Research on Learning Graphs via Enumerative Queries and its Applications

枚举查询学习图及其应用研究

• 批准号: 20K11998
• 负责人: Parque Victor
• 金额: $ 2.83万
• 依托单位: Waseda University
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Scientific Research (C)
• 财政年份: 2020
• 资助国家: 日本
• 起止时间: 2020-04-01 至 2024-03-31
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-20K11998/
• 关键词: graph learning; optimization; networks; design; robotics; planning; succinct representation; succint representation

The algorithmic foundations and applications of learning and optimization of graphs via enumerative queries in computer-aided design and robotics were proposed: (1) the generation of obstacle-avoiding paths in grid maps by enumerative encoding scheme, the bijection to tree structures, and its rigorous evaluations/comparisons with gradient-free heuristics; (2) the sampling of robot locomotion gaits by enumerative representation schemes and its through evaluations by gradient-free optimization heuristics, enabling the efficient sampling and generation of adaptive gaits for legged robots; (3) the representation of curvature in edges of graph-based membrane folding applications, allowing the efficient folding/unfolding of network-based structures; (4) the generation of robot manipulator trajectories by using graph-based representation and extreme learning schemes, enabling the planning of robot manipulator trajectories with utmost efficiency, in the order of milliseconds; (5) the learning/optimization of cable-driven robot mechanisms by graph-based representations and gradient-free optimization schemes, enabling the possibility to devise new configurations of cable-driven robot structures; and (6) the study on the visualization of resource distribution networks on the plane, enabling the possibility to design optimal networks interactively.

提出了计算机辅助设计和机器人学中基于枚举查询的图的学习和优化的算法基础和应用：(1)通过枚举编码方案生成栅格地图中的避障路径，对树结构进行双射，并与无梯度启发式进行严格的评估/比较；(2)通过枚举表示方案对机器人运动步态进行采样，并通过无梯度优化启发式进行评估，使得能够有效地采样和生成腿部机器人的自适应步态；(3)基于图的薄膜折叠应用的边缘曲率表示，允许基于网络的结构的高效折叠/展开；(4)通过使用基于图形的表示和极限学习方案来生成机器人操作手轨迹，使得能够以最高的效率规划机器人操作手的轨迹，以毫秒级的数量级；(5)通过基于图形的表示和无梯度优化方案来学习/优化缆索驱动的机器人机构，使得能够设计缆索驱动的机器人结构的新构型；以及(6)对平面上的资源分配网络的可视化研究，使得能够交互地设计最优网络。

项目成果

Towards higher order fairness functionals for smooth path planning

• DOI: 10.1145/3449726.3459430
• 发表时间: 2021-07
• 期刊: Proceedings of the Genetic and Evolutionary Computation Conference Companion
• 作者: V. Parque

Electromagnet-Based Three-Dimensional Self-Assembly System for Hierarchical Modular Space Structures

基于电磁铁的分层模块化空间结构三维自组装系统

• DOI: 10.2514/1.a34884
• 发表时间: 2021
• 期刊: Journal of Spacecraft and Rockets
• 影响因子: 1.6
• 作者: Torisaka Ayako;Hasegawa Shoichi;Miura Satoshi;Parque Victor;Miyashita Tomoyuki;Yamakawa Hiroshi;Natori M. C.
• 通讯作者: Natori M. C.

On Path Regression with Extreme Learning and the Linear Configuration Space

极限学习和线性配置空间的路径回归

• DOI: 10.1109/irc55401.2022.00074
• 发表时间: 2022
• 期刊: IEEE International Conference on Robotic Computing (IRC)
• 作者: Victor Parque;Tomoyuki Miyashita
• 通讯作者: Tomoyuki Miyashita

Learning Motion Planning Functions using a Linear Transition in the C-space: Networks and Kernels

使用 C 空间中的线性变换学习运动规划函数：网络和内核

• 发表时间: 2021
• 期刊: Proceedings of the 2021 IEEE 45th Annual Computers, Software, and Applications Conference (COMPSAC)
• 作者: 酒井 麻帆;神野 健哉;成田雅彦;Victor Parque
• 通讯作者: Victor Parque

Packaging of Thick Membranes using a Multi-Spiral Folding Approach: Flat and Curved Surfaces

• DOI: 10.1016/j.asr.2020.09.040
• 发表时间: 2020-12
• 期刊: ArXiv
• 作者: V. Parque;Wataru Suzaki;Satoshi Miura;A. Torisaka;T. Miyashita;M. Natori