CASIA OpenIR  > 毕业生  > 硕士学位论文
面向多机器人协同搬运的路径规划与编队控制方法研究
蔡奇昂
2024-05-15
Pages82
Subtype硕士
Abstract

多机器人协同搬运是指多机器人系统以互相配合的方式协作完成货物运输的过程。相较于单机器人系统,多机器人系统具有更好的鲁棒性与容错能力,可以通过协作的方式搬运单一机器人难以运输的大件或形状复杂的货物。因此,多机器人协同搬运系统的研究具有十分重大的现实意义。然而,协同搬运任务具有约束条件多,控制任务复杂等特点。同时,协同搬运任务对多机器人系统的协调性也提出了很高要求。以上问题为多机器人系统协同搬运规划与编队控制方法带来了巨大挑战。

本文面向多机器人协同搬运任务场景开展研究,旨在研究多机器人编队路径规划方法以及编队控制方法。本文的主要工作内容与创新点如下:

(1) 针对协同搬运编队队形范围规模约束下的编队全局可通行路径规划问题,提出了一种基于知识数据混合驱动方法的多机器人编队路径规划方法(Neural Formation A*, NFA*)。该方法由特征提取神经网络、地图重构模块以及可微分编队A*搜索模块构成。首先,利用特征提取神经网络从输入的环境地图中提取关键特征,为后续的路径节点搜索提供重要指导信息;其次,利用地图重构模块,通过形态学操作对环境中的障碍物进行扩展,从而构建编队路径规划节点搜索空间;最后,利用可微分编队A*搜索模块,在特征地图与编队搜索空间上搜索可行路径节点。仿真表明,在多种路径规划场景下,NFA*规划的编队路径具有较高的可通行性,同时在路径最优性、节点探索减少率以及可通行性等指标之中取得良好平衡。

(2) 针对协同搬运编队控制过程中不同控制任务建模以及各任务之间协调的问题,提出了一种基于分层二次规划方法的多机器人编队控制方法。首先,针对控制过程中的刚性队形生成、编队跟踪控制以及避障等子控制任务建立对应的优化目标函数与约束条件,将协同搬运编队控制问题转化为最优控制问题;其次,将编队控制过程划分为队形生成阶段与编队跟踪控制阶段,并针对不同阶段选取不同的优化目标与约束条件,使机器人完成不同控制任务;最后,针对编队跟踪控制阶段的多任务协调问题,应用分层二次规划方法,将编队维持任务作为更高优先级的控制任务,约束编队跟踪控制任务的控制量求解,使机器人编队可以在维持刚性编队形状的条件下完成编队跟踪控制任务。仿真结果表明,该方法可以控制机器人在静态或动态环境中以较高精度完成队形生成任务,并在遵守各项约束的条件下完成编队跟踪控制任务。

(3) 针对协同搬运场景中全局路径规划算法与编队控制算法的结合问题,设计了综合仿真模拟协同搬运场景中的规划控制任务。首先利用本文提出的编队全局路径规划算法规划编队中心全局路径。随后将全局路径转化为目标点序列,利用本文提出的编队控制算法控制机器人编队中心跟踪全局路径,完成编队控制任务。仿真结果表明,本文提出的编队全局路径规划算法可以为多机器人编队规划可行路径,使机器人编队可以顺利穿越环境中的狭窄通道;同时,本文提出的编队控制算法可以有效控制多机器人系统在复杂场景中完成队形生成控制任务与编队跟踪控制任务。

综上所述,本文围绕面向多机器人协同搬运的路径规划与编队控制开展研究,深入研究了协同搬运路径规划与编队控制问题。本文提出的知识数据混合驱动的编队路径规划方法以及基于分层二次规划的编队控制方法,为协同搬运系统路径规划与编队控制方法提供了基础与思路。

Other Abstract

Multi-robot cooperative transportation refers to the process where a multi-robot system, under a certain control scheme, cooperatively completes the task of transporting goods. Compared to single-robot systems, multi-robot systems exhibit superior robustness and fault tolerance, which enables them to handle large or complex-shaped objects that are challenging for a single robot to transport. Therefore, research on multi-robot cooperative transportation systems has great practical significance. However, cooperative transportation tasks are characterized by multiple constraints and complex  requirements. Additionally, they impose high demands on the coordination of multi-robot systems. These challenges pose substantial difficulties for the planning and formation control of multi-robot systems.

