the technology are also presented.