Recent studies report high accuracy in tomato disease and ripeness recognition using MWD-YOLO, CNN pipelines, and reinforcement learning strategies. Edge deployment on Raspberry Pi validates practical field-level inference under constrained compute.

Robotic harvesting research demonstrates feasibility of 4-DOF and 6-DOF manipulators in dense foliage. Specialized end-effectors and multisensory systems improve selective handling, but many implementations focus on actuation in isolation.

LoRa/LoRaWAN provides long-range, low-power monitoring for farms. However, unstable links and congestion remain common in humid and infrastructure-limited conditions. SDN and SD-WAN studies show adaptive QoS and dynamic routing can strengthen reliability.

The RP comparison shows most prior studies optimize only one layer: either AI detection, robotic manipulation, or network architecture. Cross-layer validation from sensing to actuation with network resilience is still limited, which motivates the Agri-Sense unified design.