Unmanned Systems

In this paper, a hexarotor Unmanned Aerial Vehicle (UAV) flight dynamics is derived and its control system is designed. A complete mathematical model of the UAV is obtained using Newton–Euler formulation and it is used for simulation. However, a reduced model is derived for the control design purpose which is based on the backstepping approach for hexarotor attitude stabilization and altitude trajectory tracking. The main contribution of this paper consists of the introduction of a new compact Gravitational Search Algorithm (cGSA) and its application for improving the backstepping controller performances in terms of tracking errors. Indeed, using the compact optimization paradigm enhances the performances of classical Population-based Algorithms (PBAs). Moreover, the introduced cGSA is applied to compute the optimal gains of the considered controller. In addition, the cGSA is compared with two compact optimization algorithms that are compact Particle Swarm Optimization (cPSO) and compact Teaching Learning-Based Optimization (cTLBO). The proposed algorithm shows encouraging results.

Aiming at the problems of strong sampling randomness, difficulty in collision detection and rough paths of rapidly exploring random tree (RRT) algorithm in collaborative motion planning of dual robotic arms, we propose an RRT algorithm based on split sampling space. First, a split sampling space strategy is proposed. According to the sampling points having a fixed range in a certain axis degree, the step size of random tree generation is restricted to the respective sampling space, and combined with the hierarchical wraparound box method to achieve effective collision detection. Besides, a greedy strategy is used to speed up the growth of random trees in the respective space. Finally, the trajectory smoothing of the dual robotic arms path using the Bezier curve improves the trajectory quality while ensuring that the dual robotic arms will not collide. The feasibility and effectiveness of the algorithm are verified through simulation experiments as well as real UR experiments.

This paper presents a complete system for autonomous drone racing combining image recognition, depth mapping, visual-inertial odometry (VIO), and collision-free trajectory planning. The proposed system focuses on simple, robust, and computationally efficient techniques to enable onboard hardware applications. A loosely coupled visual-inertial localization system is devised, to ensure real-time and robust localization. A lightweight CPU-based detection module is designed, which consists of autonomous mapping and gate detection components. We also introduce a robust and efficient trajectory planner to generate smooth and collision-free trajectories in real-time. The proposed methods are tested extensively through benchmark comparisons and challenging indoor flights, while simulation results show the validness and effectiveness of our proposed system. We release our implementation as an open-source ROS-package.

This paper presents a reliable variable pitch propeller (VPP) quadcopter with a cambered airfoil propeller called Heli-quad that achieves full-attitude control under a complete failure of one actuator. The idea of employing a cambered airfoil in the propeller blade plays a pivotal role in the full attitude control under the failure of an actuator. Experimental data shows that the cambered airfoil propellers generate significantly higher torque than symmetric airfoil propellers, enabling yaw control even under a complete failure of an actuator. The theoretical analysis clearly indicates that Heli-Quad with three actuators is sufficient to provide full-attitude control. The proposed unified fault-tolerant controller consists of a outer loop position tracking controller, a proportional-derivative inner loop attitude controller, and a novel neural-network-based reconfigurable control allocation scheme that computes the actuator commands. Experimentally validated propeller aerodynamic data has been used to train the neural network. High-fidelity software-in-the-loop simulations using the SIMSCAPE environment are carried out to analyze the Heli-quad’s performance. From the empirical result, the maximum tolerable delay in Fault Detection and Isolation (FDI) is 180 ms. The results indicate that even under the complete failure of one actuator, the position tracking performance of the Heli-quad is closer to nominal conditions.

This paper proposes robust controllers for a class of unmanned tracked ground vehicles (UTGVs), which are built to autonomously clean carryback or spillage from the conveyor belts used in the mining industry. The UTGV, a nonholonomic system in its nature, needs to follow a given path in a harsh environment with large uncertainties due to the time-varying mass and inertia when the UTGV loads and unloads as well as unknown frictions and flatness of the ground. Moreover, the input constraints coming from motors do exist. It is usually hard to design robust controllers for such complex systems. By utilizing the available autonomous driving system, which is designed to be compatible with the existing remote motion controller in unmanned systems to generate autonomous ability, this paper uses the off-the-shelf motion planner to calculate desired linear and angular velocities based on the given path and sensor perceptions. Consequently, the control design can be simplified as two decoupled linear time-invariant scalar dynamic systems with uncertainties, making the active disturbance rejection controller (ADRC) applicable. By carefully designing the parameters of ADRC with the help of an extended state observer (ESO), it is shown that the proposed ADRC and ESO can achieve good tracking performance in the presence of input saturation and can handle nonsmooth disturbances. The proposed simulation results and experimental results support the theoretical findings.

