Search

We propose a meta-ability decoupling (MAD) paradigm, which brings together various object navigation methods in an architecture system, allowing them to mutually enhance each other and evolve together. Based on the MAD paradigm, we design a multiple thinking (MT) model that leverages distinct thinking to abstract various meta-abilities. Our method decouples meta-abilities from three aspects: input, encoding, and reward while employing the multiple thinking collaboration (MTC) module to promote mutual cooperation between thinking. MAD introduces a novel qualitative and quantitative interpretability system for object navigation. Through extensive experiments on AI2-Thor and RoboTHOR, we demonstrate that our method outperforms state-of-the-art (SOTA) methods on both typical and zero-shot object navigation tasks., Comment: 17 pages

6D object pose estimation for texture-less objects from RGB images remains challenging, especially in occlusion scenarios. Instead of localizing sparse keypoints by regressing their image coordinates or heatmaps, which are sensitive to occlusion, we introduce GeoPose, a novel reconstruction guided pose estimation pipeline that predicts dense correspondences and leverages geometric consistency effectively. We first design a dense reconstruction network (ReconNet) to reconstruct pixel-wise object coordinates in normalization space. Dense 2D-3D correspondences are generated intuitively by our explicit pa-rameterization for 3D object models, which dismisses keypoint selection efforts. These 2D-3D correspondences are then utilized to estimate 6D poses by the PnP algorithm with RANSAC iterations. Furthermore, a novel Cycle Loss is proposed to provide 3D prior supervision, which significantly correlates with the pose estimation task by guiding geometric consistency between reconstruction (pixel to 3D) and reprojection (3D to pixel). In addition, a training data augmentation method is proposed to handle the insufficiency of 6D datasets, the acquisition of which is error-prone and time-consuming. Extensive experiments demonstrate that, compared with existing RGB-based methods, our GeoPose can achieve state-of-the-art (SOTA) 6D pose estimation performance on the LINEMOD, Occlusion LINEMOD and T-LESS datasets.

This paper aims to solve the problem of estimating the 6D pose of an object under occlusion using RGB-D images. Most existing methods typically use the information of color and depth images separately to make predictions, which limits their performances in the presence of occlusion. Instead, we propose a pipeline to effectively fuse color and depth information and perform region-level pose estimation. Our method first uses a CNN to extract the color features, and then we obtain the fusion features by combining the color features into the point cloud. Unlike existing methods, the fusion features are in the form of point sets instead of feature maps. We further use a PointNet++-like network to process the fusion features, obtaining several region-level features. Each region-level feature can predict a pose with confidence. The pose with the highest confidence is chosen as the final output. Experiments show that the proposed method outperforms the state-of-the-art methods on both the LINEMOD and Occlusion LINEMOD datasets, indicating that the proposed pipeline can obtain accurate pose estimation results and is robust to occlusion.

In product inspection scenarios, the objects have different characteristics of textures and different potential anomalies. Still, traditional methods cannot precisely segment structural anomalies and achieve optimal performance on all object categories. This article introduces a flexible multilevel image reconstruction (MLIR) framework for the unsupervised anomaly segmentation task. The MLIR integrates multiple generators with different downsampling ratios to serve as multiple reconstruction levels and leverage an adjustable perceptual similarity measurement to obtain the featurewise reconstruction errors, overcoming the challenge of detecting multiscale anomalies. Leveraging embeddings with different receptive fields for reconstruction, generators perform differently on the same target. Therefore, we depend on the reconstruction uncertainty to propose an adaptive attention-level transition (ALT) strategy that optimizes the detection performance automatically according to the target’s category. ALT adjusts the weights of both reconstruction levels and feature measurement scales simultaneously to utilize features at a consistent level for reconstruction and anomaly detection (AD). This strategy remarkably improves the system’s adaptivity for different applications. Moreover, we propose reducing the signal-to-noise ratio (SNR) of the images by adding pepper noise so that the reconstruction process is framed as an image denoising task, which improves feature extraction and image restoration robustness. We conduct extensive experiments on the MVTec AD dataset and demonstrate that the proposed method performs satisfactorily to different categories and achieves the state-of-the-art (SOTA) performance.

