Recently, automation using Aerial Manipulators (AM) plays a vital role in the consumer centric industries where tasks need to be accurate with stability. Rotorcraft aerial manipulators are hybrid robotic systems composed of multi-rotor unmanned aerial vehicles (UAVs) and robotic arms. They possess advantages such as high-altitude flexible maneuverability and interactive operation, rendering significant application value in fields like high-altitude maintenance, hazardous material transportation, and emergency rescue. However, the strong nonlinear force-position coupling between the UAV and the robotic arm during operation leads to unstable flight attitudes and increased positioning errors at the end-effector of the robotic arm. Traditional PID control methods struggle to cope with such complex coupling characteristics. To address this critical issue, this paper first reveals the chain coupling mechanism of "centroid offset-inertial transfer-aerodynamic disturbance" for rotorcraft aerial manipulators from three dimensions combined with dynamic simulations. On this basis, a multi-field coupling dynamic equation covering centroid offset, inertial coupling, and aerodynamic disturbance is innovatively established, providing a theoretical foundation for the design of control strategies. Furthermore, an adaptive PID control strategy based on multi-agent reinforcement learning (MAAC-PID) is proposed, which innovatively adopts a "centralized learning-distributed execution" architecture. Through two agents optimizing the UAV's PID parameters and the robotic arm's joint control respectively, global cooperative optimization of the coupled system is achieved. Simulation results show the MAAC-PID algorithm outperforms traditional PID and PPO, reducing trajectory errors and enhancing stability. Specifically, under 2KG load condition, the standard deviation of the error of the MAAC-PID algorithm in the square trajectory is reduced by approximately 16.1% compared with the PPO algorithm; in addition, the MAAC-PID algorithm achieves an improvement of 0.03m in average maximum trajectory error compared with the PPO algorithm, exhibiting stronger robustness and adaptability.

Dual-arm robot manipulation tasks are highly complex, and different tasks can be categorized as dual-arm synchronous coordination or asynchronous sequential execution. Existing action-generative policies adopt a fully parameter-sharing single-network architecture in multi-task learning, overlooking the differences between these two task types in terms of temporal dependencies and action distributions, which leads to cross-mode interference. To address this limitation, we propose LGDM (Language-Guided Dual-Mode Policy), a language-guided dual-mode policy framework. Under shared language embeddings, it parallelly constructs dual-conditional branches for language routing and visual perception. The execution layer selects independent synchronous or asynchronous policy branches according to routing signals and outputs actions by combining them with perceptual features, thereby explicitly decoupling the two motion modes. Specifically, the model organizes three functional modules around language embeddings as a hub: (1) a language mode routing layer that predicts motion modes from temporal information in semantics and provides mode-selection signals to the execution layer; (2) a vision perception layer that injects semantic information into the visual backbone via FiLM, enabling task-aware dynamic feature modulation; and (3) a dual-mode execution layer that builds independent synchronous and asynchronous policy branches, selects the corresponding branch according to routing, and generates actions with the modulated features. On the RoboTwin 2 dual-arm manipulation benchmark comprising six synchronous and asynchronous tasks, LGDM outperforms existing baselines, achieving an absolute improvement of 14.5 percentage points over RDT, and maintains robust execution performance under unseen temporal-coordination instruction variants at inference time.

To address the challenges of inadequate deep multimodal temporal features during UAV navigation, slow convergence of temporal fusion models, and difficulty in capturing diverse trajectories with traditional serial convolution kernels, this paper proposes an end-to-end UAV navigation algorithm model for unknown complex environments. First, a feature pyramid architecture is embedded in the backbone network of the feature extraction stage; leveraging its hierarchical characteristics, it can effectively extract temporal information at different depths. Second, to overcome the receptive field limitation of traditional serial Conv1D, a multi-scale trajectory generation head is constructed, integrating parallel convolution, channel attention, and residual connections. To verify the effectiveness of the algorithm model, forest and cluttered scenes with different densities were built in a simulation environment. The results show that the success rate of the proposed framework increased by 16% and 26% respectively under the flight scenario of $\mathbf{7 m} / \mathbf{s}$.

