QoS based resource management for concurrent operation using MCTS
Sebastian Durst, Kilian Barth, Tobias Müller, Pascal Marquardt
TL;DR
The paper addresses the challenge of enabling concurrent operation in multifunction RF systems by extending the quality-of-service based resource allocation model (Q-RAM) to handle multiple simultaneous tasks. It introduces a Monte Carlo Tree Search (MCTS) framework that searches a tree of feasible task combinations defined by a compatibility matrix, aiming to maximize system utility $u(\phi,e)$ under resource constraints $\sum_{i=1}^n (g_i(\phi_i))_j \le R_j$ for all $j$. A key contribution is the explicit tree-based concurrency model and the MCTS solver, which stores the best leaf utilities to accelerate inference and is agnostic to the chosen concurrency mode (e.g., split-aperture, interleaving, or specialized waveforms). Experimental verification using Fraunhofer FHR's CoRaSi demonstrates that concurrent operation yields higher total utility and improved tracking performance under realistic mission profiles, highlighting the approach's potential for real-time, cognitive MFRFS management and its applicability to diverse concurrency strategies. The work advances practical, QoS-aware resource management for complex RF systems operating under dynamic environmental conditions.
Abstract
Modern AESA technology enables RF systems to not only perform various radar, communication and electronic warfare tasks on a single aperture, but even to execute multiple tasks concurrently. These capabilities increase system complexity and require intelligent or cognitive resource management. This paper introduces such a resource management framework based on quality of service based resource allocation and Monte Carlo tree search allowing for optimal system usage and profound decision-making. Furthermore, we present experimental verification in a complex application scenario.