This study presents a planar particle image velocimetry (PIV) investigation of near-wall velocity deficit mitigation in a compact curved channel using sweeping jet actuation. A compact sweeping jet oscillator was developed based on a simplified fluidic oscillator configuration and integrated into the curved wall region of the channel. The effects of jet incidence angle, actuator placement, array layout, and jet-to-freestream velocity ratio were examined within a fixed streamwise measurement window near the curved wall. The analysis focuses on time-averaged in-plane velocity fields and representative streamwise velocity profiles, with emphasis on changes in the near-wall low-speed region and local streamwise velocity recovery. The results show that sweeping jet actuation can weaken the measured near-wall low-speed region and promote local streamwise velocity recovery within the planar PIV window. Among the tested cases, a jet incidence angle of 60° and actuator placement at the beginning of the bend produced a more balanced local recovery response. The comparison between different array layouts indicates that the observed recovery is affected by the combined influence of actuator spacing, actuator number, and total actuation input; therefore, it should not be interpreted as an isolated spacing effect. The present results provide a systematic low-speed experimental basis for selecting sweeping jet incidence angle, actuator placement, array configuration, and equivalent actuation input in compact curved channel flow control applications. The findings clarify how different actuation parameters affect local near-wall velocity deficit mitigation and streamwise velocity recovery within a controlled optical access configuration.
Self-powered sensor nodes need vibration harvesters that work over a band of frequencies, but the arrays and graded metamaterials that provide such bandwidth need many separately clamped resonators. This numerical study presents a kirigami-graded metabeam in which U-shaped slits in one 250 × 16 × 0.5 mm aluminum sheet release four tabs of graded length (44–62 mm); the sheet is folded and clamped at both ends, and a PZT-5A patch straddling each root feeds its own resistive load. A distributed-parameter cell model is verified against a three-dimensional coupled-field finite-element model within 1% in frequency and 2% in power. In the finite-element model, each cell mode is confined to one tab (off-diagonal amplitudes ≤ 0.09) and the cells deliver 0.59–1.44 µW per (m s−2)2 at 130–274 Hz; their band-mean power is 3.6 times that of a single clamped tab and 0.9 times that of four separately clamped tabs, with two clamps instead of four. Interleaving the cells so that frequency-adjacent cells are not spatial neighbors keeps the modes localized on an 18 Hz grid and gives a continuous 212–284 Hz band with 1.1 times the four-tab power. One root-compliance parameter places 18 cell frequencies within 2–3%, and the sensitivity of the results to the assumed damping is quantified. No measurement is reported; the experimental characterization of the prototype is left to a separate study.
The rapidly exploring random tree star (RRT*) algorithm is widely used for path planning, but finding an initial feasible path can require substantial sampling. Learning-based enhancements commonly predict promising sampling regions from grid maps. This paper presents a model that combines a Tree Long Short-Term Memory encoder with a Transformer decoder to predict an ordered sequence of keypoints from a hierarchical map representation. The resulting TLT-RRT* planner checks the predicted segments, repairs infeasible connections, and continues path optimization using RRT*. The multi-branch tree representation avoids explicitly storing a full high-dimensional grid when the environment admits a compact hierarchical representation. Two- and three-dimensional experiments show shorter initial-path times than the selected RRT-based baselines in the evaluated settings. Six-dimensional manipulator examples and resource benchmarks further characterize the implementation and its computational limits.
This paper addresses the optimal affine formation control problem for second-order multi-agent systems (MASs). The control objective is to steer the formation with stress matrix constraints while minimizing the sum of all the local time-varying cost functions. The key challenge is the tight coupling between position tracking, velocity regulation, and time-varying cost minimization. We develop a distributed estimator–controller framework that decouples these dynamics via three local estimators. For each agent, these estimators can track the global optimization variables using only local information from its neighbors. On this basis, a distributed optimizer is designed for the agents. Theoretical analysis shows that, using the proposed optimizer, all agents’ trajectories converge to the optimal solution by a prescribed time, and affine formation is simultaneously achieved within that horizon. Simulation results further illustrate the superior tracking performance of our algorithm in multi-robot scenarios.
This paper investigates fault-tolerant tracking for continuous-time nonlinear systems subject to input nonaffinity and prescribed state inequality constraints. A slack-function-based transformation and a dynamic precompensator are adopted to handle state constraints and input nonaffinity, respectively. The fault model applies to redundant auxiliary-drive channels of the precompensator and includes stuck, outage, bias, and loss-of-effectiveness modes. Under fault-free operation, an off-policy integral Bellman equation is formed from measured data, and the critic and actor weights are computed by regularized projected least squares. Online, the learned actor is embedded in an adaptive fault-tolerant controller. The stability analysis gives a local forward-invariance condition and retains policy-approximation, regression, and derivative-filter errors in the ultimate bound. Direct closed-loop simulations of a nonaffine continuous stirred tank reactor include single-drive outage, repeated fault switching, and measurement noise. The supplied code and raw trajectories reproduce all comparisons and quantify the regression conditioning and iteration residuals.