As drone control and communication technologies continue to mature, the use of drones equipped with cameras, robotic arms, pods, and other devices for multi-tasking and multi-scenario applications is becoming increasingly prevalent. Rotorcraft aerial manipulators have the potential to perform complex aerial operations such as grasping and transporting, offering wide-ranging applications. However, the complex coupling between the drone and the manipulator affects the stability and precision of the rotorcraft aerial manipulator. This paper aims to tackle the complex system dynamics and coupling issues that traditional control methods struggle to handle by leveraging reinforcement learning algorithms. By comparatively analyzing the performance of different reinforcement learning algorithms in rotorcraft aerial manipulator grasping tasks, a multi-objective TD3 reinforcement learning algorithm is proposed. Simulation experiments demonstrate the effectiveness and stability of this algorithm in the rotorcraft aerial manipulator system.

Pre-trained models trained with internet-scale data have achieved significant improvements in perception, interaction, and reasoning. Using them as the basis of embodied grasping methods has greatly promoted the development of robotics applications. In this paper, we provide a comprehensive review of the latest developments in this field. First, we summarize the embodied foundations, including cutting-edge embodied robots, simulation platforms, publicly available datasets, and data acquisition methods, to fully understand the research focus. Then, the embodied algorithms are introduced, starting from pre-trained models, with three main research goals: (1) embodied perception, using data captured by visual sensors to perform point cloud extraction or 3D reconstruction, combined with pre-trained models, to understand the target object and external environment and directly predict the execution of actions; (2) embodied strategy: In imitation learning, the pre-trained model is used to enhance data or as a feature extractor to enhance the generalization ability of the model. In reinforcement learning, the pre-trained model is used to obtain the optimal reward function, which improves the learning efficiency and ability of reinforcement learning; (3) embodied agent: The pre-trained model adopts hierarchical or holistic execution to achieve end-to-end robot control. Finally, the challenges of the current research are summarized, and a perspective on feasible technical routes is provided.

This paper proposes an end-to-end high-speed autonomous navigation framework for quadrotor UAVs, with innovations aimed at addressing the limitations of the original ResNet architecture: 1) Replacing the original ResNet backbone with the EfficientNet backbone, which features better parameter efficiency and stronger feature extraction capability under similar parameter counts; 2) Adding a Feature Pyramid Network (FPN) architecture based on EfficientNet-B0 to effectively fuse multi-scale features and enhance the perception ability of objects of different sizes; 3) Introducing spatial attention and channel attention mechanisms–the former enables the network to focus on key regions of images and improves obstacle detection accuracy, while the latter enhances the response of important feature channels and suppresses interference from irrelevant features. Experiments show that the improved network achieves lower trajectory loss and higher inference speed in trajectory planning tasks under highspeed flight scenarios of $7 \mathrm{m} / \mathrm{s}$, with stronger robustness and more precise obstacle avoidance in complex environments. Compared with the original architecture, the trajectory loss is reduced by 5.2% and the score loss by 1.2%, which verifies the rationality and effectiveness of the improvements. The core innovation lies in the comprehensive improvement of autonomous trajectory planning performance through the collaborative optimization of efficient backbone replacement, multi-scale feature fusion, and the introduction of attention mechanisms.

Aerial manipulators expand the application scenarios of manipulators into the air. To complete various operations, the contact force between the aerial manipulator and the target must be precisely controlled. In this study, we first established the mathematical models of the multirotor and the manipulator separately. Their mutual influence is regarded as each other’s disturbance, and the overall linkage mechanism is established through analysis. Then, a robust sliding mode control strategy is developed for accurate trajectory tracking. The controller is derived from Lyapunov theory, which can ensure the stability of the closed-loop system. To compensate for the effect of system uncertainty, an adaptive radial basis function neural network is devised to approximate the part of the controller containing the model information. In addition, an impedance controller is designed to convert force control into position control to make the manipulator contact with the target compliantly. Finally, the simulation and experimental results indicate that the proposed method can guarantee the accuracy of the contact force and has good robustness.

To solve UAVs’ problems of inefficient search, redundant paths, and local optimal solutions in complex terrain, this study proposes an improved method based on the particle swarm optimization (PSO) algorithm. First, it built a complex terrain environment model with multiple obstacles and introduced a Tent chaotic mapping function to initialize the particle swarm in the standard PSO algorithm to enhance the global search capability. Secondly, it included the Lévy flight to widen the search in space and help the particles overcome the local optimum. At the same time, it adopted the variable spiral search strategy to enable a precise search at its current position when the local optimum occurs. Finally, it used the B spline curve to smooth the optimized path. The results show that the improved algorithm generates a reliable path that meets the needs of UAVs.

