Instead of traditional ranging radar and camera, Kinect was used to ensure the robot can avoid obstacles in the process of movement. And an obstacle avoidance method of robot was proposed based on admittance control. The relative distance between the robot and human hand was real-time measured by Kinect sensor and converted into a virtual force, acting as an input of the admittance controller to control the next moment speed of the controller to avoid obstacles. Then a modification algorithm for the robot's speed direction was added on the basis of admittance control, when the operator's arm is moving at a high speed, to control the robot's avoid obstacles urgently. At the same time, a method of singular transition to the joint space was proposed to solve the problem that once the robot encountering strange points in Cartesian space it will out of control. Finally, a physical experiment platform based on Kinect sensor and UR robot was set up to conduct dynamic and static obstacle avoidance experiments. The results verified the effectiveness of the proposed method.
The prediction of the human motion plays an important role in the predictive motion controller of human-assisting mobile robot, and precise prediction requires a precise motion model. In this paper, a novel human motion prediction method is proposed. This prediction method is based on a hybrid motion model which combines the inverted pendulum model and constant velocity model. Derivations of the kinematics, especially the inverted pendulum model, are presented in this paper. Estimation of the prediction by inverted pendulum and constant velocity models based on the Unscented Kalman filter(UKF) and Kalman filter(KF) respectively, are also given. The hybrid model precisely describes the walking cycle. Experiments verifying the effectiveness of the proposed model are also presented.
This paper mainly introduces the the control system of micro AUV based on open source hardware. The control system is the most important part of the AUV and its performance has a direct impact on the stability and reliability of AUV. Due to the small size and limited energy, micro AUV puts more restrictions on the volume and power consumption of the control system. In this paper, we choose the BeagleBone Black, the open source hardware, as the main controller, the rich peripheral interface of it facilitates data acquisition and centralized processing of the vehicle. Arduino, another type of open source hardware, is used as the node controller of the control system. Compared with the traditional hardware design, this architecture has the advantages of small volume, low power consumption, low cost, short period of development and maintenance, and it also had good performance in data processing and real-time control. The experimental results showed that the control system based on this architecture run effectively.
Rope-driven rigid-flexible parallel joints are driven by flexible ropes, combining the advantages of both rigid parallel and rope-driven parallel mechanisms. This paper firstly proposes a rope-driven 4SPS-1U rigid-flexible parallel joint mechanism with a flexible spring. Then, the joint mechanism is modeled with forward/inverse position. Secondly, the position solution is numerically analyzed. The work space was simulated and Adams simulation software was used to verify the correctness of the numerical calculation method. The results of this study confirm that the rope-driven rigid-flexible parallel joint mechanism can achieve a variety of movement patterns, which provides a new idea for the innovative design of the robot mechanism.
This paper proposes a new semantic-based framework for real-time detection of arbitrary polygons and circles. In the proposed framework, the first stage is to extract Directional-Cornered Line (DCL) segments for representation of the shape contours. In the second stage, a technique of Directional-Local Searching (DLS) is proposed to group these DCL segments into shape candidates. These shape candidates can be further classified with intuitive semantic criteria. The proposed method is suitable for real-time applications with linear time complexity. Comparative experiments on natural images prove that the proposed method has good performance and the ability to simultaneously locate and detect the polygonal and circular shapes of various objects in real-time.
The command points of tool path generated by the computer-aided manufacturing systems locate near the desired curves within a certain tolerance, so the smooth tool path generation method based on the directive fitting of the command points is not accurate. To overcome this shortage and generate smooth tool path, a tool path correction and compression algorithm is proposed in this paper. Firstly, a method used to compute the parameters of a circle in the three dimension space is given. Then, the command points of tool path was corrected, and the position of new generated command points are given. Secondly, the command points locating in the fitting regions are fitted based on the NURBS curves. The experimental results show that the proposed algorithm can enlarge the regions that can be fitted and generate much smoother tool path.
Baseline error is one of the main error sources that affect the accuracy of InSAR (Interferometric Synthetic Aperture Radar) elevation measurement. When performing DEM (Digital Elevation Model) elevation inversion, high-precision baseline estimation is needed to improve the accuracy of DEM data. This paper proposes a No-GCPs (Ground Control Points) InSAR baseline vector estimation method based on EKF (Extended Kalman Filter). For lunar mapping SAR, it is difficult to obtain elevation control points. This method establishes a function model of the interferometric phase and baseline vector in the reference target region, and traverses and estimates the elevation information of the target region as an unknown quantity. Then, the baseline estimation results are filtered based on the criterion of minimum mean square error, so as to achieve high-precision baseline estimation under the condition of no control points. Finally, the effectiveness of the method is verified by simulation experiments. The results show that the proposed method can achieve high-precision baseline estimation without control points. Key words – InSAR; Control Points; EKF; Lunar Mapping
Detecting and grasping objects in unstructured environments is an important yet difficult task. Fortunately, the breakthroughs from deep convolutional networks stimulate the development of object detection and grasping. The survey aims to serve as a comparison for region-based and region-free detection framework based on deep learning, and supplies the latest research results of object grasping with deep learning. Firstly, we briefly analyze the object detection and grasping. Then, the representative object detection methods based on deep learning are overviewed. Thirdly, we introduce the application of convolutional neural networks in object grasping. Finally, the potential trends in object detection and grasping based on deep learning are discussed.
