The integration of Autonomous Guided Vehicles (AGVs) into shared industrial workspaces requires robust safety mechanisms to mitigate collision risks without compromising operational efficiency. Conventional binary safety systems often trigger abrupt emergency stops, leading to mechanical wear and production bottlenecks. This study aims to develop an adaptive decision-making framework for AGV safety based on real-time spatial risk assessment. The proposed approach uses a 3D LiDAR sensor to map unstructured environments and employs the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm to identify dynamic obstacles. A multi-tiered risk zone logic—comprising Danger, Caution, and Ready zones—is implemented to regulate braking responses compliant with ISO 12100 standards autonomously. Experimental validation was conducted through static, linear, and complex non-linear trajectory scenarios. Results indicate that the system successfully eliminates environmental noise and maintains continuous object tracking during erratic movements, triggering gradual deceleration rather than abrupt halts in non-critical situations. This research contributes a scalable safety protocol that enhances Occupational Health and Safety (OHS) compliance while maintaining smoother intralogistics flow on the production floor.