For biped robots, the ability to maintain balance under external forces is an essential requirement. Inspired by human beings’ behaviors to resist external forces, a compliant-resistant balance-control method is proposed to keep the biped robot balance subjected to an external force. A model-free trajectory generator is designed based on the central pattern generator (CPG) to generate compliant-resistant human-like behavior. The CPG pattern generator generates the desired pulse signal utilizing Matsuoka’s CPG. The signal modulator applies the defined signal to the robot’s center of mass (CoM) to generate the workspace trajectory when standing on double feet. Moreover, when standing on single foot, the output signal of the CPG will directly act on the hip joint of the robot to generate the joint space trajectory. Furthermore, the motion engine calculates the workspace trajectory into joint sequence values. The proposed control strategy can generate defined pulse signals to realize compliant-resistant balance control for biped robots. The control strategy proposed in this paper is verified in the NAO simulation environment.

SUMMARYSelf-localization in highly dynamic environments is still a challenging problem for humanoid robots with limited computation resource. In this paper, we propose a dual-channel unscented particle filter (DC-UPF)-based localization method to address it. A key novelty of this approach is that it employs a dual-channel switch mechanism in measurement updating procedure of particle filter, solving for sparse vision feature in motion, and it leverages data from a camera, a walking odometer, and an inertial measurement unit. Extensive experiments with an NAO robot demonstrate that DC-UPF outperforms UPF and Monte–Carlo localization with regard to accuracy.

Correspondence matching is a fundamental problem in computer vision and robotics applications. Solving correspondence matching problems using neural networks has been on the rise recently. Rotation-equivariance and scale-equivariance are both critical in correspondence matching applications. Classical correspondence matching approaches are designed to withstand scaling and rotation transformations. However, the features extracted using convolutional neural networks (CNNs) are only translation-equivariant to a certain extent. Recently, researchers have strived to improve the rotation-equivariance of CNNs based on group theories. Sim(2) is the group of similarity transformations in the 2D plane. This paper presents a specialized dataset dedicated to evaluating sim(2)-equivariant correspondence matching algorithms. We compare the performance of 16 state-of-the-art (SoTA) correspondence matching approaches. The experimental results demonstrate the importance of group equivariant algorithms for correspondence matching on various sim(2) transformation conditions. Since the subpixel accuracy achieved by CNN-based correspondence matching approaches is unsatisfactory, this specific area requires more attention in future works. Our dataset is publicly available at: mias.group/SIM2E., Comment: ECCV2022 Workshop Paper

"Search for" or "Navigate to"? When finding an object, the two choices always come up in our subconscious mind. Before seeing the target, we search for the target based on experience. After seeing the target, we remember the target location and navigate to. However, recently methods in object navigation field almost only consider using object association to enhance "search for" phase while neglect the importance of "navigate to" phase. Therefore, this paper proposes the dual adaptive thinking (DAT) method to flexibly adjust the different thinking strategies at different navigation stages. Dual thinking includes search thinking with the object association ability and navigation thinking with the target location ability. To make the navigation thinking more effective, we design the target-oriented memory graph (TOMG) to store historical target information and the target-aware multi-scale aggregator (TAMSA) to encode the relative target position. We assess our methods on the AI2-Thor dataset. Compared with the state-of-the-art (SOTA) method, our method reports 10.8%, 21.5% and 15.7% increase in success rate (SR), success weighted by path length (SPL) and success weighted by navigation efficiency (SNE), respectively., Comment: 12 pages, ready for AAAI2023

Prime editing is a universal and very promising precise genome editing technology. However, optimization of prime editor (PE) from different aspects remains vital for its use as a routine tool in plant basic research and crop molecular breeding. In this report, we tested MS2-based prime editor (MS2PE). We fused the M-MLV reverse transcriptase (RT) gene variant to the MS2 RNA binding protein gene, MCP, and allowed the MCP-RT fusion gene to co-express with the SpCas9 nickase gene, SpCas9H840A, and various engineered pegRNAs harboring MS2 RNA (MS2pegR). Compared with control PEs, MS2PEs significantly enhanced editing efficiency at four of six targets in rice protoplasts, and achieved 1.2~10.1-fold increase in editing efficiency at five of six targets in transgenic rice lines. Furthermore, we tested total 22 different MS2pegR scaffolds, 3 RT variants or genes, 2 MCP variants, and various combinations of the Cas9 nickase, RT, and MCP modules. Our results demonstrated a new strategy for more efficient prime editing and provide a platform for further directed evolution of PEs.

Human-vehicle cooperative driving has become the critical technology of autonomous driving, which reduces the workload of human drivers. However, the complex and uncertain road environments bring great challenges to the visual perception of cooperative systems. And the perception characteristics of autonomous driving differ from manual driving a lot. To enhance the visual perception capability of human-vehicle cooperative driving, this paper proposed a cooperative visual perception model. 506 images of complex road and traffic scenarios were collected as the data source. Then this paper improved the object detection algorithm of autonomous vehicles. The mean perception accuracy of traffic elements reached 75.52%. By the image fusion method, the gaze points of human drivers were fused with vehicles' monitoring screens. Results revealed that cooperative visual perception could reflect the riskiest zone and predict the trajectory of conflict objects more precisely. The findings can be applied in improving the visual perception algorithms and providing accurate data for planning and control.

