A low-cost, low-power, and low data-rate solution is proposed to fulfill the requirements of information monitoring for actual large-scale agricultural farms. A small-scale farm can be easily managed. By contrast, a large farm will require automating equipment that contributes to crop production. Sensor based soil properties measurement plays an integral role in designing a fully automated agricultural farm, also provides more satisfactory results than any manual method. The existing information monitoring solutions are inefficient in terms of higher deployment cost and limited communication range to adapt the need of large-scale agriculture farms. A serial based low-power, long-range, and low-cost communication module is proposed to confront the challenges of monitoring information over long distances. In the proposed system, a tree-based communication mechanism is deployed to extend the communication range by adding intermediate nodes. Each sensor node consists of a solar panel, a rechargeable cell, a microcontroller, a moisture sensor, and a communication unit. Each node is capable to work as a sensor node and router node for network traffic. Minimized data logs from the central node are sent daily to the cloud for future analytics purpose. After conducting a detailed experiment in open sight, the communication distance measured 250 m between two points and increased to 750 m by adding two intermediate nodes. The minimum working current of each node was 2 mA, and the packet loss rate was approximately 2-5% on different packet sizes of the entire network. Results show that the proposed approach can be used as a reference model to meet the requirements for soil measurement, transmission, and storage in a large-scale agricultural farm.
Stochastic computing (SC) has a substantial amount of study on application-specific integrated circuit (ASIC) design for artificial intelligence (AI) edge computing, especially the convolutional neural network (CNN) algorithm. However, SC has little to no optimization on field-programmable gate array (FPGA). Scaling up the ASIC logic without FPGA-oriented designs is inefficient, while aggregating thousands of bitstreams is still challenging in the conventional SC. This research has reinvented several FPGA-efficient 8-bit SC CNN computing architectures, i.e., SC multiplexer multiply-accumulate, multiply-accumulate function generator, and binary rectified linear unit, and successfully scaled and implemented a fully parallel CNN model on Kintex7 FPGA. The proposed SC hardware only compromises 0.14% accuracy compared to binary computing on the handwriting Modified National Institute of Standards and Technology classification task and achieved at least 99.72% energy saving per image feedforward and 31× more data throughput than modern hardware. Unique to SC, early decision termination pushed the performance baseline exponentially with minimum accuracy loss, making SC CNN extremely lucrative for AI edge computing but limited to classification tasks. The SC's inherent noise heavily penalizes CNN regression performance, rendering SC unsuitable for regression tasks.