Reinforcement Learning (Q-Learning/DRL): Applied for routing but does not scale efficiently

Theoretical Gaps: Theoretical frameworks for unified trust-anomaly are missing

 Military Use Cases: Really fast detection, very high security, and minimal delays

Metrics Vary Widely: Packet Delivery Ratio, End-to-End Delay, Throughput, Detection Accuracy are common terms

The Synthetic Data Could Be More: Traffic generated randomly generally does not reflect real-world characteristics

Scalability Limitations: DRL agents often come to failure within extremely high mobility of nodes or very large networks.

Centralized Detection System: Still Adopted but risks data aggregate and leakages

IEEE 802.11s (Wireless Mesh): At best, highly sparingly considered in either simulation or protocol

Trust Is an All-Inclusive Term: Definitions and calculation methods vary widely, and there is no consensus

Traditional Routing Protocols: AODV, DSR, OLSR dominate early solutions