Cloud systems are nowadays more and more present in industry, being part of Industry 4.0 solutions; this can be seen at production level, in manufacturing systems and in robotics. Cloud services are used to improve and interconnect the manufacturing shop-floor processes with the higher-level enterprise components (enterprise resource planning systems, logistics, supply chains and other operational systems). The paper presents a solution for cloud – manufacturing system integration where the cloud system is used to remotely access and control a manufacturing system for research, development and training purpose. The access to the manufacturing system is configured in the cloud as a service, the connection to the manufacturing system being implemented through a set of virtual machines which are deployed and custom configured in cloud. The paper describes the architecture of the system, the deployment scenarios and presents the limitations and performances of the system obtained during the tests.
In this paper we evaluate the accuracy of a contactless 6-DoF master device for a teleoperation platform. The master device is an optical tracking system, i.e. Leap Motion, capable of recognize and track the hand movements. A method to evaluate its accuracy by using a Qualisys motion capture system composed of 8 high-resolution cameras is proposed. In addition, the teleoperation control architecture is presented, where a torque-controlled robot is employed as the slave device, i.e. a 7-DoF Franka Emika robot. The results presented in this paper allow to validate the use of the Leap Motion as an accurate master device in a teleoperation chain while also to evaluate its operational limitations.
This paper addresses the contribution of structural compliance on stiffness and safety of a R-CUBE Haptic Device. Structural compliance is determined in several poses via FEM analysis and addressed by referring to local and global indices of performance. Results are also compared with evidences from experimental tests. Comparison of numerical and experimental data allows to identify and separate the contributions to the overall compliance that are due to the structural stiffness, and other contributions such as joint clearance, pose and loading conditions.
A robot is a flexible tool and handling device. However, by eliminating safety fences in robotic applications, flexibility is limited as regulations on personal safety must be followed. Any modification on the system or application will require a new risk assessment before the system can be put back into operation. This circumstance costs time and money and stands in contradiction to the nature of a robot as a versatile and adaptable device. By introducing safety-rated modification dimensions and determining admissible variations indicating the limits up to which the system or application can be changed in compliance with safety regulations, a potential solution to overcome this restriction is presented. A model-based strategy for estimating and validating the aforementioned modification limits is proposed, up to which changes on the system can be made without conducting a new risk assessment. The proposed approach gives rise to a novel safety concept for collaborative robotic applications, which ensures flexibility while taking into account safety standards.
In this paper, a feedforward neural network is trained towards the generalisation of the inverse kinematics of a 3R metamorhic manipulator with one pseudojoint. A two hidden-layered network is trained with data produced by several anatomies of the metamorphic manipulator and tested to an unforeseen anatomy. The data are separated to aspects per anatomy and combined to one training set. Various configurations of the network is trained using the Levenberg-Marquardt backpropagation method and the best of them is derived. As it will be shown the derived network can achieve very low generalisation errors.
It is anticipated that the field of industrial robot deployment will continue to grow in the coming years. However, an even higher growth rate is expected in deployment of cooperative robots. This means that the area of robotics will be one of the decisive factors in increasing competitiveness on the variable product market. It is logical that such an expansion of robotics cannot be based on solo robotic workplaces. Rather, ever increasing collaboration of different robot types is expected to take place. In co-operation with the SjF, MANEX sees the near future in implementing automation to its production, namely automation of lines and workplaces where industrial robots-cobots-men work together or in a complementary manner. The issue of this integration is relatively new and it is still difficult to find comprehensive methodologies of how to approach it. The article presents a framework proposal for deployment of different robots and the man in the area of final palletizing and packaging. It goes without saying that newer and newer approaches and methodologies and practical experience will be garnered in the future in both, the academia and in practical applications.
Recent studies highlighted the importance of assisting workers for human efforts reduction in manual handling and lifting tasks by using wearable exoskeletons. In this paper, several configurations of a trunk exoskeleton in terms of hinge joint positions are investigated with the attempt to identify the best ones for human efforts reduction. Both human joints loads and interface forces are considered and compared through simulations. The proposed computational approach may be the starting point for the analysis of design and development of effective human assistance devices.
