each interaction. The system integrates CAS operational data with the spatial context defined in Stage 1, enabling contextualized risk assessment aligned with real operational conditions. Since safety constraints differ across zones, a multi-expert architecture is adopted, with an independent fuzzy inference system per functional zone. The methodology, aligned with the overall framework shown in Figure 1 and detailed in Figure 2, is organized into the following steps. Figure 2 – Step-based fuzzy inference methodology for collision risk severity estimation. Step 1. Interaction characterization. Each CAS event is represented through a compact set of interaction descriptors derived from standard telemetry and alert logs, capturing the relative operational state of the interacting assets and the contextual conditions of the encounter. The descriptor set integrates complementary aspects of the interaction, combining motion-related signals, contextual indicators, classification attributes aligned with industry safety frameworks, or any other variable that properly captures a risky condition from a mine operation perspective. This representation remains portable across operational zones and equipment combinations, without relying on site-specific sensor configurations. Step 2. Fuzzy representation and operational calibration. Given the variability of mining traffic and the uncertainty inherent to dynamic operating environments, the system avoids rigid thresholds and instead employs fuzzy representations to model gradual transitions between risk conditions. The fuzzy partitions are defined through a hybrid data-driven and expert-driven process, where data-driven components capture prevailing operational patterns and expert-driven criteria ensure alignment with established safety practices and operational policies. To maintain alignment with evolving traffic behavior, routing configurations, and operating conditions, the fuzzy representations are periodically updated through batch processing cycles using recent operational data, enabling controlled adaptation to changing conditions. Step 3. Expert-driven fuzzy severity inference. The fuzzy rule base is constructed using expert knowledge from mine safety and CAS specialists. Within each contextual module, this expertdriven rule base integrates the interaction descriptors and produces a continuous severity index on a normalized scale (0-100%). This inference process enables consistent, context-aware risk estimation across different operational zones. The resulting severity values are mapped to a
RkJQdWJsaXNoZXIy MTM0Mzk2