adaptive valve-sequence planning

This study addresses the limited adaptability of conventional ship ballast water systems under hydraulic anomalies—such as valve failures or pipe blockages—and their heavy reliance on dense sensor arrays for fault diagnosis. The authors propose a novel approach that integrates graph theory with deep reinforcement learning, modeling ballast routing as a set of 54 feasible fluid transfer paths. By employing frame-stacked water level observations and action outcomes to approximate a partially observable environment, the method incorporates failure-action memory and dynamic action masking to enable adaptive rerouting. Notably, it implicitly infers blockage states without explicit high-dimensional POMDP modeling and introduces a fault-history scoring mechanism reliant only on sparse sensing to rank suspect components. Experimental results demonstrate 100% task success across all single-point blockage scenarios, reducing average decision steps from 61.0 to 41.5; the fault-scoring mechanism achieves 100% top-3 hit rate, with strict and inclusive top-1 hit rates of 66.7% and 83.3%, respectively.

This work addresses the problem of feedback motion planning for continuous-time stochastic nonlinear systems under Signal Temporal Logic (STL) specifications by proposing a novel framework that integrates predicate erosion with probabilistic reachable tubes. Predicate erosion is employed to transform stochastic STL constraints into tightened deterministic ones, while probabilistic reachable tubes quantify the deviation of stochastic trajectories from their nominal counterparts. Leveraging contraction theory, a tracking controller is designed to establish a closed-loop planning pipeline. The proposed approach significantly reduces the conservatism inherent in conventional methods, achieving high STL satisfaction probability without compromising planning performance. Simulations and real-world experiments on a quadrupedal robot demonstrate that the method outperforms baseline approaches in both STL satisfaction rate and computational efficiency.

Large-scale pneumatic soft robots lack scalable, high-precision real-time pressure control systems and dynamic modeling tools suitable for real-time control. Method: This paper introduces PneuDrive—a modular embedded pressure control system—and the first real-time-control-oriented tri-model dynamic modeling toolkit. PneuDrive features a novel scalable RS-485 bus architecture enabling multi-node daisy-chaining, closed-loop control of 16 valves (0–100 psig), and reliable communication over distances exceeding 10 meters. The modeling toolkit integrates data-driven, physics-based, and hybrid models, supporting hysteresis compensation, fluid–structure interaction modeling, and quantitative performance benchmarking. Contribution/Results: Evaluated on a three-segment continuum robot, the system achieves coordinated trajectory tracking across 12 actuation channels. All three model types are experimentally calibrated and validated via real-time simulation, establishing both a hardware platform and a modeling paradigm for real-time control of pneumatic soft robots.

This work addresses discrete-time nonlinear optimal control problems by unifying classical algorithms—including gradient descent, Gauss–Newton, Newton’s method, and differential dynamic programming (DDP)—within a differentiable programming framework. Methodologically, it introduces the first modular, end-to-end differentiable algorithm template library built upon linear/quadratic approximations (e.g., LQR), enabled by automatic differentiation. Theoretically, it provides a unified derivation of computational complexity and sufficient optimality conditions across all methods. Practically, it incorporates adaptive line search and regularization strategies, and validates efficacy on benchmark tasks such as autonomous racing with a bicycle model. All implementations are open-sourced, demonstrating both efficient gradient propagation and strong generalization across diverse control problems.

This study addresses the challenges of control design in complex industrial processes characterized by multivariable coupled dynamics by proposing an automated control strategy generation framework that integrates large language models (LLMs) with Bayesian optimization. The approach decomposes control design into structured code generation steps, ensuring physical consistency through execution-based validation and feedback-driven repair. It pioneers the automatic synthesis of decentralized PI controller architectures and their tuning environments directly from dynamic process models. Evaluated on a nonlinear gas preheater benchmark, the generated control schemes—subsequently refined via Bayesian optimization—achieve a 26.5% improvement in closed-loop performance and significantly enhance the transient response of pressure loops, thereby demonstrating the method’s effectiveness and novelty.

Active flow control remains highly challenging due to its nonlinear dynamics, partial observability, and high computational cost; existing deep reinforcement learning approaches rely heavily on extensive simulation and often lack interpretability. This work proposes a novel framework that introduces a code-generating agent to iteratively propose, evaluate, and refine explicit, executable feedback control laws under a constrained heuristic learning protocol within a public benchmark interface, thereby replacing conventional neural network optimization. The method matches or outperforms the strongest deep reinforcement learning baselines on 10 out of 13 benchmarks, yielding compact, physically interpretable controllers that demonstrate strong generalization and transfer capabilities across varying Reynolds numbers, Rayleigh numbers, actuator configurations, and sparse observation conditions.

This work addresses the robust satisfaction of Signal Temporal Logic (STL) specifications under tracking errors and model mismatch by proposing a unified planning-and-control framework. The approach uniquely translates STL specifications into time-varying convex sets in configuration space and embeds them within a Graph of Convex Sets (GCS) framework for trajectory planning. Continuous-time constraint satisfaction is achieved through B-spline parameterization, while a feedback controller is designed to prioritize specification compliance during execution. By employing a shared convex-set representation across both planning and control layers, the method enhances system consistency and robustness. Simulations and real-world experiments on a space robot demonstrate that the proposed framework generates smooth, collision-free trajectories that robustly satisfy STL specifications even in the presence of disturbances.

This work addresses the challenge of safe navigation in dynamic environments for systems with unknown dynamics and actuator input constraints. It proposes a novel real-time control framework that, for the first time, explicitly embeds input constraints into the design of an extended spatio-temporal tube (STT). By integrating finite-time reachability analysis with a control authority matching mechanism, the approach rigorously guarantees that Euler–Lagrange systems satisfy reachability, obstacle avoidance, and dwell-time specifications within a finite horizon—without requiring an explicit system model or online optimization—while always respecting actuator limits. The method features offline-verifiable feasibility conditions and has been validated through simulations on mobile robots, quadrotors, and spacecraft, as well as hardware experiments demonstrating safe, constraint-compliant navigation.