It is vital for mobile robots to maintain safe and stable operation in environments with slopes, e.g. airports and shopping malls. Different slopes correspond to different accelerations of robots, and the robot also require different driving forces or braking forces to go up or down different slopes. If the robot does not have the ability to recognize of the slope, it cannot generate suitable commands according to the change of slope. That is to say, when going uphill, the lack of driving force may cause the robot to stagnate on the slope; when downhill, insufficient braking force may cause the robot to lose control. To solve such problems, a slope estimation algorithm is proposed in this paper based on RGB-D camera. Firstly, we use the RGB-D camera to obtain the information of the road surface. Secondly, the edge of the slope is detected by Canny Edge Detection Algorithm. Finally, the slope is estimated based on the mathematical geometry. Experimental studies are carried out to evaluate the proposed slope estimation algorithm. Experimental results show that the slope of the road can be obtained in real time and the average slope estimation error is less than 5°.
The goal of this project is to use a new method-hinging hyperplane approximation to solve the problem of discontinuities between track segments and track trajectories. Based on this, a new speed planning method is proposed to solve the problem of uniaxial jumps. Because of the difference in the actual cutting process, the vericut software obtains the cutting amount and calculates the real-time cutting force, and adjusts the speed of each segment according to the cutting force.
To meet the requirement of high-performance motion accuracy and complex trajectory variation in modern industry, a neural network feedforward control framework is proposed in this article. This framework is based on the latest recurrent network structure (RNN) method with gate recurrent unit (GRU). By predicting the tracking error of a normal PID controller using deep GRU network, the proposed framework can achieve equal control performance with iterative learning control (ILC) without any actual iterations. Furthermore, the proposed framework can effectively tackle the problem of trajectory variation which is the widely recognized obstacle for ILC. Comparative experiments between traditional PID, ILC, and the proposed GRU control scheme (GRUC) are carried out on a linear-motor-driven stage. Essentially, the proposed GRUC scheme provides a rather excellent feedforward control scheme without iteration and trajectory repetition, and has good potential in industrial applications.
This paper proposes a neural-network compensation (NNC) strategy for precision contouring motion control of multi-axis motion systems. Firstly, some typical contouring tasks are carried out on a biaxial linear-motor-driven motion system, and the true value of contouring error is obtained by numerical calculation method as the training data of an artificial gated recurrent unit (GRU) neural network. Essentially, the GRU network can be viewed as a data based black-box error model which can capture the characteristics of contouring motion rather accurately. Then, the well trained GRU network can predict contouring error of some tasks those have not been conducted during the training session. Finally, the predicted contouring error is compensated into the reference contour as a kind of feedforward control to improve contouring performance. Comparison between the predicted contouring error and the actual one proves the effectiveness of the proposed GRU neural network. Comparative experiments between NNC and iterative learning control (ILC) validate the excellent nature of the proposed NNC scheme. Actually, NNC is easy to implement and can achieve excellent contouring motion performance as ILC, significantly without need of motion repetition and iteration.
According to the basic principle of defense in depth, the corresponding protection is carried out at different security zones. AS the key technology of perimeter defense, industrial firewall is used between enterprise zone and manufacturing zone. By utilizing an industrial demilitarized zone in industrial firewall, resources are shared between enterprise zone and manufacturing zone without direct traffic. To enhance the cyber security of industrial demilitarized zone, paired firewalls are applied. This paper will investigate the performance of industrial demilitarized zone based on main factors, such as response time, link utilization, throughput and FTP server load using the Riverbed Modeler (academic edition) simulation program.
The application of the intelligent technologies has been a trend in launch site. In the paper, the key elements of traditional launch site are analyzed. On the basis of the research on functional architecture, we present the hierarchical framework for an Intelligent Launch Site (ILS). The framework is composed of the physical layer, perception layer, network layer, and application layer. The physical layer includes physical objects and actuators of the launch site. Perception layer consists of sensors and data processing system. Network layer supplies access gateways and backbone network. Application layer serves application systems through middleware platform. The core of the intelligent system is controller crossing the four layer. Then the hierarchical framework and control system model of the ILS system is described in the formal method. In practice, the architecture and these formal definitions contribute to construction, simulation, and verification for the intelligent system on launch site.
