Unmanned surface vehicle (USV) swarms operating in communication-denied maritime environments face degraded formation control when inter-agent state exchange is disrupted. This paper presents a three-layer control architecture integrating (1) LLM-assisted strategic mission planning with formal safety verification, (2) predictive tactical coordination combining physics-based motion extrapolation with online-learned neighbor behavior models, and (3) DMPC + ADMM execution for constrained formation control. The predictive coordination module in simulation-based evaluation across five representative scenarios reduces formation error by 76.0% (under simulation conditions) during 60 s communication outages compared to zero-hold prediction (Cohen d = 0.88, p < 0.001, DMPC + ADMM validated). The adaptive topology manager dynamically selects among star, mesh, and tree configurations via a utility function balancing communication quality, threat exposure, and overhead, reducing communication overhead by 71.7% (for the particular cases investigated) under the evaluated conditions while maintaining formation accuracy. Stability analysis using multiple Lyapunov functions and average dwell time theory guarantees global uniform exponential stability under topology switching, with explicit error bounds under communication denial derived via Gronwall-type arguments. The framework is validated through 1250 simulation trials across five scenarios with rigorous statistical analysis. The three-layer temporal decoupling architecture provides a practical template for safely integrating LLM-assisted planning with real-time multi-agent control in contested environments.

Prescribed performance control (PPC) guarantees that tracking errors evolve within user-defined envelopes, yet the design of the performance certificate $\rho \left(t\right)$ itself has been restricted to hand-tuned parametric forms since the theory’s inception. We establish that the standard exponential family ${\mathcal{F}}_{\exp }$ constitutes a nowhere-dense, three-dimensional submanifold of the infinite-dimensional space of valid certificates, and we prove that any learnable function with architecturally enforced monotonicity and positivity constitutes a valid, stability-preserving performance certificate. Three theorems formalize this principle. Theorem 1 proves that a feedforward neural network with constrained non-positive time-input weights, non-negative cross-channel weights, and Softplus output activation generates a valid certificate for all conditioning inputs. Theorem 2 establishes a modularity property; closed-loop stability is preserved for any certificate in the family whenever the controller renders the transformed error uniformly ultimately bounded, decoupling certificate validity from certificate origin. Theorem 3 proves that every piecewise-monotone decreasing positive certificate can be uniformly approximated by a constrained monotone network for a suitable conditioning vector; this expressivity is achieved jointly over the network parameters $\theta$ and the conditioning vector $\mathit{c}$, with $\mathit{c}$ acting as a continuous (but not dense) semantic-steering mechanism for fixed $\theta$. The theoretical results are illustrated through an analytical comparison of certificate families and a numerical example on a Duffing-type oscillator, illustrating, in the considered example, improved tracking performance under identical Lyapunov stability guarantees.

Unmanned surface vehicle (USV) clusters operating in contested maritime environments face communication degradation from jamming, satellite denial, and partial node loss. Existing countermeasures either react after link failure or rely on a single signal quality indicator. A framework is presented that integrates two complementary mechanisms for resilient USV cluster communication: (1) a multi-indicator communication quality metric $Q_{\mathrm{comm}}$ that fuses signal-to-noise ratio, packet loss rate, latency, and temporal stability into a single scalar; and (2) a multi-modal backup communication chain spanning acoustic modem, optical link, and multi-hop RF(Radio Frequency) relay that provides physical-layer redundancy when primary radio frequency links are degraded. Across five representative failure scenarios simulated on a seven-vehicle cluster with 50 independent runs per configuration, the multi-indicator $Q_{\mathrm{comm}}$ metric achieves a mean of 0.521, outperforming single-indicator baselines on the composite $Q_{\mathrm{comm}}$ metric (Cohen’s d = 8.020, p < 0.0001, n = 250 per method); $Q_{\mathrm{comm}}$ is the framework’s own optimization target; this comparison is, therefore, presented as an internal consistency demonstration rather than an independent validation. The framework reduces recovery time from 113.8 s (single-indicator) to 15.3 s—an 86.6% improvement. With the backup communication chain enabled, delivery rates exceed 94% across all methods and scenarios. A direct Monte Carlo ablation (50 seeds × 5 scenarios) reveals that multi-indicator fusion is the primary driver of assessment accuracy and recovery speed, while the multi-modal backup chain is the dominant delivery driver: without it, delivery falls from 94.6% to 24.2%. A sigmoid-blended topology utility function is included as an architectural design component; its independent validation requires extended-duration threat experiments identified as future work. The contributions of this work are the validated multi-indicator fusion metric and the multi-modal backup chain architecture, which together provide a practical foundation for resilient USV communication.

