Answers are grounded in this paper and cite their sources.
Although smart agriculture is rapidly evolving toward data-driven cyber-physical ecosystems, existing Internet of Things (IoT) frameworks remain heavily dependent on centralized cloud computing. This dependency introduces communication bottlenecks, execution latency, and security vulnerabilities, limiting real-time responsiveness in resource-constrained rural environments. To address these challenges, this paper proposes Farming of Things (FoT), a five-layer cyber-physical architecture comprising Perception, Transmission, Ingestion, Cognition, and Execution layers that integrates IoT, edge computing, and cloud intelligence. The framework dynamically distributes computational workloads across mobile, edge, and cloud tiers using a Block Coordinate Descent (BCD)-based task-offloading algorithm to jointly optimize end-to-end latency and energy consumption. To protect sensitive agricultural telemetry, a Lightweight Secure Data Transmission (LSDT) protocol is introduced using AES-128 encryption and HMAC-SHA256 authentication. The proposed architecture is evaluated through a physical testbed and large-scale discrete-event simulations. Experimental results show that LSDT achieves secure transmission with only 1.5% communication overhead and a cryptographic delay of 0.94 ms. The distributed multi-model AI stack demonstrates strong inference performance, with on-device TinyML achieving 99.32% accuracy, Convolutional Neural Networks (EfficientNet-B4) achieving 97.13% accuracy, and a Large Language Model (Qwen 2.5-3B) supporting global agronomic decision synthesis. Scalability experiments show that static Mobile-Only, Edge-Only, and Cloud-Only strategies experience queue exhaustion at task arrival rates of λ = 10 , λ = 20 , and λ = 40 , respectively, whereas dynamic FoT orchestration maintains sub-0.2-second latency up to λ = 30 and remains scalable up to λ = 60 . Despite these improvements in real-time automation, performance remains constrained by the battery capacity of localized edge nodes and wide-area network propagation delays under maximum-load cloud offloading.
No takes yet. Share an insight, caveat, or question.
Das et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: