“Gongzhu” is a card game popular in Chinese circles at home and abroad, which belongs to incomplete information game. The game process is highly reversible and has complex state space and action space. This paper proposes an algorithm that combines the Monte-Carlo (MC) method with deep neural networks, called the Deep Monte-Carlo (DMC) algorithm. Different from the traditional MC algorithm, this algorithm uses a Deep Q-Network (DQN) instead of the Q-table to update the Q-value and uses a distributed parallel training framework to build the model, which can effectively solve the problems of computational complexity and limited resources. After 24 h of training on a server with 1 GPU, the “Gongzhu” agent performed 10,000 games against the agent that uses a Convolutional Neural Network (CNN) to fit the strategies of human players. “Gongzhu” agent was able to achieve a 72.6% winning rate, and the average points per game was 63. The experimental results show that the model has better performance.
In view of the difficulty to realize the meteorological and hydrological observation on the unmanned islands or reefs, the complexity of using wireless sensor network cannot be too high and the requirement of reliability is higher, so a master-slave dual-cluster head network trust model based on adaptive weighted D-S evidence theory fusion calculation is proposed. At the same time, the reliability is adaptively adjusted according to the time sliding statistics of the node measurement data. The base station determines whether to receive data and determines the master-slave cluster head in the next cycle according to the fusion result difference and their historical data evaluation, which reduces the complexity of trust calculation and develops the remote unattended Islands and reefs meteorological and hydrological wireless sensor monitoring technology. The test results show that the technology is effective and feasible, with the advantages of convenient layout and reliable data transmission.
Personal record keeping, behaviour tracking, accurate feeding, disease prevention and control, and food traceability require the identification of dairy cows. This work proposes a unique identification that combines Mask R-CNN and ResNet101 to identify individual cows in milking parlours accurately. Using 265 Holstein cows in various positions, a facial image dataset of their faces was created using the milking hall’s webcam. The feature pyramid network-based Mask R-CNN instance segmentation model was trained to separate the cows’ faces from their backgrounds. The ResNet101 individual classification network was trained using the segmentation above data as input and cow individual numbers as output. Combining the two techniques led to the creation of the cow individual recognition model. According to experimental findings, the Mask R-CNN model has an average accuracy of 96.37% on the picture test set. The accuracy of the Resnet101-based individual classification network was 99.61% on the training set and 98.75% on the validation set, surpassing that of VGG16, GoogLeNet, ResNet34, and other networks. The study’s suggested individual recognition model outperformed the combined effect of the YOLO series model and ResNet101 in terms of test accuracy (97.58%). Furthermore, it outperforms the pairing of Mask R-CNN with VGG16, GoogleNet, and ResNet34. This research offers precision dairy farming an excellent technical basis for individual recognition.
As an important technique of driverless vehicles, 3D Point Cloud Object Detection algorithm can provide semantic information and geometrical information of various types of objects. For the problem of feature information loss in 3D Point Cloud Object Detection, this paper proposes an improved Pointpillars algorithm where a multi-scale columnar feature extraction network based on the attention mechanism. In this algorithm we extract features to obtain pseudo-images of different scales, and splicing point cloud pseudo-images of multiple scales to obtain fusion feature maps. In addition, by introducing a Convolutional Block Attention Module (CBAM), our MSCS-Piontpillars can effectively suppress the noise in the pseudo-image of the point cloud and amplify the important feature information for target classification. Comparision experimental was carried on both the KITTI 3D Object dataset and a real experiment. The results show that compared with Pointpillars, MSCS-Pointpillars (Multi-Scale Channel Spatial Attention Pointpillars) have improved the detection accuracy of cars, cyclists, pedestrians and other targets.
Due to the global COVID-19 pandemic, there is a strong demand for pharyngeal swab sampling and nucleic acid testing. Research has shown that the positive rate of nasopharyngeal swabs is higher than that of oropharyngeal swabs. However, because of the high complexity and visual obscuring of the interior nasal cavity, it is impossible to obtain the sampling path information directly from the conventional imaging principle. Through the combination of anatomical geometry and spatial visual features, in this paper, we present a new approach to generate nasopharyngeal swabs sampling path. Firstly, this paper adopts an RGB-D camera to identify and locate the subject’s facial landmarks. Secondly, the mid-sagittal plane of the subject’s head is fitted according to these landmarks. At last, the path of the nasopharyngeal swab movement in the nasal cavity is determined by anatomical geometry features of the nose. In order to verify the validity of the method, the location accuracy of the facial landmarks and the fitting accuracy of mid-sagittal plane of the head are verified. Experiments demonstrate that this method provides a feasible solution with high efficiency, safety and accuracy. Besides, it can solve the problem that the nasopharyngeal robot cannot generate path based on traditional imaging principles. It also provides a key method for automatic and intelligent sampling of nasopharyngeal swabs, and it is of great clinical value to reduce the risk of cross-infection.
