Contemporary fruit harvesting operations are confronted with several challenges, including labor shortages, low picking efficiency, and complex operational environments. These factors underscore the imperative for developing intelligent harvesting equipment endowed with high-precision perception and autonomous operation capabilities to comprehensively enhance the efficiency and quality of fruit picking. Sensor technology plays a critical role in key tasks of fruit-picking robots, such as path planning, fruit recognition, localization, and grasping control. In unstructured orchard environments, the collaborative utilization of vision, tactile, and laser sensors facilitates target identification, positional awareness, and obstacle avoidance, significantly improving the robot's adaptability and operational accuracy in complex settings. However, existing sensors still have technical limitations. For example, visual sensors are susceptible to deleterious effects from ambient sunlight interference, occlusion from branches and leaves, and the dense clustering of fruits, all of which impede robust target detection. Tactile sensors often display sensitive to fluctuations in temperature and humidity, which poses challenges for the accurate quantification of complex mechanical feedback, thereby hindering precise control of gripping force. In addition, path optimization in unstructured environments remains challenge, and the high procurement cost of laser sensors limits their large-scale application. Single-sensor systems suffer from limitations such as single-dimensional perception, poor adaptability to environmental variations, and insufficient recognition of fruit characteristics, making them suboptimal for deployment in unstructured orchard conditions. Therefore, this paper explores the future applications of sensor technologies in fruit-picking robots, focusing on the challenges of multi-sensor fusion, including data heterogeneity, temporal synchronization, and computational complexity. It highlights that the integration of multi-spectral imaging technologies such as infrared and ultraviolet, high dynamic range (HDR) imaging, and flexible electronic skin combined with biomimetic structural designs in multi-sensor fusion systems holds great promise for widespread application.