仿真环境下不规则物体的高斯过程隐式曲面核函数优化
Optimization of Gaussian Process Implicit Surface Kernel Function for Irregular Objects in Simulation Environment
DOI: 10.12677/mos.2024.133332, PDF,   
作者: 顾浩宇, 张国庆：上海理工大学光电信息与计算机工程学院，上海；上海理工大学机器智能研究院，上海；李清都：上海理工大学机器智能研究院，上海
关键词: 高斯过程；隐式曲面；核函数优化；参数调优；MuJuCo仿真；Gassian Process； Implicit Surface； Kernel Function Optimization； Hyperparameter Tuning； MuJoCo Simulation
摘要: 对于一个物体进行探索时，首要目标就是获取物体的形状，而当针对形状复杂的物体如不规则方块时，隐式曲面往往是最优选择。本文针对MuJoCo仿真平台下的建立的不规则物体，采用RBF (Radial Basis Function)高斯核函数和TPS (Thin Plate Spline covariance)薄板协方差核函数分别建立高斯过程，完成对物体的隐式曲面建模，通过对比，相比于最初的TPS核函数，改用调参后的RBF误差减小了93.07%。通过仿真实验结果表明，对于仿真平台中的不规则物体，高斯过程隐式曲面能够有效通过少量的采集数据完成曲面建模。
Abstract: When exploring an object, the primary objective is to acquire its shape. Implicit surfaces often emerge as the optimal choice, particularly when dealing with complex shapes such as irregular blocks. This paper focuses on the implicit surface modeling of irregular objects established within the MuJoCo simulation platform. We employ Radial Basis Function (RBF) Gaussian kernel and Thin Plate Spline Covariance (TPS) kernel to construct Gaussian processes for object modeling. Through comparison, utilizing the optimized RBF kernel results in a 93.07% reduction in error compared to the initial TPS kernel. Simulation experiment results demonstrate the effectiveness of Gaussian process implicit surfaces in modeling irregular objects within the simulation platform using minimal data collection.
文章引用：顾浩宇, 张国庆, 李清都. 仿真环境下不规则物体的高斯过程隐式曲面核函数优化[J]. 建模与仿真, 2024, 13(3): 3643-3652. https://doi.org/10.12677/mos.2024.133332

参考文献

[1] Allen, P.K. (2012) Robotic Object Recognition Using Vision and Touch. Springer Science & Business Media, Berlin.
[2] Moll, M. and Erdmann, M.A. (2004) Reconstructing the Shape and Motion of Unknown Objects with Active Tactile Sensors. Springer Berlin Heidelberg, Berlin, 293-309. [Google Scholar] [CrossRef] 
[3] 周少山. 隐式曲面建模和体网格生成方法研究[D]: [硕士学位论文]. 武汉: 武汉理工大学, 2018.
[4] Williams, O. and Fitzgibbon, A. (2006) Gaussian Process Implicit Surfaces.
https://www.microsoft.com/en-us/research/wp-content/uploads/2006/06/gpc.pdf
[5] Dragiev, S., Toussaint, M. and Gienger, M. (2011) Gaussian Process Implicit Surfaces for Shape Estimation and Grasping. 2011 IEEE International Conference on Robotics and Automation, Shanghai, 9-13 May 2011, 2845-2850. [Google Scholar] [CrossRef] 
[6] Dragiev, S., Toussaint, M. and Gienger, M. (2013) Uncertainty Aware Grasping and Tactile Exploration. 2013 IEEE International Conference on Robotics and Automation, Karlsruhe, 6-10 May 2013, 113-119. [Google Scholar] [CrossRef] 
[7] Ottenhaus, S., Miller, M., Schiebener, D., et al. (2016) Local Implicit Surface Estimation for Haptic Exploration.2016 IEEE-RAS 16th International Conference on Humanoid Robots (Humanoids), Cancun, 15-17 November 2016, 850-856. [Google Scholar] [CrossRef] 
[8] Rosales, C., Spinelli, F., Gabiccini, M., et al. (2018) GPAtlasRRT: A Local Tactile Exploration Planner for Recovering the Shape of Novel Objects. International Journal of Humanoid Robotics, 15, Article ID: 1850014. [Google Scholar] [CrossRef] 
[9] Jaillet, L. and Porta, J.M. (2017) Path Planning with Loop Closure Constraints Using an Atlas-Based RRT. Springer International Publishing, 100, 345-362. [Google Scholar] [CrossRef] 
[10] 倪天杰, 何良华. 基于形状先验的隐式三维建模算法研究[J]. 计算机科学与应用, 2023, 13(11): 2012-2021.
[11] Watanabe, S. (2023) Tree-Structured Parzen Estimator: Understanding Its Algorithm Components and Their Roles for Better Empirical Performance. arXiv preprint arXiv: 2304.11127.
[12] Todorov, E., Erez, T. and Tassa, Y. (2012) MuJoCo: A Physics Engine for Model-Based Control. 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems, Vilamoura-Algarve, 7-12 October 2012, 5026-5033. [Google Scholar] [CrossRef] 
[13] Zhu, Y., Wong, J., Mandlekar, A., et al. (2020) Robosuite: A Modular Simulation Framework and Benchmark for Robot Learning. arXiv preprint arXiv: 2009.12293.

为你推荐