CRII:IIS:Topology Aware Configuration Spaces

CRII:IIS:拓扑感知配置空间

• 批准号: 1850319
• 负责人: Chinwe Ekenna
• 金额: $ 17.41万
• 依托单位: SUNY at Albany
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-06-01 至 2022-05-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1850319&HistoricalAwards=false
• 关键词: CRII; IIS; Topology; Aware; Configuration

Motion planning is an important concept in robotics with a simple definition - find a path for a robot from its start to a goal position, where the start and goal is determined by the task the robot is expected to perform e.g., navigate a factory floor, manipulate an arm to pick up an object etc. This however is difficult to model and compute. To efficiently plan motions neccessitates having complete information about the robot's environment beforehand, then building efficient algorithms that produces a trajectory the robot can utilize; unfortunately, this has proven difficult to achieve. To mitigate this shortcoming, sampling-based algorithms were developed that approximate the information given about the environment. Analysis of these algorithms shows that if a trajectory exists, these sampling algorithms will find such trajectory. The question that arises, though, is how well was the environment information approximated to ensure a high success rate. This project will develop algorithms that better approximate the environment information planning space by producing a measure for this approximation using topology and geometric based formulations. This project will give more control to sampling algorithms to help determine what areas in the environment needs more attention and representation, which will in turn improve on the time needed to perform motion planning.Motivated by the increasing need for motion planning algorithms that are fast and accurate even when they approximate the planning space, and a need to also measure such approximations so as to better manipulate and control the planning process, this project aims to devise a set of formal representations and algorithms to a) better characterize the topology of the planning space, b) combine formal methods to plan and approximate the planning space, c) provide a measure of such approximations and utilize this information to better guide planning in difficult regions, and d) subsequently reduce planning time and memory use which will potentially lead to more real time planning algorithms. In addition, the algorithms and theorems developed in this project will be adaptable to any planning method in any type of environment (dynamic, complex or heterogeneous). This research thus removes a major barrier in the current practice of motion planning where approximate techniques are the state of the art. This project will greatly increase the scalability of motion planning algorithms because it will be topologically rich with information, measure the sampling approximations made in the planning space, and subsequently give a more informed description of these planning spaces. This project will capture in a unique and novel way the topology of the planning space using mathematical formulations such as the Vietoris Rips complex, graph "gluing", and an innovative use of simplicial collapses in planning algorithms.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

运动规划是机器人技术中的一个重要概念，其定义很简单-找到机器人从其起点到目标位置的路径，其中起点和目标由机器人预期执行的任务确定，例如，在工厂车间中导航、操纵手臂拾取物体等。然而，这难以建模和计算。为了有效地规划运动，需要事先获得有关机器人环境的完整信息，然后构建有效的算法，以产生机器人可以利用的轨迹;不幸的是，这已被证明很难实现。为了缓解这一缺点，开发了基于采样的算法，该算法近似于给定的关于环境的信息。对这些算法的分析表明，如果轨迹存在，这些采样算法将找到这样的轨迹。然而，出现的问题是，环境信息的近似程度如何才能确保高成功率。这个项目将开发算法，更好地近似环境信息规划空间，通过使用拓扑和几何基础的公式产生这种近似的措施。该项目将为采样算法提供更多的控制，以帮助确定环境中哪些区域需要更多的关注和表示，这反过来又会改善执行运动规划所需的时间。由于对运动规划算法的需求日益增加，即使在接近规划空间时也能快速准确，并且需要测量这种近似，以便更好地操纵和控制规划过程，该项目旨在设计一组形式表示和算法，以a）更好地表征规划空间的拓扑，B）联合收割机形式化方法来规划和近似规划空间，c）提供这种近似的测量并利用该信息来更好地指导困难区域中的规划，以及d）随后减少规划时间和存储器使用，这将潜在地导致更多的真实的时间规划算法。此外，在这个项目中开发的算法和定理将适用于任何类型的环境（动态，复杂或异构）中的任何规划方法。因此，这项研究消除了一个主要的障碍，在目前的运动规划实践中，近似技术是最先进的。这个项目将大大增加运动规划算法的可扩展性，因为它将是拓扑丰富的信息，测量采样近似在规划空间，并随后给出了更明智的描述这些规划空间。该项目将以独特和新颖的方式捕捉规划空间的拓扑结构，使用数学公式，如Vietoris Rips复合体，图形“胶合”，以及在规划算法中创新地使用单纯形折叠。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

Uncertainty Measured Markov Decision Process in Dynamic Environments

• DOI: 10.1109/icra40945.2020.9197064
• 发表时间: 2020-05
• 期刊: 2020 IEEE International Conference on Robotics and Automation (ICRA)
• 作者: Souravik Dutta;Banafsheh Rekabdar;Chinwe Ekenna

Identifying Valid Robot Configurations via a Deep Learning Approach

通过深度学习方法识别有效的机器人配置

• DOI: 10.1109/iros51168.2021.9636742
• 发表时间: 2021
• 期刊: Identifying Valid Robot Configurations via a Deep Learning Approach
• 作者: Tran, Tuan;Ekenna, Chinwe
• 通讯作者: Ekenna, Chinwe

Approximating Cfree space topology by constructing Vietoris-Rips complex

通过构造 Vietoris-Rips 复合体近似 Cfree 空间拓扑

• 发表时间: 2019
• 期刊: Proceedings of the International Conference on Intelligent Robots and Systems
• 作者: Aakriti, Upadhyay;Weifu, Wang;Chinwe, Ekenna
• 通讯作者: Chinwe, Ekenna