Soft gripper is used in various fields for grasping, handling and manipulating objects. However, soft pneumatic gripper, which is one of the most widely studied grippers in robotics, has limitations in handling different objects and cannot qualify complicated operation tasks with inherent structures and entirely composed of soft materials. Therefore, we present a rigid-flexible coupled soft gripper with treble modular fingers, which has movable frames, air chamber actuator modules and membrane-type sensing module. The rigid frames provide rigidity to allow the gripper generating a large grasping. The actuator modules can be inflated and vacuumed, thereby enabling gripper to execute different grasping operation. The sensing module, inflating a preset pressure, is able to sense grasping force. Furthermore, the work range of gripper is adjustable, changing by contraction and elongation motion of each finger. The gripper transforms from original state to contraction state by vacuuming actuator modules, and from original state to elongation state by inflating actuator modules. Through a multi-channel pneumatic control system with positive and positive-negative pressure output, the gripper achieves flexible and stable grasping for a variety of objects, such as cherry tomato (15 g), apple (204 g), banana (420 g), etc.
In this paper, we introduce an autonomous exploration and rescue robot system based on a tracked mobile robot platform equipped with a 7 degree-of-freedom (DoF) manipulator, which realizes autonomous navigation in indoor environments and autonomous stair climbing for safe and efficient search and rescue tasks. In Sect. 2, the hardware design of the robot system is presented, which allows flexible movement and high passability to complete obstacle crossing and stair climbing. In Sects. 3 and 4, the indoor navigation algorithm and the stair detection algorithm of the robot system are presented, respectively. The ROS-based system uses the cartographer algorithm for map construction, ROS navigation for autonomous navigation and obstacle avoidance, and a depth camera for stair detection. The process of a four-flipper tracked mobile robot stair climbing is designed. The robot system is experimentally verified in Sect. 5.
The Center of Mass (CoM) is considered to be the most ideal position for robot grasping. Grasping far from the CoM is likely to cause the object to deviate from the expected pose. To robustly grasp unknown objects in unstructured and changing environments, it is necessary to rapidly predict and correct the imminent failed grasping in time and guide the robot to explore a stable grasping pose. This paper is the first to incorporate tactile sensing into reinforcement learning (RL) for robot CoM-based regrasping problem. The regrasping agent is developed and automatically optimize in a simulated environment without explicit knowledge of objects, the tactile information i.e. slip and tilt are integrated into the reward function. In 440 regrasping tests of 8 new random objects in the PyBullet simulation and 14 household objects in the real world, the average number of regrasps was 2.04 and 2.35, respectively. In comparative tests with binary search and heuristic step size adjustment strategy, our method has the highest average regrasping efficiency.
In the field of aerial manipulation, heterogeneous aerial systems with complex configuration become more and more popular, as they can overcome problems of traditional aircrafts in aerial manipulation. However, recent photo-realistic aircraft simulators do not support the accurate contact and collision behavior simulation of aircraft or the dynamics simulation of heterogeneous aerial systems. Besides, high-fidelity images are required in many machine learning-based perception and action algorithms. Therefore, we develop a simulator to provide a solution to aerial manipulation robots training, algorithm tests, and display: AeroBotSim. By using modular design, we decouple rendering engine and physics engine to obtain high-frequency states data while retrieving high-fidelity images. Also, we synchronize the contact information in rendering engine and the physics engine, and design interfaces to custom contact behavior for operation, separation, and recombination simulation. In this paper, we present our framework design and dynamic models of aircrafts under physical interaction. Then, we validate our simulator framework in three aspects: baseline controller in ROS, vision-based algorithm, and contact simulation respectively.