An improved YOLOv4 mobile robot environment awareness method was proposed to address the problems of occlusion and blurring of images and low detection accuracy and slow detection speed in the case of complex backgrounds in mobile robot obstacle detection.Firstly,the YOLOv4 backbone feature extraction network was replaced with a Ghostnet lightweight network structure,and the computational cost of the network was reduced.Secondly,a spatially separable residual fusion structure was added to the feature extraction network to improve the accuracy of target detection.Finally the obstacle detection algorithm was validated on a mobile robot.The experimental results showed that the mAP of the improved model reached 91.7%,the frame rate was up to 40.1 FPS,the weight size was reduced by 70%,and the model had a better detection effect on multiple targets,overlapping targets and fuzzy targets,which met the real-time and accuracy requirements of the mobile robot obstacle detection.

Plant protection UAVs are becoming the preferred plant protection method for agricultural pest control. At present, the evaluation of droplet distribution in aerial spraying is collected and evaluated after the completion of prevention and control operations, and there is a lack of real-time evaluation methods. Based on the flight parameter during the UAV plant protection process, real-time estimation of droplet distribution is the key to solving this problem and further improving the effectiveness of aerial spraying. This study proposes a merging algorithm for arbitrary polygonal regions, meshing the boundaries of the region, divide the mesh segments based on the overlapping meshes between the two regions, and connect the valid mesh connection segments of the two regions according to certain rules to obtain the intersection, union, and residual operation results between the regions. Afterwards, software based on this algorithm was developed and applied to generate spraying coverage regions, leakage spray regions, and repeated spray regions. The experimental results on theoretical and irregular routes show that the algorithm can accurately generate droplet distribution regions. The error of the calculation results with a mesh scale of 0.05 m is within 7‰, and the operating speed is above 30 Hz, meeting the real-time requirements. The smaller the mesh scale is, the higher the accuracy of the calculation results is, but the slower the calculation speed. Therefore, in practical applications, it is necessary to choose an appropriate mesh scale based on hardware computing power and accuracy level requirements. This study solves the problem of cumulative calculation of droplet distribution during the operation of plant protection UAVs, providing a basis for objectively evaluating the operation quality of plant protection UAVs and optimizing the setting of operation parameters.

Smart transformation and green development are the core research directions of electric vehicles. An electric tractor is powered by the vehicle battery. The motor converts electric energy into mechanical energy and drives the wheels through the drive train. Therefore, the electric tractor model is a modular mathematical model for the battery, motor, drive train, and drive wheel. A class of high-order terminal sliding mode control strategies is adopted to establish the relative rotational angles of drive wheels, driving angular speeds, and motor angular speeds as input, and driving angular speeds and motor angular speeds as output. This process ensures stable operating speed and good working quality under the operating conditions and achieves small-scale unattended driving. The output is a nonlinear system state equation. An n-order derivative continuous function is introduced to design the terminal sliding surface of the sliding mode. A control function to reduce chattering is also designed to ensure that the output function converges at the finite time and the existing sliding stage achieves zero steady-state error. Simulation results of the whole electric tractor model show that the speed remains stable under the condition of outside interference, and experiments verify the feasibility of the control strategy.

为满足移动机器人障碍物检测精度与实时性的需求,解决识别网络复杂度高、体积大的问题,提出一种基于改进YOLOv4的移动机器人障碍物检测识别方法.首先,使用MobileNetv3代替原主干特征提取网络,同时使用深度可分离卷积块代替加强特征提取网络中的普通卷积块;其次,使用CDIoU提高对障碍物区域的检测能力;最后,基于移动机器人平台进行障碍物检测实验.实验结果表明,改进的模型可以达到90.2的mAP和42.04的FPS,与其他目标检测模型相比,参数量减少了40％,检测速度得到了大幅度的提升,所提障碍物检测算法具有检测精度高、检测速度快的优点,可以满足移动机器人障碍物检测的实时性和准确性要求.

Unmanned Aerial Manipulation (UAM) is a novel type of Unmanned Aerial Vehicle (UAV) equipped with manipulators instead of manual operation in hazardous and unreachable environments. The combination of UAV and manipulator unavoidably causes a significant predicament due to the increase of nonlinearity and coupling of the UAM system. Consequently, the system’s robustness becomes more vulnerable in the presence of system uncertainty and external disturbance. In addition, as a real-time embedded system, rapid and precise tracking of the desired trajectory is an essential aspect of UAM performance. This study aims to establish the dynamic model of UAM and propose a global fast terminal sliding mode controller for trajectory tracking. The controller is derived from Lyapunov theory to ensure the stability of the closed-loop system. We propose a set of illustrative metrics to evaluate the performance of the designed controller and compare it with the other two controllers by simulation. The results show that the proposed controller can effectively reduce the convergence time of tracking error and has good robustness and mechanical properties. And experimental results also verified its effectiveness.