Bipedal walking is the basis for the humanoid robot to execute all the work. This paper proposed an online learning upper body posture control strategy for humanoid walking. The proposed posture controller is based on real-time adjustment of the hip joint, and keeps upper body posture upright stabilizing the robot. A backpropagation neural network (BPNN) controller is trained using the dataset from different PID controllers. A policy gradient reinforcement learning method is proposed to adjust the network parameters online according to the sensor feedback in different environments. Experiments on a NAO humanoid robot show the effectiveness of the proposed methods.

Developing a friendly and efficient obstacle avoidance algorithm for mobile robot in dynamic environments is challenging in the scenarios where robot plans its paths without observing other obstacles' intents. Recent works have shown the power of deep reinforcement learning techniques to learn collision-free policies. However, many of them ignore the interactions between obstacles, which may cause unnatural trajectory. Meanwhile, the performance of these methods is not well in simulation environment. Combined with the interactions of obstacles, we propose a mobile robot obstacle avoidance algorithm based on deep reinforcement learning in dynamic environments. Firstly, we not only consider the direct impact of obstacles on robots, but also take the interaction between obstacles into account using angel pedestrian grid. Therefore, robot can extract more abundant environmental prior information. Subsequently, the temporal characteristics of obstacles are extracted through attention mechanism, then we can obtain the joint impact of obstacles on the obstacle avoidance strategy. Finally, the robot obtains the control output through value-based reinforcement learning. The feasibility and effectiveness of our proposed algorithm in dynamic environments is verified through simulation experiments.

Toxoplasma gondii is an obligatory intracellular parasite that critically depends on active invasion and egress from infected host cells to complete its propagation cycle. T. gondii rhoptry proteins (TgROPs) are virulent factors associated with host cell invasion, growth. In this study, we analyzed the functions of ROP9 in the process of T. gondii infection. The TgROP9 knockout RH strain (RH△ROP9) and its recovery strain (RH-ReROP9) were constructed using the CRISPR/Cas9 system. The invasion, proliferation, and egress efficiency of the RH△ROP9 strain were evaluated and their pathogenicity to mice was analyzed. Compared with RH wild-type (RH-WT) and RH-ReROP9 strains, the invasion percentage of RH△ROP9 to Vero cells was reduced by about 28.0% (p< 0.01) at 1.5 h, and the relative proliferation percentage was decreased by about 35.0% (p< 0.01) after infection with 102 or 103 parasites. In addition, the RH△ROP9 strain also showed prolonged egress time from host cells. The survival time of the mice (12.6 ± 1.6 or 10.1 ± 1.1 days) were delayed (p < 0.001) after infection with either 200 or 1000 RH△ROP9 parasites. These evidences suggested that ROP9 facilitated T. gondii infection in vitro and in vivo.

Additional file 3: Table S2. Primer sequences used for donor construction and the detection of hDHFR gene identification.

Additional file 4: Table S3. Primer sequences used for Southern blot.

Additional file 8: Table S7. Gene expression of five differentially expressed genes at six time points.

Additional file 5: Table S4. Overview of the RNA-Seq data.

Additional file 2: Table S1. Primer sequences used for sgRNA synthesis.

Additional file 6: Table S5. The annotation of novel transcripts/genes.

In this paper, we are interested in the unbalanced assignment problem with constraints to agents. The modified Hungarian algorithm with dummy tasks and agents are most common methods to solve the unbalanced problem. However, it is impractical in the real scenarios sometimes since some tasks are unassigned actually using the above method. Instead, we propose a graph based twin cost matrices method with improved ant colony optimization algorithm to solve the assignment problem elegantly and uniformly. We start from the generation of the twin cost matrices, and modify the ant colony algorithm to be capable to assign tasks to agents with different numbers. Random mutation is conducted for the stable pheromones to help the ant colony keep diversity during the operation. Experiments demonstrates that using our proposed method to solve the unbalanced assignment problem can stably obtain better results than previous methods.