In this study, a plasmonic nano-sensor with excellent performance based on aperture-coupled square resonator is proposed. Utilizing the 2D finite element algorithm, the transmission characteristics of MIM waveguide structure are systematically explored. Simulation shows that a sharp fano line-shape is formed in a succinct structure with its FOM (figure of merit) over 2000. In addition, we systematically investigate the coupling distance, the geometric parameter of the resonator, in detail. Optimizing the structural parameter results and introducing asymmetry into the structure by two means, such as horizontally moving the center of square resonator or adding a tiny groove besides the square resonator. Both of these can cause extra fano profile in transmission pattern while the former fano profile maintains stable, thus this property makes the metrics in sensor of this structure more abundant. Results prove that this compact and asymmetric structure has great potential application in nano-sensor, optical switches and nonlinear devices in future highly integrated optical circuits.
Unmanned tractor-trailer vehicles are widely used in factory transportation scenarios. However, the trailer hitching process is still manually operated. The automatic trailer hitching is the precondition of fully unmanned logistics. Existing auto hitching methods directly detect the trailer or coupler features, making them hard to generalize and deploy to various types of trailers and couplers. To address this problem, this paper proposes a trailer hitch system using fiducial markers. The system is divided into two modules: hitch coupler pose estimation and visual servoing. An algorithm based on AprilTag detection is used for hitch coupler pose estimation. The pose messages guide the tractor to reverse to the trailer. An algorithm based on decoupled lateral-longitudinal control is used for visual servoing. The proposed system is experimented on 4 different tractor-trailer vehicles with 254 tests under various conditions. Large scale test condition variations include initial position and orientation, light illumination, indoor/outdoor scenario, and fiducial marker damage. Overall success rate of 95
Flexible printed circuit boards (FPCBs) are widely used in the electronic information industry. Due to their different shapes and porous, the traditional vacuum suction has problems such as leakage. Dry adhesive technology has strong potential to replace existing techniques since it provides stable handling. To this end, this paper proposes a negative pressure-driven adhesion gripper based on an annular wedge-shaped microstructure. The deformable chamber is designed to be cylindrical, which provides shear loads for annular wedge-shaped dry adhesion. The adhesion switch by shear loads and pick-and-place operations for FPCBs can be realized. During the operations, the contact state is critical for adhesion performance. To realize better contact and stability, this paper focuses on the plane adaptation technology of the negative pressure-actuation adhesion gripper, and a corrugated connector is modeled and mounted in the gripper. A series of experiments demonstrate that the adhesion becomes more stable and the adhesion force is increased by more than 20
In this paper, an LED (Light-Emitting Diode) screen number detection method is proposed to guide the autonomous navigation of UAV (Unmanned Aerial Vehicle) for further task implementation. Aiming at the LED module identification in the task of detecting bright and dark lines on the LED screens, an improved YOLOv5 object detection algorithm has been developed with a series of operations, i.e., the backbone network, activation function, optimization strategy and post-processing screening prediction box. The LED dataset is designed, along with data augmentation so as to achieve a high detection accuracy of the trained detection model in too strong or too dark light complex scenes. The trained model can be transplanted to the Android platform, which provides support for subsequent UAV automatic navigation and portable screen calibration. With extremely high detection accuracy (0.843 mAP@0.5), the detection speed of the proposed model is also quite fast, achieving almost real-time performance (85 FPS).
Current service robots without learning ability are not qualified for many complex tasks. Therefore, it is very significant to decompose the complex task into repeatable execution unit. In this paper, we propose a complex task representation method based on dynamic motion primitives, and use hierarchical knowledge graph to represent the analytic results of complex tasks. To realize the execution of complex robot manipulation tasks, we decompose the semantic tasks into the minimum motion units that can be executed by the robot and combine the multi-modal information: posture, force and robot joint parameters, obtained by the sensors. We use the knowledge graph to record the end-effector required by the robot to perform different tasks and make appropriate selection of end-effector according to different needs. Finally, Taking the long sequence complex task of service scene as an example, we use UR5 robot to verify the effectiveness and feasibility of this design.
Aiming at the needs of civil market tasks such as autonomous swimming display of pools in the aquarium and special tasks such as intelligent autonomous patrol in specific water areas, this paper proposes a task driven control method for manta robot to circumnavigate around the shores. We equip the robot fish with a complete information sensing network and realize its driving and yaw control basing on the CPG phase oscillator network. We generate an offline look-up table by using fuzzy control method, and realize the closed-loop control of heading by looking up this table. Basing on the requirements of the circumnavigation task, we propose a real-time obstacle avoidance strategy combined with the infrared range sensor information. Finally, we build an experimental pool platform to conduct underwater alongshore circumnavigating experiments of the robot fish, and the experimental results prove the effectiveness of the overall scheme.
