Search

6D object pose estimation is a longstanding computer vision problem. Existing deep learning-based methods have achieved inspiring results in this task. However, large-scale training data with annotations is extremely needed to guarantee these methods’ performance, and acquiring real 6D object pose annotations is fairly labor-intensive and time-consuming. To overcome this drawback, we propose a semi-supervised pose estimation method using labeled synthetic data and unlabeled real data. For unlabeled real data, we form a self-supervised pipeline by minimizing the distance between the input point cloud, which is under ground-truth pose, and the model points transformed based on predicted pose. The labeled synthetic data is used to supervise the network to converge correctly. And we utilize a feature mapping to eliminate the domain gap between the real and synthetic features to further enhance the network’s performance. Moreover, we propose an attention-based pose estimation network, which can concentrate more on the distinguishing features, thus improving the accuracy of pose estimation. Experiments show that our proposed semi-supervised method is able to achieve good performance without the real annotations and outperforms all other methods relying on synthetic data or self-supervision strategy, indicating that the proposed method is effective. [ABSTRACT FROM AUTHOR]

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 parameterization 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. [ABSTRACT FROM AUTHOR]

Real-time and high-performance segmentation is a crucial but challenging perception task for computationally constrained robots, such as the humanoid NAO robot used in the RoboCup Soccer Standard Platform League. However, most existing convolutional neural network (CNN)-based models for semantic segmentation suffer from massive computational costs, which prevents them from being applied to performing real-time inference with a NAO. In this article, we first publish meticulously annotated datasets for training and evaluating semantic segmentation models. Then, we propose a fast downsampling module that downsamples the image while maintaining the spatial information and a novel dense learning module that learns high-level semantic information while recovering the spatial details. Based on these operations, by using a multiscale fusion method to recover the resolution, we propose a more efficient and real-time segmentation model called RoboSeg primarily aimed at offering better speed and accuracy tradeoffs. Finally, to accommodate practical engineering applications, we offer a promising deployment guideline for the CNN model describing how to deploy it on computational resource-limited robots and achieve real-time performance. The experimental results show that the RoboSeg exceeds the state-of-the-art networks in RoboCup scene segmentation: we attain a mean IoU of 87.35% and a pixel accuracy of 96.88% on our dataset using a model that contains only 0.29M parameters and performs just 0.73 GFLOPs. Under the proposed deployment strategies, the network can run at above 30 FPS on NAO robots with downsampled frames. [ABSTRACT FROM AUTHOR]

Medical information is private, and medical data are valuable for medical research. Thus, medical information sharing is challenging because the data might be manipulated improperly and revealed during the operational process. The accuracy and integrity of medical information should be guaranteed throughout the sharing process. Medical institutions require shared information for scientific research and development; however, the issue of privacy inhibits the sharing process. In this article, we propose a new business process and a blockchain-based platform for medical information sharing. Our method exploits the advantages of blockchain in medical information recording and sharing. Information can be stored, shared, and credibly verified among parties in the distributed network. In addition, we propose a new consensus algorithm and a universal anonymous sharing model. These methods improve the efficiency and security of medical information sharing between users. In this way, both the information and the traces of the transaction can be stored in a distributed manner to prevent manipulation and fraud. Consequently, the value of medical information can be fully utilized. [ABSTRACT FROM AUTHOR]

Alternating current (ac) impedance spectra facilitate lithium-ion battery management. Realizing a low-cost and low-complexity onboard impedance measuring system is a vital issue for the management based on the ac impedance. In the article, a novel impedance measuring system combined with a high-power dual active bridge (DAB) converter and distributed sampling units is proposed and verified. The DAB converter is designed to generate the ac disturbance to ensure a quasi-steady measurement of the battery impedance. The distributed signal sampling units simultaneously measure the voltage and current of all the series-connected battery cells in a module to measure their impedance. The measured impedance in a frequency range of 0.1–500 Hz shows the feasibility of the system. The root-mean-square errors of the measured impedance phase from 0.2 to 200 Hz and the magnitude from 0.1 to 500 Hz are less than 6.9% and 4.0%, respectively. The errors in some frequency ranges are slightly larger, which are analyzed. The novelty is reflected in that the system is easily integrated into a bidirectional onboard charger and compatible with the battery management system, thus reducing costs and complexity. It provides a basis for the onboard application of the battery impedance. [ABSTRACT FROM AUTHOR]

