CAJ | 학술논문

传统农业采用人工方式对温室进行控制，但是随着现代农业的快速发展，这种人工投入大、精度低的控制方式已不能满足现代农业需要。该文基于分层有限状态机和事件晶格的概念，建立3层的信息物理系统模型，并提出一种基于分层有限状态机的信息物理系统时空建模方法，同时利用该建模方法开发了新的温室控制系统。该系统能够将物理层传感器感知到的温室物理环境数据通过物理-信息层汇聚节点融合后上报信息层决策节点得到决策信息，物理-信息层控制节点分析决策信息得到控制信息后下传物理层执行器进行控制。由于该系统模型考虑了各层状态机中事件的时空属性，能够将温室控制的正确率由传统基于物联网的温室控制方法的80.20%提高到87.20%，错误肯定率和错误否定率由7.50%和12.30%下降到3.60%和9.20%，保障温室环境满足作物生长对温度、湿度和光照的要求。
전통농업채용인공방식대온실진행공제，단시수착현대농업적쾌속발전，저충인공투입대、정도저적공제방식이불능만족현대농업수요。해문기우분층유한상태궤화사건정격적개념，건립3층적신식물리계통모형，병제출일충기우분층유한상태궤적신식물리계통시공건모방법，동시이용해건모방법개발료신적온실공제계통。해계통능구장물리층전감기감지도적온실물리배경수거통과물리-신식층회취절점융합후상보신식층결책절점득도결책신식，물리-신식층공제절점분석결책신식득도공제신식후하전물리층집행기진행공제。유우해계통모형고필료각층상태궤중사건적시공속성，능구장온실공제적정학솔유전통기우물련망적온실공제방법적80.20%제고도87.20%，착오긍정솔화착오부정솔유7.50%화12.30%하강도3.60%화9.20%，보장온실배경만족작물생장대온도、습도화광조적요구。
Traditional agriculture uses manual way to control temperature and moisture in a greenhouse, but with the rapid development of modern agriculture, this high manpower investment and low accuracy control method cannot meet the needs of the modern agriculture. This paper used the concept of hierarchical finite state machine and lattice-based event to build a 3-layer cyber physical system model, and put forward a hierarchical finite state machine based spatiotemporal cyber physical system modeling method to design a new greenhouse control system. In these modeling methods, the cyber physical system was divided into three layers: physical layer, physical-cyber layer and the cyber layer. There were also two flows in cyber physical system: information gathering flow and decision control flow. The physical layer had sensor nodes, sensor motes and actors, the physical-cyber layer had sink nodes and controller nodes, and the cyber layer had a decision node. The hierarchical finite state machine can easily express the 3-layers system, the state transition between each layer, and the conversion relationship between the two flows in mathematical expressions. In the information gather flow, sensor nodes monitored the physical environment and generated sensor events, sensor motes used physical layer’s hierarchical finite state machine to transform sensor events into physical events, and then passed physical events to sink nodes. Sink nodes used physical-cyber layer’s hierarchical finite state machine to transform physical events into physical-cyber events. In the decision control flow, decision node used another physical-cyber layer’s hierarchical finite state machine to transform the physical-cyber events into cyber events, and passed cyber events to the controller nodes, controller nodes used another physical layer’s hierarchical finite state machine to transform cyber events into control events, and passed control events to the actors. At last, actors used the control events to change the physical environment. The lattice-based event modeling method can be used to divide cyber physical system event into three parts: event attributes, the observer information, the occurrence time and location and attribute information of the event. Event attributes referred that which type the event belonged to, the occurrence time referred that when the event happened, the occurrence location referred that where the event happened, and the attribute information of the event referred the physical environment. Because the 3-layer spatiotemporal model method considered the spatiotemporal attribute into the events of each layer’s state machine, it improved the event detection and control accuracy in the greenhouse effectively, and ensured the greenhouse environment to meet the plant growth demands for temperature, humidity and light. The experiment proved that 3-layer spatiotemporal modeling method which realized the joint modeling with spatial and temporal attributes, reduced the error detection, improved the detection accuracy and the model performance was good. Compared with the traditional control methods based on “Internet of Things”, we found that using 3-layers spatiotemporal cyber physical system modeling in facilities of agriculture, can improve the control accuracy from 80.20% to 87.20%, decrease the false control positive rate from 7.50% to 3.60% and the false negative rate from 12.30% to 9.20%, and it can also be adapted to the modern agriculture requirements of high precision and high automation.