This paper presents a hybrid optimization methodology for parameter estimation of reactive transport systems. Using reduced-order advection-diffusion-reaction (ADR) models, the computational requirements of global optimization with dynamic PDE constraints are addressed by combining metaheuristics with gradient-based optimizers. A case study in preparative liquid chromatography shows that the method achieves superior computational efficiency compared to traditional multi-start methods, demonstrating the potential of hybrid strategies to advance parameter estimation in large-scale, dynamic chemical engineering applications.

High-energy-density batteries are gaining popularity with the rise of electric vehicles (EVs). State-of-charge (SOC) estimation is crucial for informed decision-making, operational safety, and the longevity of these batteries; however, model-based SOC estimation often struggles due to the nonlinear dynamics of batteries, unpredictable measurement noise, and dynamic loading conditions. Model-free data-driven methods provide an alternative solution. However, they often struggle to process a long sequence of temporal dependencies in the input data. Furthermore, uncertainty awareness associated with the estimated SOC is often unavailable. This work presents an uncertainty-aware SOC estimation framework integrating a Gated Recurrent Unit (GRU) network with deep evidential regression. Our framework processes sequential time-series data to estimate SOC and its associated uncertainty in a single forward pass. Validation results on real-world battery cycling datasets show that, for in-distribution data, our method achieves predictive performance comparable to computationally intensive ensemble methods.

The motivation for this paper is the implementation of nonlinear state feedback control, designed based on the continuous-time plant model, in a sampled control loop under relatively slow sampling. In previous work we have shown that using one-step predictions of the target dynamics with higher order integration schemes, together with possibly higher order input shaping, is a simple and effective way to increase the feasible sampling times until performance degradation and instability occur. In this contribution we present a unifying derivation for arbitrary orders of the previously used Lobatto IIIA collocation and Hermite interpolation schemes through the Hermite-Obreschkoff formula. We derive, moreover, an IDA-PBC controller for a magnetic levitation system, which requires a non-constant target interconnection matrix, and show experimental results.

Autonomous Underwater Vehicles (AUVs) have been widely used in diverse marine applications. Although previous studies have investigated various guidance law designs for underactuated AUVs, compensation for the sideslip motion remains a critical and underexplored challenge, particularly in the presence of extreme time-varying external disturbances. Inspired by the nature principle that a system’s position and velocity changes are uniquely determined by its acceleration, we develop a finite-time guidance law design for AUVs depth tracking task. This guidance law executes an online planning-time-based depth-tracking acceleration constrained by the depth-tracking errors of position and velocity. Simulations were conducted to verify that the proposed guidance law could make the depth-tracking motion converge within a given finite-time duration if the state-constrained acceleration was promptly executed. Compared to the selected methods, the proposed method can rapidly and accurately compensate for sideslip effects via the estimation of sideslip angle, resulting in precise depth-tracking performance.