Over the last decade, the focus on the use of computer tools to assess flows around submarines and guide their design has been considerably developed. In particular, with the emergence of parallel computational capabilities, RANS simulations of viscous flows have seen a greater role in predicting these flow fields. The shift to a much more CFD-based design and analysis approach leads to the ability to achieve better designs in the shortest time. This makes it possible to classify several ranges of designs as well as to provide an entire image of the flow field, which can lead to a better understanding of the flow field’s physics. The aim of this study consists of a 3D numerical simulation of the turbulent flow around a submarine without and with appendages to quantify the effect of each appendage on total drag and to study the interaction between the submarine hull and its appendages. The hydrodynamic turbulent flow behavior around the well-defined shape is described using the ANSYS CFX code to resolve the RANS governing equations. Good agreement is obtained from the validation of the numerical results confirming the efficiency of the method in terms of computational time and robustness. The numerical approach is adopted to predict the flow field and forces and moments acting on the underwater vehicle for different maneuvering cases and simulation results have been presented.

Relative measurements play an important role in cooperative localization since they combine the observation from other agents and improve the state estimation accuracy. Due to the presence of uncertainty from agents’ location information, the model of relative measurements has to be modified to take the uncertainty into account. To this end, we present an equivalent relative observation model based on the unscented transformation to incorporate the relative measurements. The model enables each agent with the relative measurement from its neighboring agents to contribute to the estimation performance. In particular, the scheme of relative measurements is able to handle the outlier embedded in each agent’s measurement from environments, which prevents the estimate from being unbounded in this case. Meanwhile, we present two update schemes to incorporate the innovation information from the relative observations. One scheme absorbs the relative measurement after the update with anchor nodes while the other scheme utilizes the relative measurement in the sense of track-to-track fusion via a consistent fusion, which guarantees the consistency of the estimate. A simulation with four robots demonstrates that the performance of the proposed algorithm is superior to other conventional approaches, e.g. EKF without relative measurements.

This paper proposes a water surface garbage cleaning vehicle. Its main functions include the detection, tracking and collection of garbage. The object detection algorithm based on the YOLOv3 network is applied to embedded systems. The sliding mode control is designed to guarantee the stability of the tracking process of the unmanned vehicle. Experiments and comparative simulation results are given to show the effectiveness of the proposed solution.

This paper addresses two substantial aspects in the field of robotics: trajectory planning and trajectory tracking for multi-robot systems. The collective movement of the multi-wheeled mobile robots is based on a formation control with a leader–follower strategy. To ensure safe navigation toward the destination within a workspace containing multiple obstacles, we skillfully combine the Modified Artificial Potential Field (MAPF) method and the Improved Particle Swarm Optimization (IPSO) algorithm for a group of nonholonomic mobile robots. A global planner based on IPSO algorithm is assigned to the leader robot, to find the optimal collision-free path. Compared to recent nature-inspired methodologies, the improvement suggested enhance the search capabilities of the algorithm, resulting in a 2% reduction in path length and a 29% decrease in computation time. Simultaneously, the follower robot has the ability to employ either a formation policy or a local planner using MAPF. The proposed modifications address the primary limitations of the conventional method, notably the susceptibility to local minima and the challenge of reaching goals near obstacles. Furthermore, this local planner is adept at evading both static and dynamic obstacles. Extensive real hardware experiments are conducted within multiple scenarios using the mobile robot Qbot-2e to corroborate the obtained simulation results and validate the effectiveness of the proposed approaches both in tracking the desired trajectories, finding the optimal feasible collision-free path and avoiding static and dynamic obstacles for multi-robot systems.

Aerial object detection is crucial in various computer vision tasks, including video monitoring, early warning systems, and visual tracking. While current methods can accurately detect normal-sized objects, they face challenges distinguishing small objects from cluttered backgrounds. Developing methods that can be deployed on edge devices to achieve fast, accurate, and energy-efficient performance is also an urgent challenge. This paper proposes a network for aerial object detection by incorporating an attention mechanism to enhance feature extraction and elevate the accuracy of aerial moving object detection. Additionally, we optimize the channel dimensions of the feature extraction framework, resulting in a reduction in model parameters, acceleration of inference speed, and alleviation of the computational burden. Ulteriorly, we optimize the Spatial Pyramid Pooling (SPP) module to enhance detection accuracy and processing speed. Inspired by the ResNet and RepVGG structures, we design a feature fusion module to combine early-extracted features, improving speed and accuracy. Based on the design mentioned above principles, we develop a neural network method with an impressively small model size of only $4.5$M. The proposed approach achieves state-of-the-art performance on five benchmark datasets. Besides its superior performance, our method demonstrates excellent throughput on edge computing devices. Experimental results show that even when running on low-performance computing devices, the CPU and GPU temperatures remain below 50^∘C and achieve a detection speed of 14.8 frames per second (fps) and power consumption of only 2.9W. These findings suggest that a high-accuracy, low-power, low-latency, and low-memory footprint aerial object detection solution is achievable.