The cooperative operation of multiple Unmanned Aerial Vehicles (multi-UAV) is emerging as a pivotal trend in future complex autonomous systems. To enable accurate mission understanding and efficient collaboration among UAVs in complex, dynamic, and uncertain operational environments, this paper introduces CogMUS, a novel cooperative mission understanding framework based on the Soar cognitive architecture. We first construct a mission understanding framework for UAV operations centered around five typical mission categories. Building on this foundation, we design a distributed cognitive model where each UAV is equipped with a Soar agent. This model leverages the synergy of working memory (WM), long-term memory (LTM), and the decision cycle (DC) to achieve key functionalities, including hierarchical mission decomposition, dynamic task allocation, and proactive airspace conflict detection and resolution. Through comprehensive simulation experiments, we validate the performance of the proposed CogMUS framework across key metrics, including task understanding accuracy, cooperative efficiency, and overall task completion rate. The results demonstrate that CogMUS exhibits superior adaptability to diverse scenarios, as well as remarkable scalability and robustness.

Autonomous dynamic mission planning for a spacecraft performing distributed satellite swarms rendezvous and proximity operations missions is challenging due to complex orbital dynamics and stringent timing constraints. This paper proposes a mission planning method based on a hierarchical framework and a hybrid solution strategy. First, the overall mission execution flow and practical constraints for distributed satellite swarms proximity operations missions are analyzed. An orbital maneuver model is then established based on the Lambert transfer problem within a layered planning architecture. Second, a Genetic Algorithm (GA) optimizes the Lambert transfer velocity increments, thereby constructing a static cost matrix that quantifies the maneuver costs between all target pairs. Next, a hybrid algorithm is applied to rapidly solve the resulting traveling salesman problem (TSP) formulated using the cost matrix, producing an initial target visitation sequence. To address temporal constraints, a combined strategy involving waiting, transfer, and replanning is implemented to refine the initial sequence. This specifically enables sequence correction for targets located in the same orbital plane, yielding a mission execution plan that adheres to timing constraints. Furthermore, the complete maneuver sequence is simulated via numerical orbit propagation using the spacecraft and target orbital parameters. Finally, numerical simulations under various scenarios are conducted to verify the feasibility and robustness of the proposed approach.This study provides a theoretical reference for autonomous satellite swarms proximity operations of spacecrafts.

Modern recommender systems have achieved remarkable success in capturing user interests. However, their heavy reliance on correlations within observational data renders them susceptible to confounding biases, such as item popularity and exposure effects. These biases cause the model to learn spurious correlations between user behaviors and item attributes, thereby compromising fairness, novelty, and the ability to generalize to users' true preferences. To address this fundamental challenge, we propose a novel debiasing framework, Causal-Guided Disentanglement for Debiasing Recommendation (CGDR), which causally separates latent preference factors from confounding effects at the representation level. We first construct a Structural Causal Model (SCM) to explicitly delineate the generative process of user-item interactions, postulating that a user's action is determined by both their intrinsic preference and a set of confounding factors. Guided by this causal graph, our framework employs a novel Causal-Guided mechanism that leverages auxiliary supervisory tasks to provide the disentanglement process with a clear causal direction. This decomposes the latent representation into two independent components: one capturing robust, transferable true preferences, and the other explicitly modeling dynamic confounding effects. Upon achieving this disentanglement, we generate debiased recommendations via Counterfactual Reasoning, performing interventions on the confounder representation to simulate a bias-free environment while leveraging the preference representation for prediction. Extensive experiments on multiple real-world datasets demonstrate that our proposed model not only significantly outperforms state-of-the-art baselines in recommendation accuracy, particularly in enhancing long-tail item performance, but also exhibits superior robustness and interpretability.