Continuous genetic algorithm for grasping an object of a priori unknown shape by a robotic manipulator

Objectives. The problem of providing the interaction of a robotic manipulator with a priori unknown objects in a given workspace is of great interest both to the research community and many industries. By developing a solution to this problem, it will be possible to reduce the time taken for robots to adapt to new environments and objects therein. One of the primary stages of providing the interaction of the robotic manipulator with objects is the search for the target position of the robot gripper based on the onboard sensor subsystem, which can be carried out by a number of methods. Methods associated with machine learning and self-learning technologies may not be suitable for some applications (for example, during rescue operations) when it is necessary to quickly search for the target position of the gripper for an a priori unknown object, about which there is no relevant information in the robot database. Therefore, for this problem, heuristic approaches – for example, genetic algorithms – seem to be applicable. The objectives of this work are to implement a search based on a continuous genetic algorithm for the target position of the robot gripper including collision avoidance and study its performance under virtual simulation.

Methods. A heuristic search algorithm (continuous genetic algorithm) is used. The complex scene analysis algorithm uses classical image processing methods. In order to evaluate the effectiveness of the algorithm, virtual simulation is used.

Results. The possibility of using a continuous genetic algorithm is analyzed in the problem of grasping an object of an a priori unknown shape avoiding collisions with other objects of a static scene. A complex scene analysis algorithm and implementation of a continuous genetic algorithm are presented for finding the target position of the gripper of a Kuka LBR iiwa 7 R800 robotic control system with redundant kinematics. The results of an experimental virtual simulation of the obtained algorithm are presented.

Conclusions. The conducted research demonstrates the effectiveness of the continuous genetic algorithm in obtaining the target position of the gripper of the robotic manipulator under conditions when the static scene represents randomly located objects of various shapes.

1. Lei Q., Wisse M. Fast grasping of unknown objects using cylinder searching on a single point cloud. In: Ninth International Conference on Machine Vision (ICMV 2016). 2016. V. 10341. https://doi.org/10.1117/12.2268422

2. Sundermeyer M., Mousavian A., Triebel R., Fox D. Contact-GraspNet: efficient 6-DoF grasp generation in cluttered scenes. arXiv [Preprint]. 2021. 7 p. Available from URL: https://arxiv.org/abs/2103.14127

3. Breyer M., Chung J., Ott L., Siegwart R., Nieto J. Volumetric grasping network: Real-time 6 DOF grasp detection in clutter. In: 4th Conference on Robot Learning (CoRL 2020). P. 1602–1611. Available from URL: https://proceedings.mlr.press/v155/breyer21a/breyer21a.pdf

4. Lippiello V., Ruggiero F., Siciliano B., Luigi V. Visual grasp planning for unknown objects using a multifingered robotic hand. IEEE/ASME Transactions on Mechatronics. 2013;18(3):1050–1059. https://doi.org/10.1109/TMECH.2012.2195500

5. Lei Q., Meijer J., Wisse M. A survey of unknown object grasping and our fast grasping algorithm-C shape grasping. In: 2017 3rd International Conference on Control, Automation and Robotics (ICCAR). P. 150–157. https://doi.org/10.1109/ICCAR.2017.7942677

6. Back S., Lee J., Kim T., Noh S., Kang R., Bak S., Lee K. Unseen object amodal instance segmentation via hierarchical occlusion modeling. arXiv [Preprint]. 2021. 8 p. Available from URL: https://arxiv.org/abs/2109.11103

7. Abhijit M., Federico T., Perez-Gracia A. Grasping unknown objects in clutter by superquadric representation. In: 2018 Second IEEE International Conference on Robotic Computing (IRC). P. 292–299. https://doi.org/10.1109/IRC.2018.00062

8. Miller A., Knoop S., Christensen H., Allen P. Automatic grasp planning using shape primitives. In: Proceedings of the 2003 IEEE International Conference on Robotics and Automation. P. 1824–1829. https://doi.org/10.1109/ROBOT.2003.1241860

9. Firman M., Aodha O., Julier S., Gabriel J. Structured prediction of unobserved voxels from a single depth image. In: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). P. 5431–5440. https://doi.org/10.1109/CVPR.2016.586

10. Gridin V.N., Solodovnikov V.I. A continuous genetic data preprocessing algorithm searching coefficients of the approximating function. Novye informatsionnye tekhnologii v avtomatizirovannykh sistemakh = New Information Technologies in Automated Systems. 2018;21:302–306 (in Russ.)

11. Rossi C., Abderrahim M., Diaz J. EvoPose: A modelbased pose estimation algorithm with correspondences determination. 2005 IEEE International Conference Mechatronics and Automation. V. 3. P. 1551–1556. https://doi.org/10.1109/ICMA.2005.1626786

12. Jabalameli A., Behal A. From single 2D depth image to gripper 6D pose estimation: A fast and robust algorithm for grabbing objects in cluttered scenes. Robotics. 2019;8(3):63. https://doi.org/10.3390/robotics8030063

13. Xiang Y., Xie C., Mousavian A., Fox D. Learning RGB-D feature embeddings for unseen object instance segmentation. In: 4th Conference on Robot Learning (CoRL 2020). P. 461–470. Available from URL: https://proceedings.mlr.press/v155/xiang21a/xiang21a.pdf

14. Ronneberger O., Fischer P., Brox T. U-Net: Convolutional networks for biomedical image segmentation. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. 2015. P. 234–241. https://doi.org/10.1007/978-3-319-24574-4_28

15. Dune C., Marchand E., Collowet C., Leroux C. Active rough shape estimation of unknown objects. In: 2008 IEEE/RSJ International Conference on Intelligent Robots and Systems. P. 3622–3627. https://doi.org/10.1109/IROS.2008.4651005