Contact force control for Unmanned Aerial Manipulators (UAMs) is a challenging issue today. This paper designs a new method to stabilize the UAM system during the formation of contact force with the target. Firstly, the dynamic model of the contact process between the UAM and the target is derived. Then, a non-singular global fast terminal sliding mode controller (NGFTSMC) is proposed to guarantee that the contact process is completed within a finite time. Moreover, to compensate for system uncertainties and external disturbances, the equivalent part of the controller is estimated by an adaptive radial basis function neural network (RBFNN). Finally, the Lyapunov theory is applied to validate the global stability of the closed-loop system and derive the adaptive law for the neural network weight matrix online updating. Simulation and experimental results demonstrate that the proposed method can stably form a continuous contact force and reduce the chattering with good robustness.

采用Solidworks软件建立了 WJ-2型弹条扣件三维有限元模型,对弹条在自由状态和标准安装状态下的固有频率和振型进行了分析,得知该弹条在服役状态下与轮轨波磨激励一致时易引发共振,且其共振频率位于尾部支承位置,与现场断裂位置一致.基于此,在不影响WJ-2型弹条扣件配套设施的前提下,对该弹条结构进行优化设计,并分别对弹条内腔掏空后无填充阻尼材料、填充橡胶及填充铁素体球墨铸铁等3种工况进行仿真分析.结果表明,弹条内腔填充灰口铸铁时,其减振性能良好,但力学性能不佳;弹条内腔填充铁素体球墨铸铁时,其力学性能良好,但收效同样不明显.

The aerial manipulator is a novel flying robot consisting of an unmanned aerial vehicle (UAV) and a multi-degree-of-freedom (DoF) robotic arm. It can actively interact with the environment to conduct dangerous or inaccessible tasks for humans. In this paper, we propose a composite control scheme considering force and position for the aerial manipulator to operate in steady contact with the environment when influenced by external disturbances. First, a contact force control method without employing the force sensor is obtained on the mechanical relationship of the system's contact with the environment. Second, we regard the system's internal coupling and external disturbance as lumped disturbances and design an extended state observer (ESO) to estimate them. Combined with the disturbance estimation and the nonsingular global fast sliding mode algorithm, a controller derived from the Lyapunov theory is proposed. Finally, we compare the proposed controller with the other four controllers through simulations and actual flight experiments. The results show that the proposed controller can effectively restrain disturbances, reduce convergence time, and guarantee steady contact under external disturbances.

An unmanned aerial manipulator (UAM) is a novel flying robot consisting of an unmanned aerial vehicle (UAV) and a multi-degree-of-freedom (DoF) robotic arm. It can actively interact with the environment to conduct dangerous or inaccessible tasks for humans. Owing to the underactuated characteristics of UAVs and the coupling generated by the rigid connection with the manipulator, robustness and a high-precision controller are critical for UAMs. In this paper, we propose a nonsingular global fast terminal sliding mode (NGFTSM) controller for UAMs to track the expected trajectory under the influence of disturbances based on a reasonably simplified UAM system dynamics model. To achieve active anti-disturbance and high tracking accuracy in a UAM system, we incorporate an RBF neural network into the controller to estimate lumped disturbances, including internal coupling and external disturbances. The controller and neural network are derived according to Lyapunov theory to ensure the system’s stability. In addition, we propose a set of illustrative metrics to evaluate the performance of the designed controller and compare it with other controllers by simulations. The results show that the proposed controller can effectively enhance the robustness and accuracy of a UAM system with satisfactory convergence. The experimental results also verify the effectiveness of the proposed controller.