In severe fracture cases, a bone can be completely separated into two fragments. In order to guarantee a re-ossification of the bone, it is mandatory to reposition the bone fragment together. This process requires a delicate surgery called “bone reduction surgery”. The most advanced technique relies on the use of a robots to manipulate the bone fragments with higher precision and stability. The present work introduces the kinematic design of a new hybrid mechanical architecture to perform this task. It is made of one 3-PRP planar mechanism and one 3-RPS tripod mechanism. Its kinematic model is resolved while taking account the tripod parasitic motion. The workspace of this mechanism is then compared to the standard hexapod mechanism that is widely used in bone reduction surgery. It reveals that the proposed mechanism benefits from a larger workspace.
An adaptive multilocator for online recognition and tracking of moving objects is considered. The vision system is based on the fast real time sub image localization module aimed to recognize patterns in order to track the motion of the selected object. In order to reduce mistakes concerned with the identity, the system is able to recognize several similar objects simultaneously. Nevertheless, the operator can point out an object on the screen to correct the tracking process. An adaptation of the actual patterns according to the lastly localized objects in the area of interest is performed in each step of the real time process. The aim of the paper is to present an effective adaptation and correction procedure based on constant primary patterns defined by the operator and varying secondary patterns defined by the automatic adaptation.
A deep encoder-decoder network was previously proposed for learning a mapping from raw images to dynamic movement primitives in order to enable a robot to draw sketches of numeric digits when shown images of same. In this paper, the network architecture, which was previously constructed entirely with fully-connected linear layers, is modified to include convolutional layers in order to improve the image encoder component and make the network more robust to noise. The convolutional layers are pre-trained as part of an MNIST digit classifier and adapted for use in the encoder-decoder network, before the network is trained using a dataset composed of digit images and corresponding writing trajectories. This architecture was tested on several challenging noisy digit datasets and the use of convolutional layers is shown to provide a robust improvement in results.
Speech-based robot instruction is a promising field in private households and in small and medium-sized enterprises. It facilitates the use of robot systems for non-experts as well as experts, even while the user executes other tasks. Considering force-based robot motions, the common approach is to map verbs to robot motions depending on the tool and the work piece. While this method works well for a wide variety of applications, it limits the user to a fixed force for a given manipulation task and does not allow extensions like “hard” or “soft”. To overcome this drawback, we contribute an approach for a defuzzification of uncertain force parameters to numerical robot motions. To proof the reliability of our approach, we apply it on a motion with varying material parameters.
In this paper, a gain scheduled PID force feedback controller of a manipulator is designed and implemented for the robotized sewing of fabrics, using a commercial sewing machine. The proposed manipulator controller should keep a constant tension of the fabric, to achieve high quality of cloths seams, as the length of the fabric is shortened along the sewing process. For the controller design, a non-linear model of the fabric is considered, varying with the actual length between the grasping and sewing points. The model is based on a simplified Kelvin-Voigt model with non-linear spring and damper coefficients, which also depend on the type and the length of the fabric and are estimated experimentally. The Model-based PID tuning process from Simulink is used, for tuning and scheduling the gains of the PID controller corresponding to fabric's different actual lengths. The determined sets of gains are used for controlling on-line an Adept Cobra s800 robot to manipulate a woven piece of fabric during the sewing process, where the proposed approach is tested. The proposed force control approach maintains a stable tensional force on the fabric and therefore the quality of the seam and hence of the cloth could be enhanced. Finally, it is compared with a PID controller with constant gains.
The share of service robots is increasing. Large number of those robots are the professional service robots such as customer service and logistics robots. One of the main requirements of these robots is the ability to navigate in an environment where a GPS system cannot be used. This article investigates existing methods for indoor localization and navigation. It proposes a new complete approach. This approach does not require the integration of a new indoor navigation system. Instead, the existing 24/7 video surveillance infrastructure, as well as the capabilities of the service robot, are considered. The proposed approach can reduce both the initial cost of integrating the robotic system as well as the maintenance costs. Subsequent maintenance can be done entirely remotely. The suggested approach is validated experimentally on a mobile robot.