This article presents a flexible optical goniometer for the measurement of single-axis joint angular displacement. The design philosophy behind the apparatus lies that the agent's angular displacement is converted to the laser's distance measurement thanks to geometric conversion. We propose a dual-laser ranging model for joint angle measurement with the aid of a reference plane. This model converts the agent's rotational angle to the distance of laser range measurements under triangular constraints. Compared with the commonly used rotary encoders, the proposed sensor shows its competence in easy installation and flexible deployment. Furthermore, a model parameter calibration approach is introduced to decrease the angular measurement error. Extensive validation experiments with a prototype sensor on a robot arm prove the aforementioned claims and show that the prototype sensor can reach an angular precision of 0.3° with a measurement range of −30° – 30°. [ABSTRACT FROM AUTHOR]

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. [ABSTRACT FROM AUTHOR]

This paper investigates the distributed formation control problem of multiple nonholonomic wheeled mobile robots within real environments. The formation pattern of the system adopts leader–follower structure and the communication topology among the multirobot system is modeled by a directed graph. Slippage is hard to avoid due to the possible existence of ice, sand, or muddy roads. To overcome the effect of slippage, an adaptive trajectory tracking controller for leader robot is designed, such that the leader robot can keep up with the virtual reference trajectory and estimate the real value of unknown slipping ratio accurately. In addition, it is difficult for each follower robot to obtain leader’s states in a large formation system, so distributed formation controllers are designed based on distributed observers. It is shown that the proposed controllers can realize formation objective and overcome the slippage constraints at the same time. Finally, the effectiveness of the proposed controllers is verified by simulation results. [ABSTRACT FROM AUTHOR]

In this paper, an online foot position compensator (FPC) is proposed for improving the robustness of humanoid walking based on orbital energy conservation and discrete control Lyapunov function (DCLF), with which the asymptotic stability of the humanoid system can be maintained and, thus, the foot placement control is achieved. The online FPC is developed based on linear model predictive control (MPC) by replanning the trajectories of the center of mass (CoM) and properly placing the footsteps to resist external disturbances and recover the walking posture. To further improve the robustness of the humanoid robots to suppress strong external disturbance, a strategy of upper body posture control is proposed. The presented controller stabilizes the humanoid robot by utilizing hip joints to modulate the upper body posture online. Webots simulations and real experiments on a full-body NAO humanoid robot verify the effectiveness of the proposed methods. [ABSTRACT FROM AUTHOR]

In this article, the problem of steering control based on Takagi–Sugeno (T–S) fuzzy vehicle lateral dynamics is investigated for autonomous driving with nonlinearities, system uncertainties, and actuator constraints. During normal vehicle cruising, the vehicle velocity always changes due to the different road conditions and/or steering wheel maneuvers, and moreover, the vehicle dynamics is also significantly influenced by the tire/road forces under different road surface conditions, which brings many difficulties in steering controller design. By adopting fuzzy modeling techniques and varying look-ahead control strategy, an approach to the T–S fuzzy antiwindup output feedback controller design is proposed for the steering control in path tracking within T–S fuzzy-model-based analysis framework, where the actuator amplitude saturation and rate limit are simultaneously taken into consideration. Finally, valuation results with Carsim/MATLAB joint simulation are shown to demonstrate the effectiveness of the developed methods, and some comparison results in path tracking performance with fixed look-ahead distance control rule and the driver model controller embedded in Carsim are provided, which illustrate the advantages of the developed controller design method. [ABSTRACT FROM AUTHOR]