We propose a optimized lidar odometry and mapping method based on loam algorithm for pose estimation of ground vehicles.Loam algorithm proposed by Zhang Ji has achieved excellent results in 6-DOF lidar motion analysis. However, for ground vehicles, the main motions are the translational motion in the ground and the rotation of Yaw angle.The ground points have no contribution to the analysis of such motions and even cause the error.Based on this,this paper has done additional processing for ground points to get the better results for ground vehicles.In this paper,laser points are segmented to omit some outliesrs and two kinds of features of line and surface are extracted.Similar to the loam algorithm, the 6-DOF laser motion is solved for line, surface features and ground points with Levenberg-Marquardt optimization method. The pose-graph optimization algorithm is used to optimize the loop-closure, which solves the the drift problem of loam algorithm with a long time. On the Kitti data set, we get the very good result.

Research on kart racing games provides an opportunity to develop real autonomous driving controllers using unsupervised learning methods and therefore worth studying. In this paper we bring forward a feasible end-to-end control technique using a novel DQN architecture. Recurrence is introduced to eliminate the effect caused by the partially observable environment. A weighting layer is created to strengthen the impact differences of the lengthy episode. A modified relay mechanism is applied to solve the problem of sparsity and speed up learning. The method is tested and verified under the Mario Kart64 environment. With the screen pixels being the only input during both training and testing period, a direct speed control signal being the output of the system, this method makes the agent perform out similar learning behaviours just like a human player. The resulting agent is able to recognize the environment features well and could act out advanced behaviours which are undiscovered in previous methods.

Toxoplasmosis is a worldwide zoonosis caused by the protozoan parasite Toxoplasma gondii, an obligate intracellular parasite. Currently, no viable vaccine or effective drug strategies exist to prevent and control toxoplasmosis. T. gondii microneme protein 3 (MIC3), rhoptry protein 9 (ROP9), and surface antigen 2 (SAG2) are related to active invasion of the parasite. Hence, we constructed recombinant adenoviruses expressing TgMIC3, TgSAG2, or TgROP9 and evaluated the recombinant adenoviruses as potential vaccines against acute T. gondii infection in BALB/c mice. Mice immunized with the recombinant adenoviruses were measured the antibody levels, percentages of lymphocytes and actived T lymphocytes, cytokine productions, and the survival rates and time to evaluate the protective efficacy. Results showed that immunization with the bivalent or trivalent recombiant adenoviruses could strongly stimulate humoral and cellular immune responses in mice, resulting in effective immune protections against lethal challenge with the tachyzoites of T. gondii RH. These results indicated that the divalent and trivalent adenoviruses, especially Ad-ROP9-MIC3-EGFP, may be promising vaccine candidates against acute T. gondii infection.

Blockchain is the basic technology of bitcoin. With the value appreciation and stable operation of bitcoin, blockchain is attracting more and more attention in many areas. Blockchain has the characteristics of decentralization, stability, security, and non-modifiability. It has the potential to change the network architecture. The consensus algorithm plays a crucial role in maintaining the safety and efficiency of blockchain. Using a right algorithm may bring a significant increase to the performance of blockchain application. In this paper, we reviewed the basic principles and characteristics of the consensus algorithms and analyzed the performance and application scenarios of different consensus mechanisms. We also gave a technical guidance of selecting a suitable consensus algorithm and summarized the limitations and future development of blockchain technology.

This paper presents a novel control method of extended virtual constraint to mimic human movement for a three-link planar robot with moving torso. Inspired by two supine yoga movements, the dynamic model of bipedal walker is modified accordingly, which enlarges application of generalized planar biped robot and provides a more stable and robust walking gait. Due to the continuity of kinematics and discreteness of impact, the walking motion is regarded as a hybrid system, whose zero dynamics determines the stable state of robot. Hence, a within-stride feedback controller is designed based on input-output linearization and extension of virtual constraint. Moreover, poincare return map is adopted to analyze the stability of walking gait. The simulation results demonstrate the validity of proposed control law, leading to asymptotically stable walking with modified planar biped robot.

The zero moment point (ZMP) control method is usually used in stable biped walking. A novel preview control of ZMP is proposed in the paper. First, the dynamics of a biped robot is modeled as a running cart on the table which gives a convenient representation to treat ZMP. Second, the walking pattern generation process is describes in details and a presented preview controller is introduced to reduce zero steady-state error significantly. Third, several patterns in the omnidirectional walking including support-foot and swing-foot are discussed respectively. In order to demonstrate the effectiveness of this approach, the proposed approach is validated via simulations and experiments on a NAO humanoid robot. The webots simulation results verify the feasibility of the proposed control scheme in the stable and robust walking.

