A GROUND-BASED ROBOTIC SURVEILLANCE ARCHITECTURE FOR EARLY THREAT DETECTION AND REAL-TIME ALERTING IN BORDERLESS MINING ZONES

Abstract

Artisanal, small-scale, and remote mineral extraction sites frequently operate without fixed, legally demarcated perimeters, which leaves them exposed to trespass, equipment theft, illegal excavation, and personnel-safety incidents that conventional security infrastructure struggles to address. Fixed camera towers require power and communication backbones that are rarely available at remote sites, human patrols are costly and hazardous, and aerial platforms are constrained by flight endurance, weather, and airspace regulation. This paper proposes a ground-based robotic surveillance architecture that combines a multi-modal on-board sensor suite (3-D LIDAR, thermal imaging, millimetre-wave radar, acoustic sensing, and GNSS-RTK/IMU), an edge-resident Extended Kalman Filter (EKF) sensor-fusion and classification pipeline, a resilient LoRa mesh/cellular communication backhaul, and a cloud command-center dashboard for real-time alerting. A terrain-aware, three-dimensional detection-probability model is developed to capture the effect of ground occlusion on sensor coverage, and a boustrophedon coverage-path-planning formulation is used to schedule multi-robot patrols over an unbounded mining perimeter. Simulation results indicate a fused detection accuracy above 90% within a 150 m operating range, a mean end-to-end alert latency under two seconds for networks of up to twenty nodes, and a favorable trade-off between detection performance and energy consumption relative to fixed-camera and unmanned-aerial-vehicle baselines. The proposed architecture offers a practical, extensible template for continuous, low-infrastructure security monitoring of open and border-adjacent mining zones.