Unmanned Aerial Vehicles (UAVs) have recently attracted significant attention due to their potential applications in reconnaissance and search. This paper aims to investigate the issue of multi-UAV cooperative reconnaissance and search (MCRS) to ensure ample coverage of the mission area and precise localization of static targets. The MCRS problem is modeled as a multi-objective optimization problem, taking into account the credibility of search results. To achieve this, we design a belief probability map based on the Dempster-Shafer (DS) evidence theory, comprising an uncertainty map and two target maps. This representation enables a clear depiction of both the presence of the target and the uncertainty within the map. Subsequently, we reformulate this multi-objective optimization problem within the framework of Decentralized Partially Observable Markov Decision Process (Dec-POMDP). To address this reformulation, a new deep reinforcement learning approach called Double Critic Deep Deterministic Policy Gradient (DCDDPG) is proposed. Specifically, we introduce both a centralized critic and a local critic for each UAV agent to estimate the action-value function. This approach helps balance the bias in the action-value function estimation and the variance in the policy updates, thereby improving the coordination effect. Extensive simulation results demonstrate that DCDDPG outperforms existing techniques in terms of search efficiency and coverage.

To achieve model reuse in combat effectiveness simulation systems development, cognitive decision behaviors are usually implemented using a scripting language, which is separate from the programming language used to implement simulation models. Therefore, it is desirable to establish a much better grounding for cognitive behaviors modeling. In the context of domain specific modeling, metamodeling from scratch for designing such a scripting language poses some limitations, among which is the issue of integrating various models that are represented by various customized languages with different syntax and semantics, together with a large expenditure of designing, implementing, and maintaining these languages and their supporting resources. Instead, UML profile-based metamodeling is adopted, as a lightweight extension to capture the cognitive domain specific concepts, relationships, and constraints. Moreover, a unifying framework is proposed to guide the cognitive domain specific profiles design. Upon this framework, the development process is shown through constructing an anti-submarine tactical profile in combat effectiveness simulation systems domain and the feasibility of the domain specific language is illustrated with an armed escort scenario.

Unmanned Aerial Vehicles (UAVs) have garnered widespread attention in reconnaissance and search operations due to their low cost and high flexibility. However, when multiple UAVs (multi-UAV) collaborate on these tasks, a limited communication range can restrict their efficiency. This paper investigates the problem of multi-UAV collaborative reconnaissance and search for static targets with a limited communication range (MCRS-LCR). To address communication limitations, we designed a communication and information fusion model based on belief maps and modeled MCRS-LCR as a multi-objective optimization problem. We further reformulated this problem as a decentralized partially observable Markov decision process (Dec-POMDP). We introduced episodic memory into the reinforcement learning framework, proposing the CNN-Semantic Episodic Memory Utilization (CNN-SEMU) algorithm. Specifically, CNN-SEMU uses an encoder–decoder structure with a CNN to learn state embedding patterns influenced by the highest returns. It extracts semantic features from the high-dimensional map state space to construct a smoother memory embedding space, ultimately enhancing reinforcement learning performance by recalling the highest returns of historical states. Extensive simulation experiments demonstrate that in reconnaissance and search tasks of various scales, CNN-SEMU surpasses state-of-the-art multi-agent reinforcement learning methods in episodic rewards, search efficiency, and collision frequency.

Current simulation models are not only represented in the form of traditional data or formula for pure theory analysis but expanded to be simulation modeling assets that are featured with complicated structure, diverse behaviors, and abundant semantics. Semantic composability, therefore, receives a constantly growing attention in recent years. So far, one of the popular solutions to enhance semantic expressiveness is domain specific modeling based on general metamodeling (GMM) facilities. But for some particular domains, researchers identified the need of deeper semantic expressiveness therefore proposed domain specific metamodeling (DSMM). Hence, this paper aims to explore the underlying methodologies of DSMM for supporting deep semantic composability. Compared with several usual alternatives based on GMM, this paper applies the multi-level metamodeling architecture to create a set of metamodeling primitives using an example named SEvent. In fact, SEvent is a novel formalism that slightly extends Petri net to support continuous states transition and continuous event triggering. As a proof of concept, we concentrate on developing the textual syntax of SEvent and using it to represent torpedo's behaviors.