In this paper, we study the cooperative output regulation problem (CORP) of linear heterogeneous multi-agent systems subject to an uncertain leader system. A key component of our control law is the adaptive distributed observer (ADO) for an uncertain leader system proposed recently by ourselves. We first show that this ADO is capable of estimating the unknown parameter vector of the leader system exponentially fast as long as the leader's signal is persistently exciting. Then, we further solve our problem by a distributed measurement output (DMO) feedback control law utilizing the ADO.
In this paper, we propose an approach to fuse various available cues, such as vanishing point of the road, moving direction of the moving target, road edges, road textures, and cues with vertical edges, which extracted only from images of the road scene, to predict the moving direction of the autonomous vehicle. Firstly, the Gaussian models of the moving direction of the autonomous vehicle is constructed by using above cues respectively. Then, the prediction results of different cues are fused under the Bayesian framework to estimate the most reasonable moving direction of autonomous vehicle. We test the algorithm in our campus road scene. Compared with the prediction result of single cue, our fusion algorithm effectively improves the robustness of the prediction, and It has certain reference significance for local navigation and path planning for the autonomous vehicle.
With rapid development of biotechnology, laboratory automation plays a vital role in reducing the manual operation of laboratory personnel and improving the efficiency of scientific research. Especially more and more genomes have been sequenced in recent years, researchers face the task of managing vast clone libraries. As a result of the low picking efficiency and accuracy, manual picking can't meet the high-throughput experimental requirement. Automation picking techniques can reduce the experimental time, and accelerate project progress. This paper presents a novel and simple design of colony picking robot with a multi-pin synchronous manipulator. It can achieve picking, inoculation, cleaning and heating simultaneously, which provides an estimated throughput rate of 2400 colonies/hr. The sterilization method is implemented that the picking needle is sterilized twice in 2 seconds and heated at 400°C. Visual servo control as critical control principle of automated colony pickers is used to identify individual colonies. It helps determine their locations and control the picking tool to pick up these colonies. Through the static analysis of the picking structure in the picking process, it is found that the deformation of the picking structure meets the positioning accuracy of 0.1 mm. The colony picking instrument with multi-pin synchronous manipulator can ensure fast and accurate colony picking in the Automated laboratory system.
The pedestrians in human-robot coexisting environments introduce many challenges to the motion planning of mobile service robots. Besides the problems of minimizing the runtime of the robot navigation, the safety and psychological comfort of the pedestrians should be also guaranteed. This paper presents an optimal sampling-based motion planning algorithm for the human-robot coexisting environments, which models the psychological comfort of the pedestrians by constructing the Estimated Comfort and Collision Risk (ECCR) map of the working environment, and reduces the runtime of the robot navigation by adopting the optimal sampling method in the Informed-RRT* algorithm. The ECCR map combines the pose estimations and the psychological comfort estimations with the static map of the environment. By applying the ECCR map in the motion planning process, the robot can not only avoid collisions with the pedestrians but also keep enough distance to ensure their psychological comfort. The navigation runtime and the lengths of the trajectories are recorded for the evaluations in this work. The experimental results in both simulations and real-world experimental studies reveal that the proposed Risk-Informed-RRT* algorithm can achieve a better performance than the Risk-RRT algorithm in both static and dynamic environments.
Recognizing object affordances is a very important skill, especially for intelligent service robots which serve human users. This paper presents a high-level object affordance recognition approach, which is based on a deep Convolutional Neural Network (CNN). In order to realize object affordance recognition, we build a dataset and utilize a deep CNN to learn high-level object affordances. At last, multiple affordance recognition experiments are implemented on a PR2 platform to validate the feasibility and practicability of the presented approach.
The stereoscopic radiotherapy robot has more and more advantages in the field of radiotherapy because of its high precision and stability. And the respiration tracking technology of tumor motion plays a key role in the precision radiotherapy of stereoscopic radiotherapy robot. Due to the high complexity and individual differences in modeling for tumor motion during respiration, in order to achieve better accuracy and robustness, this paper proposes a method for establishing a correlation model between tumor and thoracic-abdominal surface model based on 3D point cloud, which more fully reflects three-dimensional surface information and tumor motion correlation. For the proposed method, we present a preliminary study in this paper, including a) body surface modeling scheme; b) the method of establishing a correlation model between tumor and thoracic-abdominal surface model; c) experiments to compare the effect of the two modeling methods, namely the point cloud data modeling and external marker modeling to verify the feasibility of using point cloud data modeling to replace the method of external marker modeling. As a preliminary study, the result of experimental verification lays the foundation for the further correlation model building.