A common challenge for autonomous mobile ground robots in unstructured environments is the traversal of obstacles without risking to tip over. Previous research on prevention of vehicle tip-over is mostly limited to basic mobility systems with only few degrees of freedom (DOF). In this paper, a novel whole-body motion planning approach is presented. Based on a 3D world model and a given planned path, the trajectories of all joints are optimized to maximize robot stability. The resulting motion plan allows the robot to cross obstacles without tipping over. Compared to existing approaches, the proposed approach considers environment- and self-collisions during planning. Few assumptions about the robot configuration are made which enables the adoption to different mobile platforms. This approach is evaluated for a simulated and a real robot. The platform is a tracked vehicle with adjustable flippers and a five DOF manipulator arm. In several test scenarios, it is shown that the proposed approach effectively prevents tip-over and increases robot stability.
This paper describes the design and operational principles of a device that imparts a well-controlled mechanical force or impulse, a so-called perturbation, to a pre-selected point on the surface of the human body. This perturbator will be integrated within a system aimed at measuring and evaluating human postural reaction in a clinically meaningful way. The ease of use and versatility of the device renders it suitable for manual operation but it can also be integrated in a robotized system. The hardware, control law and characterization of the perturbator are presented. Preliminary results indicate that the device is able to generate repeatable perturbations with characteristics appropriate to the intended application. Further improvements are discussed and proposed.
This paper aims to reduce gearbox errors on industrial robots with a feed forward, neural adaptive control. The algorithm combines two networks, one for control and one for system identification. In order to achieve a high precision and generality on untrained data, a Runge-Kutta Neural Network is used for black-box identification of a nonlinear robot joint. Secondary encoders as additional angle sensors measure the gearbox error and are used for supervised learning. The presented algorithm is capable of online application and reduces gearbox errors in a nonlinear simulation.
Torque-controlled robots are essential for safe human-robot interaction (HRI). In this paper we address the question of how joint level torque control can be implemented and seamlessly integrated into tasks that require precise velocity control. We present a control scheme that takes a desired velocity and torque as input to generate control output driving the joints at this velocity, and simultaneously realizing compliant behavior for safe HRI. We propose a novel method to integrate torque and velocity control into a single controller. The controller consists of an inner torque control loop embedded in an outer velocity control loop, and is hence called Torque-Based Velocity Control (TBVC). Experiments demonstrating the performance of the proposed control scheme are carried out on the humanoid ARMAR-6.
This article presents an approach to communication between robot controllers via RFID transponders, where the partners in communication read and write data into a joint transponder. In some scenarios, cable-based communication, e.g. using connections of digital inputs and outputs or fieldbuses, can be substitute. Using the new approach, it is not necessary to integrate the robot controller into the network of the factory. Since at all times no more than one robot can communicate with the transponder, an effective algorithm of collision avoidance has been implemented. We propose an approach that uses random waiting times for avoiding collisions. The approach is verified by a practical experiment, where two robots hand over a workpiece. Additionally, further scenarios of communication over RFID tags in intelligent workpieces are mentioned.
The paper describes a new swarm based algorithm inspired by the behaviour of tree bats. The algorithm is designed to search through unknown environment and look for objects of interest. Individual robotic swarm agents are equipped with memory to make a more efficient search of the space. The proposed algorithm has been evaluated in comparison with a uniform search. Changing a parameter “shout” changes the behaviour mode of the algorithm. By decreasing the parameter, parallel scan by smaller groups of agents is achieved. A higher value of the parameter leads to a sequential scanning with a large group of agents.
In this paper, we present a novel method called Overlapping Region Concept to extend Matlab's Partial Differential Equation (PDE) Toolbox to apply free-form surface loads during finite element analysis. In our institute, we are developing a toolbox called SG-Library in Matlab to combine multidisciplinary methods to achieve automatic design and manufacturing of medical robots and mechanisms. Recently, we have also integrated the PDE Toolbox into the SG-Library to develop some bionic methods like CAO and SKO to optimize the structure of our robots. During the implementation of these bionic methods, we have encountered the problem of applying free-form surface loads in FEM analysis. The typical workflow of PDE Toolbox separates the surface of a 3D geometry into one or more feature surfaces with sharp boundaries. Boundary conditions can be only applied to such feature surfaces. This is a remarkable drawback for the development of our methods since the surface profile of our model is constantly changing during the shape optimization. Therefore, we need to develop a method to apply free-form surface loads independent from the feature surfaces. In this paper, we use the concept of overlapping region to determine the free-form surface for applying boundary conditions. With this new method, we can achieve robust FEM analysis during the structural optimization process.