In the industrial production line, putting several small packages into large packages with the required arrangement rules still requires a lot of manual work to complete. In order to solve this problem, we build an automatic robot grasping system based on robotic arm, industrial camera and a uniform-speed conveyor, which can automatically put small packages into large packages. We use the YOLO algorithm to identify the position, category and number of small packages. The number of snacks identified is utilized to plan the degree of gripper closure. The trajectory of the small packages can be predicted according to the speed of the conveyor and the position of small packages on the robot coordinate system, then we fuse multiple predicted trajectories to form a complete and coherent robot arm trajectory to grasp the small packages and put them into large packages. This system can save a large amount of labor and reflect the intelligence of the robot.
With the aggravation of aging, a series of ankle injuries, including muscle weakness symptoms, caused by stroke or other reasons have made the problem of ankle rehabilitation increasingly prominent. Muscle strength training is one of the main rehabilitation methods for the ankle joint complex (AJC). Based on the current incomplete development of muscle strength training modes for all forms of ankle movement, this study developed six muscle strength training modes for the human ankle on our parallel ankle rehabilitation robot, namely continuous passive motion (CPM), isotonic exercise, isometric exercise, isokinetic exercise, centripetal exercise and centrifugal exercise, based on position inverse solution of the robot combining with the admittance control or position control. The dorsiflexion (DO) movement was used as an example to analyze the training effect of each training mode, with the results showing good function performance of the developed ankle muscle strength training methods.
At present, most of the traditional algorithms for mobile robot autonomous exploration in unknown environments use frontier as a guide and adopt greedy strategies for determining the next exploration target. They generate new frontier as new regions of the map, and thus the cycle repeats itself to finally complete the exploration of unknown environments. In this paper, we propose a region clustering-based approach for autonomous exploration of mobile robots in complex environments. Our approach incorporates the concept of region clustering at its global level based on the current advanced hierarchical exploration framework. The robot is more inclined to finish exploring a certain region of the map first, thereby minimizing the robot’s repetitive exploration of explored regions. Mobile robot autonomous exploration experiments were implemented in our college campus. The experimental results show that the average exploration trajectory length was reduced by 14.03%, and the average exploration time was reduced by 16.15%, respectively.
EIEE syndrome, known as early infantile epileptic encephalopathy, is considered to be the earliest onset form of age-dependent epileptic encephalopathy. The main manifestations are tonic-spasmodic seizures in early infancy, accompanied by burst suppressive electroencephalogram (EEG) patterns and severe psychomotor disturbances, with structural brain lesions in some cases. Specific to EIEE syndrome, this paper presents a comprehensive analysis of EEG features at three different periods: pre-seizure, seizure and post-seizure. Coherent features are extracted to characterize EEG signals in EIEE syndrome, and Kruskal-Wallis H Test and Gradient-weighted Class Activation Mapping (Grad-CAM) are used to investigate and visualize the significance of features in different frequency band for distinguishing the three stages. The study found that activity synchrony between temporal and central regions decreased significantly in the γ band during seizures. And the coherence feature in the γ band combined with the ResNet18-based seizure detection model achieved an accuracy of 91.86 γ band can be considered as a biomarker of seizure cycle changes in EIEE syndrome.
It is a difficult thing for robot working in a tight and narrow space with obstacles because of collision occurrence. For solving this problem, the paper proposes a joint trajectory generation method for obstacle avoidance. Besides of the end-effector, our work plans a collision free trajectory for each joint in the narrow space. Considering the complexity of obstacle distribution, the presented method combines Dynamic Movement Primitive (DMP) with a RRT-Connect algorithm that firstly, in the joint space DMPs generate trajectories for each manipulator joint, and then, in the cartesian space, the collision detection model checks the DMP generated trajectories. If any of the links collides with the obstacle, a collision free path will be planned on the trajectory points that encounter obstacles by employing RRT-Connect algorithm. Based on ROS platform, the experiments build a tight and narrow simulated environment, and test the method on a UR3 robot manipulator, which show the effectiveness of the presented method.