Spider-web-inspired hierarchical clustering network is an emerging research topic in wireless sensor networks (WSNs), benefitting from the particular characteristics in invulnerability. It is of great practical significance to achieve a proper parameter combination for optimizing network clustering and cluster head (CH) selection. However, there is no feasible solution to determine these parameters to lessen the energy consumption of resource-limited spider-web-inspired WSNs. Meanwhile, the existing protocols cannot adequately balance the total network energy dissipation and the network performance due to insufficient consideration of invulnerability. In this paper, a novel multi-objective optimization approach of parameter combination (MOOAPC) is proposed to solve the predicament, including the number of layers L, number of interval rounds for CH re-election m, grade communication radius within a cluster G(cr), number of sectors Z, and total number of nodes N. Specifically, the statistical methods, consisting of normality test, homogeneity of variance test, and ANOVA, are used to clarify the effect of different parameters on the network performance. Moreover, the logistic regression algorithm is applied to establish the optimization objective functions of invulnerability and average residual energy with the impact degree as the basis for setting the parameter constraints, and then, NSGA-II algorithm is adopted to acquire the optimal parameter combination. We discovered that the parameter combination (L = 6, m = 20, G(cr) = 4, Z = 4, and N = 1100) was appropriate to prolong the network lifetime. In the case of reaching the threshold of death rate of nodes, the number of death rounds of MOOAPC was 670, which was 10.56%, 4.36% and 6.35% higher than that of LEACH, HEED and EEUC, respectively. Compared with LEACH, HEED, and EEUC, MOOAPC demonstrated significant performance advantages in invulnerability and average residual energy, 42.06% and 17.99% higher on average. Based on these results, the proposed method can be utilized to increase the capability of spider-web-inspired WSNs against deterioration of quality of service and energy constraints.

针对PC(Personal Computer)在无人机火灾检测应用中的环境受限问题,提出了一种基于YOLOv2-Tiny的无人机火灾检测与云台跟踪方法.首先,在改进的YOLOv2-Tiny模型上进行预训练获得最优YOLOv2-Tiny模型,并在K210开发板上部署最优YOLOv2-Tiny模型;其次,将检测到的火灾图像传至云端并将火灾框选中心与图像中心之间的距离参数传递给PID进程,控制云台实现实时火灾跟踪;最后,通过无人机实际飞行验证火灾检测与云台跟踪的能力.实验结果表明,相较于YO?LOv2模型,YOLOv2-Tiny在测试集上具有更高的检出率,检出率达到96.66％,并且检测速度达到每秒14帧,云台跟踪中心位置像素误差(CPE)低于5,实时检测与跟踪过程中无人机姿态角保持相对稳定,该研究在火灾实时检测方面具有潜力.

Energy limitation is one of the intrinsic shortcomings of wireless sensor networks (WSNs), although it has been widely applied in disaster response, battlefield surveillance, wildfire monitoring, radioactivity detection, etc. Due to the large amount of energy consumed for data transmission, how to prolong the network lifespan by designing various hierarchical routing protocols has attracted more and more attention. As a result, numerous achievements have emerged successively. However, these presented mechanisms can rarely guarantee the satisfactory quality of service (QoS), while lowering the energy cost level of WSNs. Meanwhile, invulnerability is undoubtedly an excellent quantitative index to assess QoS. Therefore, it is critical to develop a practical routing method to optimize network lifetime by considering both invulnerability and energy efficiency. Game theory is suitable for such a critical problem as it can be used in node or at network level to encourage the decision-making capabilities of WSNs. In this paper, a novel invulnerability-aware clustering routing algorithm (IACRA) using game-theoretic method is proposed to solve the predicament. The core features of the addressed game-theory-based routing protocol include integral invulnerability awareness, optimal cluster head selection in hierarchical routing, distance-aware cluster head discovery, and cluster rotation update mechanism for lifetime optimization. Particularly, the integral network invulnerability based on weighted fusion is constructed for further defining the profit model by combining the invulnerability indicators used to evaluate the local and whole network. Meanwhile, the optimal probability function of every node elected as CH in per cluster is established through the game between invulnerability and node energy consumption. In addition, the cluster update mechanism base on cluster rotation is proposed to avoid the rapid death of nodes with large energy consumption for maximizing network lifetime. The experimental results indicated a significant improvement in energy balance as well as in invulnerability compared with the other three kinds of well-known clustering routing protocols including GEEC (Game-theory-based energy efficient clustering routing protocol), HGTD (Hybrid, game-theory-based distributed clustering protocol), and EEGC (Efficient energy-aware and game-theory-based clustering protocol). Concretely, at the 400 communication rounds, the invulnerability of IACRA was higher than that of GEEC, HGTD, and EEGC by 77.56%, 29.45% and 15.90%, respectively, and the average residual energy of IACRA was 8.61%, 18.35% and 6.36% larger than that of GEEC, HGTD, and EEGC, respectively. Based on these results, the proposed protocol can be utilized to increase the capability of WSNs against deterioration of QoS and energy constraints.