The electric unmanned aerial vehicles (UAVs) are rapidly growing due to their abilities to perform some difficult or dangerous tasks as well as many public services including real-time monitoring, wireless coverage, s...
详细信息
The electric unmanned aerial vehicles (UAVs) are rapidly growing due to their abilities to perform some difficult or dangerous tasks as well as many public services including real-time monitoring, wireless coverage, search and rescue, wildlife surveys, and precision agriculture. However, the electrochemical power supply system of UAV is a critical issue in terms of its energy/power densities and lifetime for service endurance. In this paper, the current power supply systems used in UAVs are comprehensively reviewed and analyzed on the existing power configurations and the energy management systems. It is identified that a single type of electrochemical power source is not enough to support a UAV to achieve a long-haul flight;hence, a hybrid power system architecture is necessary. To make use of the advantages of each type of power source to increase the endurance and achieve good performance of the UAVs, the hybrid systems containing two or three types of power sources (fuel cell,battery, solar cell, and supercapacitor,) have to be developed. In this regard, the selection of an appropriate hybrid power structure with the optimized energy management system is critical for the efficient operation of a UAV. It is found that the data-driven models with artificial intelligence (AI) are promising in intelligent energy management. This paper can provide insights and guidelines for future research and development into the design and fabrication of the advanced UAV power systems.
Graph neural networks(GNNs) have shown great popularity and achieved promising performance on various graph-based tasks in the past years. However, there is little work that explores the information fusion mechanism, ...
详细信息
Graph neural networks(GNNs) have shown great popularity and achieved promising performance on various graph-based tasks in the past years. However, there is little work that explores the information fusion mechanism, which plays an import role in GNNs. Besides, datasets in the real world often have noises, which make the information fusion difficult. In this paper, we give an information-theoretic explanation. Specifically, we focus on how the information from topological structures and node features fuses and how different information contributes to the downstream task. Furthermore, we propose a general framework named M-GCN to express the fusion process in GNNs. Graph embeddings and feature graph are introduced to extract the information from topological structure and node features separately in M-GCN. Extensive experiments are conducted on several benchmark datasets and experimental results show that our proposed models are more robust and outperform state-of-the-art methods.
Flat objects with negligible thicknesses like books and disks are challenging to be grasped by the robot because of the width limit of the robot's gripper, especially when they are in cluttered environments. Pre-g...
Flat objects with negligible thicknesses like books and disks are challenging to be grasped by the robot because of the width limit of the robot's gripper, especially when they are in cluttered environments. Pre-grasp manipulation is conducive to rearranging objects on the table and moving the flat objects to the table edge, making them graspable. In this paper, we formulate this task as Parameterized Action Markov decision Process, and a novel method based on deep reinforcement learning is proposed to address this problem by introducing sliding primitives as actions. A weight-sharing policy network is utilized to predict the sliding primitive's parameters for each object, and a Q-network is adopted to select the acted object among all the candidates on the table. Meanwhile, via integrating a curriculum learning scheme, our method can be scaled to cluttered environments with more objects. In both simulation and real-world experiments, our method surpasses the existing methods and achieves pre-grasp manipulation with higher task success rates and fewer action steps. Without fine-tuning, it can be generalized to novel shapes and household objects with more than 85% success rates in the real world. Videos and supplementary materials are available at https://***/view/pre-grasp-sliding.
Admittance control is one of the important methods of compliance control. It is performed by inputting force signals and outputting position signals. Position control can achieve high accuracy, but the force signals h...
详细信息
ISBN:
(纸本)9781665478977
Admittance control is one of the important methods of compliance control. It is performed by inputting force signals and outputting position signals. Position control can achieve high accuracy, but the force signals have large errors, which is an important factor affecting its accuracy. When the force sensor collects force data, there are non-negligible initial deviations due to gravity, installation, and other factors, resulting in poor admittance control accuracy. In order to solve this problem, this paper takes the six-degree-of-freedom manipulator admittance control as the research object. According to the mapping relationship between the end attitude and the sensor errors, we propose a method of obtaining the initial offset value of the sensor through data interpolation. This method traverses the initial values of the sensor in each working posture when the sensor has no external force input, and then performs secondary interpolation to obtain the offset values of the sensor in any posture, corrects the output values of the force/torque sensor, and then improve the accuracy of admittance control. The experimental results show that this method can quickly and effectively improve the calculation accuracy of the sensor’s initial deviation values and the admittance control accuracy.
The deadbeat predictive current control(DPCC) exhibits strong dynamic performance in the current control of permanent magnet synchronous motor(PMSM),whereas the performance of the traditional DPCC depends on the accur...
