EMission-Aware Multi-Objective Policy Learning for Adaptive Routing in Tactical Mobile Ad Hoc Networks

Tactical Mobile Ad Hoc Networks are those whose conditions can hardly be assumed to be constant and they are often not easy to model with fixed conditions. The ability of nodes to move suddenly and often, the varying and unpredictable needs of a mission, and the changing nature of operations, cause these changes in the priorities of missions. This paper builds on the previous studies regarding machine learning-assisted routing of MANET by proposing a mission-aware and multi- objective policy learning model on adaptive routing in tactical settings. The method that is proposed based on the building up also leaves metric-centric optimization; rather it also trains routing policies that expressively optimize between throughput and latency, packet delivery reliability and energy consumption within mission-specific constraints. The framework formulates routing as a situational decision problem instead of assuming the existence of a single overall objective, with the weighting of the performance measures depending on the operational context, that is, high-mobility patrols, power limited operations, or time- sensitive coordination problems. Lightweight learning mechanisms alongside supervised learning models that have been trained on simulation datasets of high-fidelity NS-3 are used to permit real time modification of routes according to the projected network states and mission priorities. The results of numerous simulation studies involving a wide range of mobility models, traffic workloads, and node densities show that the proposed mission-aware policy learning strategy is always superior to the classical routing protocols or metric-based learning algorithms.