Distributed model predictive control (MPC) coordinates constrained multi-agent systems while preserving computational locality, yet an implementation can cease to represent its stated convex program. We study a minimal linear plant so communication denial is a modeled source of uncertainty, and treat centralized equivalence as a hard gate: an edge-splitting consensus alternating direction method of multipliers (ADMM) formulation of the formation-tracking quadratic program is built so every distributed update can be audited against a centralized solver. Against a Lagrangian dual-decomposition baseline, ADMM attains the same accuracy without step-size tuning. Across topologies, horizons, penalties, budgets, and denial windows, the exact branch matches the centralized optimum to numerical precision, the equivalence gate and saturation signature survive scaling to eight agents, and the saturation signature also survives measurement noise. For communication denial, we prove a bounded-error result separating the summable-error regime, which recovers the optimum exactly, from the merely-bounded regime, whose limit points remain O(ε)-optimal. Coupling each edge weight to message age through an exponentially decaying confidence factor contracts the distributed problem to uncoupled local MPC, so formation error saturates as outages lengthen; a fixed-weight ablation isolates this decay as the causal mechanism. We contribute a reproducible baseline and two honest mechanisms—centralized-equivalence gating and confidence-decayed coupling—for distributed formation MPC.

Formation control of unmanned surface vehicles (USVs) in complex marine environments is required to contend with strongly coupled, high-dimensional disturbances. A Multi-Agent Active Disturbance Rejection Control (MAADRC) framework is developed for this purpose. The design centers on a distributed extended state observer (DESO) coupled with a dual-channel feedback structure—NEFL-GCO and LGL-FC—that collectively maintains formation geometry. Three main ideas underpin the approach. First, a bandwidth-efficient distributed observation scheme enables agents to share disturbance estimates while using substantially less communication bandwidth. Second, an adaptive consensus compensation mechanism accommodates parameter variations as formations evolve. Third, a formation-compatible obstacle avoidance algorithm enhances reliability in congested waters. To evaluate the control structure and optimize its parameters, a multi-agent reinforcement learning (MARL) method—specifically Multi-Agent Proximal Policy Optimization (MAPPO)—is employed. The MARL agent tunes two critical parameters: observer bandwidth and nonlinear feedback gain, thereby establishing a performance baseline. After ten million training steps, the MAPPO-optimized MAADRC achieves a tracking root-mean-square error (RMSE) of 1.18 m. This value lies within 3% of the manually tuned result of 1.21 m, indicating that the bandwidth parameterization is near-optimal. Extensive simulations incorporating realistic wind, wave and current disturbances demonstrate a dynamic obstacle avoidance success rate maintaining an expected level, alongside consistently low formation tracking errors. Collectively, these findings confirm the resilience and practical utility of the proposed framework in demanding maritime settings.

Unmanned Surface Vehicle (USV) clusters operating in maritime environments face dynamic communication conditions, including varying sea states, electromagnetic interference, and satellite denial, that render static communication topologies suboptimal. Existing approaches assess link quality through single indicators, typically the SNR, and lack mechanisms for automatic topology adaptation. This paper presents a multi-layer adaptive communication framework that achieves a mean communication quality score of 0.72 (vs. 0.51–0.66 for baselines), a message delivery rate of 94.1% under benign conditions, and a failure recovery time of 3.2 s (vs. 5.8–8.4 s for baselines) across five communication failure scenarios. The framework integrates three layers: a weighted multi-indicator communication quality metric fusing the SNR, packet loss rate, latency, and link stability into a unified score; a topology utility function that selects among centralized, distributed, and hierarchical topologies by optimizing a quality–threat–overhead objective; and a multi-modal backup communication manager with physics-based underwater acoustic, optical line-of-sight, and multi-hop relay fallback modes. Simulation results demonstrate consistent improvements over single-indicator and static-topology baselines, with particularly strong performance under satellite denial and jamming scenarios where multi-modal backup communication sustains delivery rates above 85% under simulated conditions. In summary, the framework demonstrates consistent improvements across all metrics (communication quality, delivery rate, recovery time) relative to four baselines, with the largest gains observed under the most challenging conditions (satellite denial and jamming). We emphasize that the framework adaptively selects among pre-defined canonical topologies (star, mesh, tree) based on real-time conditions rather than synthesizing optimal topologies de novo—a distinction between topology management and topology optimization.