2026-10-14T20:18:25Z https://www.scidb.cn/oai

10.57760/sciencedb.40811 2026-06-23T11:08:03Z

2026-06-23 Simulation Dataset for Fast Cooperative Sparse Control for Underactuated Vessels with Path Following and Rudder Roll Stabilization doi:10.57760/sciencedb.40811 en This dataset originates from a simulation study on fast cooperative sparse control for path following and rudder roll stabilization of underactuated marine vessels. It aims to compare the performance of five controllers in terms of path tracking and roll stabilization, and to evaluate robustness under wave disturbances. All data were generated using scripts written in MATLAB R2020a, employing a 4-degree-of-freedom ship model with state variables including cross-track error, heading angle, roll angle, roll rate, etc., and with rudder angle as the control input. The simulation time step is fixed at 0.1 s, the maximum number of simulation steps is 32900, and the path is obtained by spline interpolation and smoothing of six original waypoints.The five controllers are: a roll-stabilized controller based on quadratic programming (QP); a PPGM controller without roll stabilization; a PPGM roll-stabilized controller with a time-varying L1-weight coefficient; a roll-stabilized controller based on a projection neural network (PNN); and a roll-stabilized controller with a fixed sparse L1 weight. Each controller runs independently until it reaches the destination or the maximum step limit, with the actual number of steps varying depending on the control strategy. The specific step counts are recorded in the time vector of each curve file.After simulation, the original code generates several comparative figures, from which the data of each curve are extracted and saved as independent .mat files, organized by figure order in the Figure_Curves_Data folder. This folder contains the following subdirectories:Figure_04_PathTracking: Stores the east/north position sequences of the five controllers and the original path, as well as the coordinates of the start and end points. Each record is a two-dimensional vector (x, y) with no missing values. The units of both horizontal and vertical coordinates are meters (m), with a spatial range of approximately 0–14000 m (east) and 0–11000 m (north).Figure_05_CrossTrackError: Stores the cross-track error time series of each controller. The variable t is time (s), and y is the cross-track error (m). The length of each curve equals the actual number of simulation steps of that controller (typically 2000–32900 steps) without missing values.Figure_06_HeadingAngle: Stores the time series of the desired heading and the actual heading angles of the five controllers, converted to degrees (°). The variable t is time (s), and y is heading angle (°). The desired heading is taken from the time base of Controller 3. The length of each curve matches the actual number of steps of the corresponding controller.Figure_07_RollAngle: Stores the roll angle time series of the five controllers, in degrees (°), with the same variable definitions as above. Data have no missing values, and the curve lengths are equal to the number of steps of each controller.Figure_08_RollRate: Stores the roll rate time series of the five controllers, in degrees per second (°/s), with the same variable definitions.Figure_09_RollPDF: Stores the kernel density estimation (KDE) results of the roll angles for each controller. Each file contains the abscissa xi_* (degrees) and the probability density f_*, automatically determined by MATLAB's ksdensity function with default bandwidth and number of points. No missing values.Figure_10_WaveSpectralDensity: Contains the ITTC single-parameter wave spectrum data, including the frequency vector omega (rad/s), spectral density S (m²·s/rad), peak frequency omega_peak, and peak spectral density S_peak. The frequency range is 0–5 rad/s with a resolution of 0.01 rad/s, no missing values.Figure_11_SparsityBar: Stores the data for the bar chart of high-frequency control action proportions, including the category labels categories and the corresponding (1 − sparsity) values. Sparsity is defined as the proportion of sampling points where the rudder rate is less than 0.5°/s, with the first point ignored to avoid differencing loss.Figure_12_Radar: Stores the normalized indicator values used in the radar chart. Each controller file contains the angle vector theta (radians), the radius vector rho (dimensionless), and the indicator labels. RadarCommon.mat contains the original performance indicator matrix of all controllers and the common angles. All data are free of missing values.Potential sources of error in the data include: ① approximation errors introduced by linearization and discretization of the ship model; ② convergence tolerances of the proximal gradient method or QP solver in control increment computations; ③ the first-order filter approximation of the steering gear model; ④ the wave disturbance module based on the ITTC spectrum and a simplified second-order transfer function, which may deviate from actual sea conditions; and ⑤ parameter perturbation that only alters rudder effectiveness coefficients, without fully considering the coupling effects of hydrodynamic parameters. None of the simulation data are intentionally incomplete; if a controller terminates early, its record length is shorter than the maximum steps, which is reflected in the file by a correspondingly shortened vector length, not as missing data. The dataset uniformly adopts the .mat format and can be read with MATLAB without the need for additional tools. Underactuated vessels; path following; rudder roll stabilization; sparse control; projected proximal gradient method; cooperative optimization wei li RESTRICTED dataset Science Data Bank