苗中华等:具身智能农业机器人关键技术与发展趋势

图 1 近年来具身智能领域的代表性成果

Fig.1 Representative achievements in the field of embodied intelligence in recent years

图 2 具身智能农业机器人研究范式

Fig.2 Research paradigm for embodied intelligent agricultural robots

图 3 农业作物不同程度受遮挡示例

Fig.3 Examples of Agricultural Crops with Varying Degrees of Occlusion

图 4 基于启发式的主动感知方法

Fig.4 Heuristic-based active perception method

图 5 动态调整相机视角实现主动感知

Fig.5 Active perception through Dynamic Camera Viewpoint Adjustment

图 6 三种多模态信息融合架构

Fig.6 Three multimodal information fusion architectures

图 7 特征级别融合示意图

Fig.7 Feature level fusion schematic diagram

图 8 端到端导航架构图

Fig.8 End to end navigation architecture diagram

图 9 基于模仿学习的农业机器人应用示例

Fig.9 Application examples of agricultural robots based on imitation learning

图 10 基于视觉-语言模型的导航应用示例

Fig.10 Example of vision language model based navigation application

图 11 LLM 与经典任务规划相结合示例

Fig.11 Example of combining LLM with classic task planning

图 12 基于点云的触觉表示方法 

Fig.12 A tactile representation method based on point cloud

图 13 智慧农业多机器人任务分配与调度示意图

Fig.13 Schematic diagram of task allocation and scheduling for multiple robots in smart agriculture

图 14 具身模拟与生成的技术路线

Fig.14 Technical roadmap for embodied simulation and generation

图 15 农业场景高保真场景重建流程

Fig.15 High fidelity scene reconstruction process for agricultural scenes

图 16 自监督学习原理图

Fig.16 Principle diagram of self supervised learning

图 17 农业机器人具身诊断与运维关键技术

Fig.17 Key technologies for body diagnosis and operation of agricultural robots