详细信息
The deadbeat predictive current control(DPCC) exhibits strong dynamic performance in the current control of permanent magnet synchronous motor(PMSM),whereas the performance of the traditional DPCC depends on the accuracy of model parameters to a great *** cope with this problem,parameter sensitivity of the traditional DPCC has been analyzed in detail,with a disturbance observation compensation method proposed to improve the robustness of the current control of ***,a method of non-singular terminal sliding mode disturbance observer(NTSMDO) which employs a novel reaching law,combined with the DPCC is put forward to enhance the parameters robustness of the current loop.A harmonic compensation(HC) term has been added to the output of the current control for ignoring the harmonic electromagnetic process in the mathematical model of the PMSM to further promote the performance of the current loop *** and experimental results demonstrate that the proposed method offers robustness against parameters mismatch and control precision compared with the traditional DPCC.
Machine learning, classification, and clustering techniques use the distance functions to evaluate the proximity between data entries and deduce the best neighbouring element and the closest matching entry. The best n...
Machine learning, classification, and clustering techniques use the distance functions to evaluate the proximity between data entries and deduce the best neighbouring element and the closest matching entry. The best neighbour is not only the closest neighbour but a neighbour that is quick to respond. In view of that, a time-based isochronous metric is introduced to evaluate the best neighbours and form linkages by grouping similar entities. The proposed method uses parametric equations of the fastest descent and solves the time variables for attributes localised in curved space–time. The time metric is compared with commonly used distance metrics for accuracy in classification and clustering using benchmark and commonly used datasets. The nearest-neighbour technique is used for evaluating classification accuracy, and an adjusted random index (ARI) is used to evaluate clustering. The proposed method shows better accuracy and ARI in comparison to distance functions. It also assigns better weights to attributes of the dataset and easily identifies repeated patterns in noisy time series data.
Micro-assembly is an emerging method to fabricate microrobots with multiple modules or particles. However, there is always a lack of a flexible and efficient method to freely create the desired magnetic soft microrobo...
详细信息
ISBN:
(数字)9798350384574
ISBN:
(纸本)9798350384581
Micro-assembly is an emerging method to fabricate microrobots with multiple modules or particles. However, there is always a lack of a flexible and efficient method to freely create the desired magnetic soft microrobots. In this paper, an automated assembly system based on a two-fingered microhand is presented for fabricating magnetic soft microrobots. Our proposed system can automatically pick and place components to assemble microrobots with a two-fingered micromanipulator, and orient these components through an external magnetic field. The automated assembly has the advantages of high accuracy, high speed, and high success rate. It can endow magnetic microrobots with flexible material selection, arbitrary geometry design, and programable magnetization profile. We can make full use of this system to fabricate multiple magnetic soft microrobots. The experiment results demonstrate that this system can efficiently fabricate microrobots with excellent mechanical properties, which have application potential in robotics, biomedical engineering, and environmental governance.
This study aims to investigate the problem of attitude control for a spacecraft with inertial uncertainties, external disturbances, and communication restrictions. An event-triggered active disturbance rejection contr...
详细信息
This study aims to investigate the problem of attitude control for a spacecraft with inertial uncertainties, external disturbances, and communication restrictions. An event-triggered active disturbance rejection control approach is proposed for attitude tracking of the spacecraft. An event-triggered mechanism is introduced together with an extended state observer to jointly monitor the systemstates and total disturbances. The observation error is proved to be uniformly bounded. Based on the proposed control scheme,the integrated tracking system is shown to be asymptotically stable, implying successful attitude tracking of the spacecraft for the desired motion. Numerical results illustrate the effectiveness of the control strategy in achieving satisfactory tracking performance with a reduced data-transmission cost.
The present paper considers the model-based and data-driven control of unknown linear time-invariant discretetime systems under event-triggering and self-triggering transmission schemes. To this end, we begin by prese...
详细信息
This paper proposes a self-attention based temporal intrinsic reward model for reinforcement learning (RL), to synthesize the control policy for the agent constrained by the sparse reward in partially observable envir...
详细信息
ISBN:
(纸本)9781665426480
This paper proposes a self-attention based temporal intrinsic reward model for reinforcement learning (RL), to synthesize the control policy for the agent constrained by the sparse reward in partially observable environments. This approach can solve the problem of temporal credit assignment to some extent and deal with the low efficiency of exploration. We first introduce a sequence-based self-attention mechanism to generate the temporary features, which can effectively capture the temporal property of the task for the agent. During the training process, the temporary features are employed in each sampled episode to elaborate the intrinsic rewards, which is combined with the extrinsic reward to help the agent learn a feasible policy. Then we use the meta-gradient to update this intrinsic reward model in order that the agent can achieve better performance. Experiments are given to demonstrate the superiority of the proposed method.
暂无评论