Robust Classification of Grasped Objects in Intuitive Human-Robot Collaboration Using a Wearable Force-Myography Device

Kahanowich, Nadav D.; Sintov, Avishai

doi:10.1109/lra.2021.3057794

Cited by 18 publications

(8 citation statements)

References 67 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…People and robots coexist or work together. On the premise of ensuring that robots do not harm humans, it is necessary to explore efficient human–machine cooperation schemes to complement the advantages of humans and robots (Kahanowich, 2021 ). Generally, human motion is obtained in the form of a video or image and then recognized.…”

Section: Research Purpose and Current Situationmentioning

confidence: 99%

Return Strategy and Machine Learning Optimization of Tennis Sports Robot for Human Motion Recognition

et al. 2022

View full text Add to dashboard Cite

At present, there are many kinds of intelligent training equipment in tennis sports, but they all need human control. If a single tennis player uses the robot to return the ball, it will save some human resources. This study aims to improve the recognition rate of tennis sports robots in the return action and the return strategy. The human-oriented motion recognition of the tennis sports robot is taken as the starting point to recognize and analyze the return action of the tennis sports robot. The OpenPose traversal dataset is used to recognize and extract human motion features of tennis sports robots under different classifications. According to the return characteristics of the tennis sports robot, the method of tennis return strategy based on the support vector machine (SVM) is established, and the SVM algorithm in machine learning is optimized. Finally, the return strategy of tennis sports robots under eight return actions is analyzed and studied. The results reveal that the tennis sports robot based on the SVM-Optimization (SVM-O) algorithm has the highest return recognition rate, and the average return recognition rate is 88.61%. The error rates of the backswing, forward swing, and volatilization are high in the return strategy of tennis sports robots. The preparation action, backswing, and volatilization can achieve more objective results in the analysis of the return strategy, which is more than 90%. With the increase of iteration times, the effect of the model simulation experiment based on SVM-O is the best. It suggests that the algorithm proposed has a reliable accuracy of the return strategy of tennis sports robots, which meets the research requirements. Human motion recognition is integrated with the return motion of tennis sports robots. The application of the SVM-O algorithm to the return action recognition of tennis sports robots has good practicability in the return action recognition of tennis sports robot and solves the problem that the optimization algorithm cannot be applied to the real-time requirements. It has important research significance for the application of an optimized SVM algorithm in sports action recognition.

show abstract

Section: Research Purpose and Current Situationmentioning

confidence: 99%

Return Strategy and Machine Learning Optimization of Tennis Sports Robot for Human Motion Recognition

et al. 2022

View full text Add to dashboard Cite

show abstract

“…Proposed by Kahanowich and Sintov (2021), we track the scores of the classes based on the predictions for each sample provided by any chosen classifier. An iterative classification (IC) process to do so is described in Algorithm 1.…”

Section: Iterative Classificationmentioning

confidence: 99%

Simple Kinesthetic Haptics for Object Recognition

Sintov¹,

Ben-David²

2022

Preprint

Self Cite

View full text Add to dashboard Cite

Object recognition is an essential capability when performing various tasks. Humans naturally use either or both visual and tactile perception to extract object class and properties. Typical approaches for robots, however, require complex visual systems or multiple high-density tactile sensors which can be highly expensive. In addition, they usually require actual collection of a large dataset from real objects through direct interaction. In this paper, we propose a kinesthetic-based object recognition method that can be performed with any multifingered robotic hand in which the kinematics is known. The method does not require tactile sensors and is based on observing grasps of the objects. We utilize a unique and frame invariant parameterization of grasps to learn instances of object shapes. To train a classifier, training data is generated rapidly and solely in a computational process without interaction with real objects. We then propose and compare between two iterative algorithms that can integrate any trained classifier. The classifiers and algorithms are independent of any particular robot hand and, therefore, can be exerted on various ones. We show in experiments, that with few grasps, the algorithms acquire accurate classification. Furthermore, we show that the object recognition approach is scalable to objects of various sizes. Similarly, a global classifier is trained to identify general geometries (e.g., an ellipsoid or a box) rather than particular ones and demonstrated on a large set of objects. Full scale experiments and analysis are provided to show the performance of the method.

show abstract

“…The mask and pose of objects are exported by pybullet [12]. To stick as close as possible to real depth images, instead of only adding Gaussian noise to the perfect depth image as in [35], we follow the steps below to generate a more realistic depth image: (1) The values in the image are probabilistically set to 0, and the probability is calculated from the local gradient of the point. The processed image is:…”

Section: B Data Collectionmentioning

confidence: 99%

“…Object grasping is a core issue in the field of robots [1] [2]. Recently, many algorithms have claimed to be effective for handling stacked scenarios as well as grasping novel objects.…”

Section: Introductionmentioning

confidence: 99%

On-Policy Pixel-Level Grasping Across the Gap Between Simulation and Reality

Wang¹,

Chang²,

Liu³

et al. 2022

Preprint

View full text Add to dashboard Cite

Grasp detection in cluttered scenes is a very challenging task for robots. Generating synthetic grasping data is a popular way to train and test grasp methods, as is Dex-net and GraspNet; yet, these methods generate training grasps on 3D synthetic object models, but evaluate at images or point clouds with different distributions, which reduces performance on real scenes due to sparse grasp labels and covariate shift. To solve existing problems, we propose a novel on-policy grasp detection method, which can train and test on the same distribution with dense pixel-level grasp labels generated on RGB-D images. A Parallel-Depth Grasp Generation (PDG-Generation) method is proposed to generate a parallel depth image through a new imaging model of projecting points in parallel; then this method generates multiple candidate grasps for each pixel and obtains robust grasps through flatness detection, force-closure metric and collision detection. Then, a large comprehensive Pixel-Level Grasp Pose Dataset (PLGP-Dataset) is constructed and released; distinguished with previous datasets with off-policy data and sparse grasp samples, this dataset is the first pixellevel grasp dataset, with the on-policy distribution where grasps are generated based on depth images. Lastly, we build and test a series of pixel-level grasp detection networks with a data augmentation process for imbalance training, which learn grasp poses in a decoupled manner on the input RGB-D images. Extensive experiments show that our on-policy grasp method can largely overcome the gap between simulation and reality, and achieves the state-of-the-art performance. Code and data are provided at https://github.com/liuchunsense/PLGP-Dataset.

show abstract

Robust Classification of Grasped Objects in Intuitive Human-Robot Collaboration Using a Wearable Force-Myography Device

Cited by 18 publications

References 67 publications

Return Strategy and Machine Learning Optimization of Tennis Sports Robot for Human Motion Recognition

Return Strategy and Machine Learning Optimization of Tennis Sports Robot for Human Motion Recognition

Simple Kinesthetic Haptics for Object Recognition

On-Policy Pixel-Level Grasping Across the Gap Between Simulation and Reality

Contact Info

Product

Resources

About