To address the challenges of heavy data processing volume and the difficulty in meeting real-time requirements for industrial applications in 3D point cloud–based manipulator grasping, this paper proposes a novel visual grasping method based on negative space analysis of point cloud bird's-eye view (BEV). First, the YOLOv8 network is employed to perform fast and accurate 2D localization of targets in RGB images, and a 3D frustum is constructed to preliminarily filter the scene point cloud, followed by the random sample consensus (RANSAC) algorithm to robustly segment the desktop support plane. The core innovation involves a geometric manifold projection strategy that reduces the dimensionality of sparse 3D point clouds onto a 2D BEV plane. Based on the theory of image moments, the contour of the "negative space" occupied by the object is analytically parsed, thereby solving the target's six-degree-of-freedom (6-DoF) grasping pose with a linear computational complexity of O(N). Experimental results demonstrate that, compared with the baseline method combining single-shot multiBox detector (SSD) and PointNetGPD, the proposed method achieves a 5% improvement in the total system success rate (rising from 65% to 70%). Moreover, the average computation time per grasp is significantly reduced from 550 to 210 ms, exhibiting a speed advantage of more than 2.6 times. This work verifies the feasibility of replacing complex 3D deep-learning models with lightweight geometric analysis in specific structured scenes.
Numerous studies have appeared in the control literature regarding the regulation of the angular speed of DC motors driven by DC-DC converters. In this work, we considered a buck converter-driven DC motor system. The implementation of the proposed adaptive controller requires only one sensor to measure the motor angular speed. An estimator was developed to estimate the inductor current and the output voltage of the converter, the armature current of the DC motor, and to generate a simple adaptation law for the estimate of the external load torque. A linear sliding surface was used to derive an adaptive controller that is simple in its design and guarantees closed-loop system stability and precise speed regulation in the presence of the inductor parasitic and parameter uncertainties of the converter/motor system. Furthermore, the controller was shown to be robust against external disturbances. A simple systematic procedure was outlined to select the three gains of the proposed adaptive controller. The robustness and effectiveness of the controller were validated by computer simulations and compared to the proportional integral derivative (PID) control.
In three-phase three-wire power systems, unified power quality conditioners (UPQCs) compensate for current/voltage fluctuations while enhancing load-side power quality. However, current solutions face three key limitations: (1) limited responsiveness to rapid disturbances, (2) sensitivity to grid variations, and (3) inadequate phase-locked loop (PLL) performance, all of which undermine compensation effectiveness. Traditional proportional-integral (PI) controllers further exacerbate these issues through overshooting during control quantity disturbances. This paper proposes an enhanced finite control set model predictive control (FCS-MPC) system integrated with linear active disturbance rejection control (LADRC) to improve UPQC's anti-disturbance capabilities. The architecture combines a series active power filter (APF)-side voltage compensation module based on FCS-MPC, a shunt APF-side current compensation module based on FCS-MPC, and a LADRC-optimized mixed second/third-order generalized integrator-based PLL (MSTOGI-PLL). Through rigorous MATLAB/Simulink simulations, the proposed UPQC demonstrates superior robustness compared to conventional controllers, achieving 85% voltage sag compensation within 12.5 ms response time. Simulation result validation confirms that the LADRC-enhanced FCS-MPC system significantly improves disturbance rejection, yielding effectively reduced voltage total harmonic distortion (THD) and enhanced post-compensation power quality across various grid anomalies.
This paper proposes and evaluates a triangular formation control method for unmanned aerial vehicles (UAVs) based on a leader-follower structure using a Mamdani fuzzy PID controller on a fully nonlinear 6-DOF dynamic model. The controller was designed with a fuzzy inference mechanism to adjust PID gains via a rule-based, nonlinear fuzzy gain-scheduling mechanism based on state deviations, thereby improving trajectory tracking quality under nonlinear conditions and disturbances. The system's performance was evaluated through transient metrics and energy error indicators (MSE, RMSE) in three scenarios: no noise, external noise, and external noise combined with wind disturbance. Simulation results in MATLAB/Simulink showed that under noise-free conditions, the fuzzy PID achieves a rise time of 2.03 s, a percentage overshoot of 50.30%, and a steady-state error of 0.0315 m, significantly improving upon the PID (3.29 s; 55.74%; 0.0365 m). When external noise is present, the overshoot decreases from 227.50% to 88.58%, and the steady-state error decreases from 1.0272 to 0.4060 m (approximately 60.5% improvement). In conditions with combined wind disturbance, MSE decreases by up to 87.49% along the z-axis, and RMSE decreases from 1.1538 to 0.4081, demonstrating superior disturbance rejection. Quantitative results confirm that fuzzy PID significantly enhances trajectory tracking accuracy, stability, and robustness of the multi-agent UAV system in uncertain environments.