Imitation or programming by demonstration from humans is a kind of effective method for robots to learn human's behaviors. This paper proposes a task-prior algorithm to solve the physical kinematic constraints of a full-body humanoid robot on real-time imitation. Redundant kinematic system is built to track the trajectory of the wrist position as a primary task. In the meanwhile, joint velocity limits and self-collision avoidance are introduced into the null space as additional tasks. Experiments on both simulation and a full-body humanoid are designed to verify the validity of the final policy.With the proposed method, the full-body humanoid robot succeeds in imitating humans' movement with Kinect sensor and comparison between with and without considering the constraints are making to show the necessaries of this algorithm.

The Chinese GNP in 1993 was 13.4%, against 12.7% in 1992 and this paper highlights the role of the construction industry in China and pinpoints those areas that will have new legislation following the introduction of the Construction Industry Development Act as well as issues dealing with reform and international practice.

At present, Kalman filtering is widely used in dynamic positioning. With continuous development of deep-sea exploration, the accuracy of Kalman filtering is affected by more complex oceanic condition. As erratic performance of Kalman filtering with fixed system noise matrix and measurement noise matrix in complex oceanic condition, a new filtering method called Kalman filtering based on real-time updating of noise matrix is put forward. The algorithm obtains real-time calculation results of the system noise and measurement noise which are devoted to revising variance of noise array. Compared with Kalman filtering, it is more accurate and more efficacious in improving the performance of dynamic positioning in complex oceanic condition proved by computer simulation.

The standard simulated algorithm has been applied into vehicle routing problem, and it has the common defects of slow convergence and easily being trapped into local minima. In this paper, a new stochastic approach called the simulated annealing genetic algorithm is proposed to solve stochastic vehicle routing problems and the solution is then compared with that from simulated algorithm. Results from case studies show that the proposed algorithm can avoid getting stuck in local minima and has better convergence property and find the optimal or near-optimal solution effectively as well as time and quickly convergence property. So, it is an efficient method for vehicle routing problem.

Ibis paper presents a kind of strategy of task division to process the discontinuous trajectory tracking. Using cubic spline with string parameter to fit the discontinuous trajectory result in: (a) the desired trajectory for macro manipulator is continuous, of which, the discontinuous parts will be tracked by the micro manipulator; and (b) macro-micro manipulator - the redundant system can be regarded as two non-redundant sub-systems. In trajectory planning, a computing algorithm for maximum velocity and maximum acceleration are given based upon trapezoidal velocity profile and torque constraints. Besides tracking the divided trajectory, micro manipulator compensates the trajectory errors caused by macro manipulator. Simulations and experiments of discontinuous trajectory tracking verify the effectiveness of our strategy.

A small manipulator is attached at the tip of a large manipulator, this kind of system can be regarded as a macro-micro manipulator system. This paper presents a macro-micro manipulator system, which uses laser in endpoint operating and can be applied to trajectory tracking. This system has low inertia and wide bandwidth. A multi-stage rotational transformer with thirty pairs of poles as joint angle measurement of macro manipulator and a kind of errors correction strategy is given, which can reduce the measurement errors caused by excitation power. High accuracy and low inertia capacitive position sensor is used for joint angle sensing of micro manipulator. Joint controllers are independent, high computation ability is guaranteed by parallel process network. Fast and reliable data exchange between parallel process network and joint controller is implemented by using peripheral high speed bus chip. Experiments on continuous trajectory and Chinese character tracking showed the effectiveness of the macro-micro manipulator control system.

Combining depth information and color image, D-RGB cameras provide a ready detection of human and associated 3D skeleton joints data, facilitating, if not revolutionizing, conventional image centric researches in, among others, computer vision, surveillance, and human activity analysis. Applicability of a D-RBG camera, however, is restricted by its limited range of frustum of depth in the range of 0.8 to 4 meters. Although a D-RGB camera network, constructed by deployment of several D-RGB cameras at various locations, could extend the range of coverage, it requires precise localization of the camera network: relative location and orientation of neighboring cameras. By introducing a skeleton-based viewpoint invariant transformation (SVIT), which derives the relative location and orientation of a detected humans upper torso to a D-RGB camera, this paper presents a reliable automatic localization technique without the need for additional instrument or human intervention. By respectively applying SVIT to two neighboring D-RGB cameras on a commonly observed skeleton, the respective relative position and orientation of the detected humans skeleton for these two cameras can be obtained before being combined to yield the relative position and orientation of these two cameras, thus solving the localization problem. Experiments have been conducted in which two Kinects are situated with bearing differences of about 45 degrees and 90 degrees; the coverage can be extended by up to 70% with the installment of an additional Kinect. The same localization technique can be applied repeatedly to a larger number of D-RGB cameras, thus extending the applicability of D-RGB cameras to camera networks in making human behavior analysis and context-aware service in a larger surveillance area.