随着我国农村人口老龄化程度不断加深，提升水稻生产智能化水平刻不容缓，其中，收获环节的智能化需求尤为迫切，且实现难度较大。当前，智能水稻收获机因粮仓容量有限，在作业过程中需频繁卸粮，这严重影响了作业效率。针对这一问题，本文结合国内外研究现状，系统梳理了定点卸粮与跟车卸粮这两种协同作业模式的关键技术研究进展。在定点卸粮技术方面，通过运用高精度空间几何建模、停车距离补偿预测以及立体视觉检测技术，实现了纵向偏差小于0.20 m、横向偏差小于0.10 m的精准对位，有效提高了卸粮的准确性和效率。在跟车卸粮技术方面，基于改进的机间通信协议（采用电台/4G双模）以及卡尔曼滤波延时补偿方法，成功将通信误差降低82.00%以上。同时，结合增益自调整单神经元控制算法，使动态协同纵向偏差稳定控制在±0.08 m以内，显著提升了跟车卸粮的稳定性和可靠性。在路径规划方面，本文构建了基于改进蚁群算法的多目标优化模型。仿真结果表明，采用该模型可使协同作业效率提升13.58%，进一步优化了收获作业流程。通过集成上述技术，所构建的协同系统可使水稻收获效率达到0.42 hm²/h，相较于单机作业效率提升了26.00%。然而，现有研究成果在复杂农田环境适应性以及不规则地块适用性方面仍存在一定局限。未来，需要进一步强化系统的鲁棒性，并开展多场景验证工作，以推动智能水稻收获技术的广泛应用和发展。

With the continuous deepening of rural population aging in China, it is urgent to enhance the intelligence level of rice production. Among various production stages, the demand for intelligence in the harvesting process is particularly pressing and challenging to achieve. Currently, due to the limited capacity of grain bins in intelligent rice harvesters, frequent unloading is required during operations, which significantly impacts the operational efficiency. To address this issue, this paper systematically reviews the research progress on key technologies of two collaborative operation modes, namely fixed-point unloading and vehicle-following unloading, by integrating the current research status at home and abroad. In terms of fixed-point unloading technology, through the application of high-precision spatial geometric modeling, parking distance compensation prediction, and stereo vision detection technology, precise alignment with longitudinal deviation less than 0.20 m and lateral deviation less than 0.10 m has been achieved, effectively improving the accuracy and efficiency of unloading. In the aspect of vehicle-following unloading technology, based on an improved inter-machine communication protocol (utilizing radio/4G dual-mode) and the Kalman filter delay compensation method, the communication error has been successfully reduced by over 82.00%. Meanwhile, combined with the gain self-adjusting single-neuron control algorithm, the dynamic collaborative longitudinal deviation is stably controlled within ±0.08 m, significantly enhancing the stability and reliability of vehicle-following unloading. Regarding path planning, this paper constructs a multi-objective optimization model based on an improved ant colony algorithm. Simulation results indicate that the adoption of this model can improve collaborative operation efficiency by 13.58%, further optimizing the harvesting process. By integrating the aforementioned technologies, the constructed collaborative system enables a rice harvesting efficiency of 0.42 hectares per hour, representing a 26.00% increase compared to single-machine operations. However, existing research still has certain limitations in terms of adaptability to complex farmland environments and applicability to irregular plots. In the future, it is necessary to further strengthen the system's robustness and conduct multi-scenario validation to promote the widespread application and development of intelligent rice harvesting technology.