Knowledge Reasoning for Unmanned Systems Based on Multi-layer Behavior Modeling

Intelligent unmanned systems are increasingly deployed in dynamic and complex environments, where their behaviors exhibit multi-dimensional characteristics. Effectively modeling these heterogeneous behaviors and reasoning over their temporal and causal relationships are critical for anomaly tracing and risk prediction. In this paper, we propose a dynamic reasoning framework that models system behaviors across three semantic layers: functional, network, and device, and constructs an integrated multi-layer behavior model to capture cross-dimensional dependencies. Building on this representation, we design dynamic reasoning mechanisms combining backward tracing and forward reasoning, which incorporate historical knowledge and real-time updates to reconstruct the root cause of anomalies and predict potential risks. Experiments conducted on real-world and public datasets demonstrate that the proposed method delivers competitive performance in reasoning accuracy, efficiency, and scalability, and effectively reduces data redundancy while maintaining high information completeness.