Blockchain, the underlying technology of digital currency, such as Bitcoin, has the characteristics of decentralization, stability, security, anonymity, and nontampering. Blockchain is being applied in an increasing number of fields. Supply chain finance is a financing model in which banks connect core enterprises with upstream and downstream enterprises to provide flexible financial products and services. In the traditional supply chain finance, the core enterprises with strong competitiveness and a large scale play an irreplaceable role in managing the supply chain information flow, logistics, and capital flow, which results in inequality and information asymmetry. Fraud in supply chain finance is also very serious. To solve the problems of the traditional supply chain finance, in this article, we build a new type of supply chain financial platform that uses blockchain technology to manage the whole process. This supply chain financial platform solves the problem of nontrust among the participants in the supply chain, improves the efficiency of the capital flow and information flow, reduces costs, and provides better financial services to the relevant parties in the supply chain. To protect users' privacy, we propose a new method of using homomorphic encryption in the blockchain to meet the needs of sensitive data privacy protection in supply chain financial scenarios. [ABSTRACT FROM AUTHOR]

In this article, a novel sampling framework with smart exploration and exploitation (sampling-SEE) is proposed for sampling-based optimal motion planning problem. Exploration represents uniform sampling in the configuration space to find a path with better performance than the previous one. Exploitation denotes biased sampling in the surrounding region of the current path to rapidly improve this path. Sampling-SEE fully utilizes the current path cost and location information to decrease the space for exploration and increase the capability for exploitation. Additionally, a biasing ratio is introduced to dynamically adjust the sampling proportion for the two procedures. Sampling-SEE can be easily integrated into the single or batch sampling-based optimal anytime planners to improve the sampling efficiency and accelerate the convergence. Meanwhile, the property of asymptotic optimality remains unchanged. Simulation experiments are carried out in various environments to verify the effectiveness of the proposed framework. [ABSTRACT FROM AUTHOR]

Considering input constraints and parameter uncertainty of the nonholonomic wheeled mobile robots, a new control law has been designed in this paper to solve the tracking and stabilization control problems simultaneously. The asymptotic convergence of the stabilization or tracking errors is achieved by a smooth controller with time-varying feedback parameters subject to set conditions. A geometric-based parameter design strategy is employed to overcome the difficulties resulting from input constraints. With the parameter design strategy, the designed parameters satisfy the set conditions and the control inputs stay in the constrained domain. Experimental studies have been conducted to verify the effectiveness of the proposed control law. [ABSTRACT FROM AUTHOR]

This paper examines the problem of membership-function-dependent controller design for a class of discrete-time T-S fuzzy systems. Based on the partition method of premise variable space, the original T-S fuzzy model is equivalently converted into a piecewise-fuzzy system. Then, by employing some staircase functions, the continuous membership functions are approximated by a series of discrete values via which the information of membership functions is brought into the stability analysis to reduce the design conservatism. With piecewise-Lyapunov functions, the approaches to the piecewise-fuzzy state feedback and observer-based output feedback controller design are proposed, respectively, in terms of linear matrix inequalities such that the closed-loop system is asymptotically stable with a prescribed $\mathcal {H}_{\infty }$ performance level. It is shown that the membership functions of the fuzzy model and fuzzy controllers are not necessarily the same, which allows more design flexibility. Finally, two illustrative examples are provided to show the effectiveness of the developed methods. [ABSTRACT FROM AUTHOR]

In this paper, the problem of steering control is investigated for vehicle path tracking in the presence of parametric uncertainties and nonlinearities. In practice, the vehicle mass varies due to the number of passengers or amount of payload, while the vehicle velocity also changes during normal cruising, which significantly influences vehicle dynamics. Moreover, the vehicle dynamics are strongly nonlinear caused by the tire/road forces under different road surface conditions. With fuzzy modeling method, the original nonlinear path tracking system with parameter variations is first formulated as a T–S fuzzy model with additive norm-bounded uncertainties, and then an approach to the fuzzy observer-based output feedback steering control for vehicle dynamics is proposed under a fuzzy Lyapunov function framework. By employing matrix inequality convexifying techniques, a sufficient condition is developed in the form of linear matrix inequalities such that the closed-loop path tracking error system is asymptotically stable with a guaranteed $\mathcal {H}_{\infty }$ level. Finally, the effectiveness of the proposed fuzzy observer-based output feedback controller is demonstrated in Carsim/Matlab joint simulation environment, via which the advantage of a T–S fuzzy observer-based output controller over the closed-loop driver model embedded in Carsim is also shown with parametric uncertainties and nonlinearities. [ABSTRACT FROM AUTHOR]