Abstract
The convergence of Industry 4.0 technologies with precision agriculture demands intelligent, adaptive systems capable of autonomous decision-making under energy and latency constraints. This paper presents AMAI-PF (Agentic Multi-Agent Intelligence for Precision Farming), a novel framework that re-imagines cross-layer routing optimization through the lens of autonomous agents and multi-agent reinforcement learning (MARL). Unlike traditional centralized approaches, AMAI-PF models each sensor node as an independent intelligent agent with localized decision-making capabilities, enabling emergent collective behavior for energy-efficient, reliable data transmission. The framework integrates role-specialized agents for cluster-head selection, route discovery and network monitoring, coordinated through a decentralized governance mechanism. Our approach combines cross-layer optimization (network, physical, and MAC layers) with Q-learning-based adaptive routing, achieving superior performance metrics. Experimental validation using NS-2 simulations on large-scale precision farming networks (100-500 nodes across 1000×1000m2 farm areas) demonstrates that AMAI-PF reduces average energy consumption by 31.4%, extends network lifetime by 220+ rounds, achieves 97.2% packet delivery ratio and maintains end-to-end latency below 8ms. This represents a significant advancement over state-of-the-art protocols including LEACH, EECRP and FEEC-IIR. The agentic architecture enables autonomous adaptation to dynamic environmental conditions making it particularly suitable for precision agriculture applications in smart farming ecosystems.