Policy Benchmarking for Delay‐Aware Multi‐Target Engagement in Gimbaled and Turret‐Based Defense Systems

Abstract

ABSTRACT Modern gimbaled electro‐optical systems and turret‐based defensive platforms increasingly operate in dense and time‐compressed environments in which multiple heterogeneous targets must be serviced under non‐negligible slewing, acquisition, servicing, and cooldown delays. In such settings, the problem is not limited to estimating target threat, but extends to determining an engagement order that balances tactical value, temporal feasibility, and protected‐zone safety. This study presents a reproducible benchmark framework for delay‐aware multi‐target prioritization and engagement sequencing in a two‐dimensional operational environment. The proposed framework is not a learning‐based artificial intelligence (AI) decision system; rather, it is an optimization‐inspired heuristic scheduling benchmark in which candidate targets are selected using explicitly defined utility functions and online decision policies. The framework combines semantic target characterization, geometric urgency, explicit execution delays, and multiple online scheduling policies within a unified simulation architecture. In addition to reactive baselines, a robust lookahead decision layer is included to evaluate candidate actions under alternative escalation conditions. The benchmark comprises 400 runs covering five scenario families, eight decision policies, and ten random seeds per condition. Performance is assessed using weighted handled priority, completed targets, high‐priority timeliness, waiting delay, breach count, and switching rate. The results show that no single policy is uniformly dominant across all objectives and scenarios. Nevertheless, throughput‐ratio, robust look ahead, and hybrid policies consistently form the upper performance tier. Overall, throughput‐ratio achieves the strongest global compromise, yielding the highest weighted handled priority (7.4070) and improving weighted priority and completed targets by 8.24% and 25.53%, respectively, relative to the greedy baseline, while leaving breach count nearly unchanged. Paired analysis further indicates a statistically significant weighted‐priority advantage of throughput‐ratio over greedy (, ). These findings suggest that delay‐aware engagement in slew‐constrained defensive systems is better interpreted as a structured policy‐selection problem than as a search for a universally dominant rule.