借助监督式机器学习（ML）方法，对空间翻滚目标的运动状态预测问题进行研究，为空间机器人抓捕空间翻滚目标提供可靠的数据依据。基于物理模型的运动预测方法依赖理想的建模假设，需要连续的视觉反馈信息，解决目标预测问题的能力有限。因此，本文采用机器学习中纯数据驱动方式的稀疏伪输入高斯过程（SPGP）回归方法进行空间翻滚目标的运动预测。给定空间翻滚目标运动状态的历史观测数据，通过连续优化真实观测数据，得到稀疏的伪训练数据集，进而在线快速预测目标的运动状态，预测的计算效率达到毫秒级。此外，利用马尔科夫链蒙特卡洛（MCMC）法处理连续优化过程，克服由于随机初始值造成的优化过程陷入局部极小值问题。利用Snelson数据验证了所提稀疏伪输入高斯过程回归方法的正确性，并通过4组仿真算例验证了所提方法对于空间翻滚目标运动预测的有效性和鲁棒性。
Based on supervised machine learning (ML), this paper addresses the motion prediction issue of space tumbling targets to provide reliable data of the target motion for space robots when they capture the target. Physics-based motion prediction methods find it hard to solve this problem due to their ideal modelling assumptions and constant requests for vision feedbacks. Hence, a purely data-driven learning-based method, named Sparse Pseudo-input Gaussian Process (SPGP), is employed. Given observed data for the motion state of the space tumbling target, this method continuously optimizes the real data to obtain a sparse pseudo training dataset, making it feasible for a fast online motion prediction implementation with the computational time of prediction within milliseconds. Moreover, the Markov Chain Monte Carlo(MCMC) method is adopted for the continuous optimization, overcoming the local minima problem resulted from the random initial guess during the optimization process. Snelson's data is employed to validate the correctness of the proposed SPGP regression method, and several simulation cases are conducted to demonstrate its effectiveness and robustness.
[1] 王明明, 罗建军, 余敏. 冗余空间机械臂抓捕自旋卫星后的消旋控制[J]. 宇航学报, 2018, 39(5):550-561. WANG M M, LUO J J, YU M. Detumbling control for kinematically redundant space manipulator post-grasping a rotational satellite[J]. Journal of Astronautics, 2018, 39(5):550-561(in Chinese).
[2] 路勇, 刘晓光, 周宇. 空间翻滚非合作目标消旋技术发展综述[J]. 航空学报, 2018, 39(1):012302. LU Y, LIU X G, ZHOU Y. Review of detumbling technologies for active removal of uncooperative targets[J]. Acta Aeronautica et Astronautica Sinica, 2018, 39(1):021302(in Chinese).
[3] HILLENBRAND U, LAMPARIELLO R. Motion and parameter estimation of a free-floating space object from range data for motion prediction[C]//8th International Symposium on Artificial Intelligence, Robotics and Automation in Space, 2005.
[4] BOISSONNAT J D. Geometric structures for three-dimensional shape representation[J]. ACM Tram Graphics, 1984,3(4):266-286.
[5] OBLONSEK C, GUID N. A fast surface-based procedure for object reconstruction from 3D scattered points[J]. Computer Vision and Image Understanding, 1998, 69(2):185-195.
[6] HOPPE H, DEROSE T, DUCHAMP T. Surface reconstruction from unorganized points[J]. ACM Proceedings of Siggraph, 1992, 26(2):71-78.
[7] 吴晟, 孙晟昕, 魏承. 基于机器人柔性毛刷的空间翻滚目标消旋[J]. 航空学报, 2019, 40(5):422587. WU S, SUN S X, WEI C. Tumbling target despun based on robotic flexible brush[J]. Acta Aeronautica et Astronautica Sinica, 2019, 40(5):422587(in Chinese).
[8] HIRZINGER G, LANDZETTEL K, FAGERER C. Telerobotics with large time delays -the ROTEX experiment[C]//IEEE/RSJ International Conference on Intelligent Robots and System, 1994.
[9] KHAMSEH B H, GHORBANI S, JANABI-SHARIFI F. Unscented Kalman filter state estimation for manipulating unmanned aerial vehicles[J]. Aerospace Science and Technology, 2019,92:446-463.
[10] GREENSPAN M, SHANG L, JASIOBEDZKI P. Efficient tracking with the bounded Hough transform[C]//IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2004:520-527.
[11] AGHILI F. A prediction and motion-planning scheme for visually guided robotic capturing of free-floating tumbling objects with uncertain dynamics[J]. IEEE Transactions on Robotics, 2012, 28(3):634-649.
[12] FOKA A F, TRAHANIAS P E. Predictive autonomous robot navigation[C]//IEEE/RSJ International Conference on Intelligent Robots and Systems, 2002.
[13] SUNG R C, DAN F, RUS D. Trajectory clustering for motion prediction[C]//IEEE/RSJ International Conference on Intelligent Robots and Systems, 2012.
[14] PENG H, BAI X L. Improving orbit prediction accuracy through supervised machine learning[J]. Advances in Space Research, 2018,61:2628-2646.
[15] RASMUSSEN E C, WILLIAMS C K I. Gaussian processes for machine learning[M]. Cambridge:The MIT Press, 2005.
[16] 何志昆, 刘光斌, 赵曦晶. 高斯过程回归方法综述[J]. 控制与决策, 2013, 28(8):1121-1129. HE Z K, LIU G B, ZHAO X J. Overview of Gaussian process regression[J]. Control and Decision, 2013, 28(8):1121-1129(in Chinese).
[17] HERAVI J E, KHANMOHAMMADI S. Long term trajectory prediction of moving objects using Gaussian process[C]//First International Conference on Robot, 2011.
[18] KIM E, CHOI S, OH S. A robust autoregressive Gaussian process motion model using l1-norm based low-rank kernel matrix approximation[C]//IEEE/RSJ International Conference on Intelligent Robots and Systems, 2014.
[19] SNELSON L E. Flexible and efficient Gaussian process models for machine learning[D]. London:University of London, 2007.
[20] ANDRIEU C, FREITAS N D, DOUCET A, et al. An introduction to MCMC for machine learning[J]. Machine Learning, 2003, 50:5-43.
[21] GAO Y S. Github[EB/OL].(2013-11-13)[2020-04-27].http:11gitub.com/tyanshuaicao/gp_cholqr.
[22] GULER D C, RAITOHARJU M, PICHE R. Nanosatellite attitude estimation using Kalman-type filters with non-Gaussian noise[J]. Aerospace Science and Technology, 2019, 92:66-76.
CJA