Automatic Modulation Classification (AMC) is a significant decision-making process in non-cooperative, 5G, and beyond communication systems. Advancements in Artificial Intelligence (AI) led to the implementation of Deep Learning (DL) to provide superior performance over the feature extraction and offline training process of AMC. In this work, we proposed a hybrid modulation classification architecture by integrating Convolutional Neural Networks (CNN) with ML classifiers such as Random Forest (RF), Support Vector Machine (SVM), and Extreme Gradient Boosting (XGBoost). The Radio Frequency Signal Classification (RFSC) dataset consists of data collected under different Signal-to-Noise Ratio (SNR) scenarios to analyze the resilience of the classifiers. Among all architectures, the Deep Convolutional Layer based RF (D-CNL-R) model achieved superior modulation recognition accuracy due to its enhanced capability to learn complex nonlinear feature distributions. We observed that the training overhead of the proposed D-CNL-R reduced to ≃ 1× with better classification accuracy performance. We also presented an experimental approach for the prediction performance of real-time signals for indoor and outdoor scenarios.
The development of robot technology has received increased interest in three-wheeled mobile robots (TWMR) due to their flexibility and adaptability in various applications. However, due to the limitation of independent motion in all directions, highly nonlinear dynamics, and effects such as noise and wheel slippage, it is difficult to effectively control the motion of TWMR. In this paper, we studied the motion of TWMR through kinematic and dynamic models considering the phenomenon of wheel slippage. The lateral and longitudinal slippage components were added to the dynamic model, thereby designing the position control law using the backstepping method and the speed control law using the SMC method. We also adjusted the parameters of the controllers to help the robot perform trajectory tracking. In addition, controlling the input torque was performed to overcome the lateral and longitudinal slippage phenomena. The method was tested through simulations on MATLAB Simulink with different trajectory cases (circular trajectory, infinite trajectory, spiral trajectory). The results showed the suitability of the controller to the desired trajectory tracking problem of TWMR, and the trajectory tracking error converged to zero. This shows promise and is practical.
Vehicle state estimation accuracy in the Adaptive Unscented Kalman Filter (AUKF) is highly dependent on the proper tuning of the process and measurement noise covariance matrices. Conventional optimization techniques often produce suboptimal covariance values, increasing the risk of estimation divergence and compromising vehicle stability. To overcome this limitation, this paper proposes a Modified Dragonfly Optimization Algorithm-based Adaptive Unscented Kalman Filter (MDOA-AUKF), in which adaptive covariance tuning is formulated as an optimization problem. By incorporating elite solution preservation and multi-behavioral swarm dynamics, the proposed MDOA effectively balances exploration and exploitation, thereby enabling adaptive updates of the process (Q) and measurement (R) noise covariance matrices at each iteration. This improves the convergence, estimation accuracy, and robustness under nonlinear driving conditions and varying vehicle mass. The proposed approach is implemented using a two-degree-of-freedom bicycle model considering lateral and yaw dynamics under a low-friction road condition (tire–road friction coefficient of 0.4) and evaluated using Fishhook (FH) and Double Lane Change (DLC) maneuvers. During the FH maneuver, the proposed method achieves root mean squared error (RMSE)/ normalized root mean squared error (NRMSE) values of 0.0008/0.21% and 0.0010/0.03% for the sideslip angle and yaw rate, respectively, while the corresponding values during the DLC maneuver are 0.0011/0.24% and 0.0009/0.02%, respectively. Compared with Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), and Grey Wolf Optimization (GWO), the proposed approach provides significantly lower estimation errors and faster convergence. Robustness is further validated under vehicle mass variations from 1800 kg to 2300 kg, demonstrating consistently low estimation errors. These results confirm that the proposed MDOA-AUKF is an accurate, robust, and computationally efficient framework for intelligent vehicle state estimation.