This dissertation aims to investigate methods for formation path planning and formation control navigation under multi-robot cooperative transportation scenarios. The main contributions and innovations of this paper are as follows:

(1) To deal with the problem of global feasible path planning under constraints of formation size for cooperative transportation formations,  a knowledge-data hybrid driven path planning method named neural formation A* (NFA*) is proposed. First, a feature extraction network is utilized to extract crucial features from the input environment map, which provides guidance for subsequent path node searches. Second, a map reconstruction module is developed to expand regions of obstacles in the environment through morphological operations, which constructs a configuration space for formation path planning. Finally, a differentiable formation A* search module is modified to explore feasible path nodes on combination of the feature map and the configuration space. 
Experimental results demonstrate that NFA* plans paths of formation with high feasibility in various path planning scenarios, meanwhile achieves a good balance among path optimality, node exploration reduction rate, and path feasibility.

(2) To deal with the problem of modeling different control tasks in cooperative transportation and coordinating these tasks with different priorities, a multi-robot formation control method based on hierarchical quadratic programming is proposed. First, optimization objective functions and constraints for each sub-control task are established, including rigid formation generation, formation tracking control as well as obstacle avoidance, which transforms the formation control problem into an optimal control problem. Second, the overall formation control process is divided into two different phases of formation generation and formation tracking control, in which different optimization objectives and constraints are selected to enable robots to accomplish distinct sub-control tasks. Finally, an optimal control method based on hierarchical quadratic programming is developed for the multi-task coordination problem in the formation tracking control phase, which treats the formation maintenance task as a higher-priority control task to constrain the control variables of the formation navigation task. This method enables the robot formation to complete the formation navigation task while maintaining the rigid formation shape. Simulation results demonstrate that this method achieves precise formation generation in static and dynamic environments and accomplishes formation tracking control tasks under strict adherence to constraints.

(3) To address the integration problem of global path planning algorithm and formation control algorithm in cooperative transportation scenario, overall simulation experiments are designed to simulate planning and control tasks in cooperative transportation scenarios. First, the proposed formation global path planning algorithm is employed to plan feasible paths for the formation center. Then, the global path is transformed into a sequence of target points, and the proposed formation control algorithm is utilized to control the formation center of the robots to track the global path and to complete the formation control task. Simulation results demonstrate that the proposed formation global path planning algorithm can plan feasible paths for multi-robot formations, enabling smooth traversal of narrow passages in the environment. Additionally, the proposed formation control algorithm can effectively control multi-robot systems to accomplish formation generation control tasks and formation tracking control tasks in complex environments.

In summary, this dissertation aims to investigate methods for formation path planning and formation control navigation under multi-robot cooperative transportation scenarios. The proposed knowledge-data hybrid driven path planning method and the formation control navigation method based on hierarchical quadratic programming provide foundations and insights for future researches on path planning and formation control methods in cooperative transportation systems.

Keyword多机器人系统 协同搬运 知识数据混合驱动 路径规划 编队控制 最优控制
Subject Area电子、通信与自动控制技术 ; 机器人控制
MOST Discipline Catalogue工学::控制科学与工程
Language中文
IS Representative Paper
Sub direction classification智能控制
planning direction of the national heavy laboratory多智能体决策
Document Type学位论文
Identifierhttp://ir.ia.ac.cn/handle/173211/57155
Collection毕业生_硕士学位论文
Corresponding Author蔡奇昂
Recommended Citation
GB/T 7714
蔡奇昂. 面向多机器人协同搬运的路径规划与编队控制方法研究[D],2024.
Files in This Item:
File Name/Size DocType Version Access License
蔡奇昂_面向多机器人协同搬运的路径规划与(6471KB)学位论文 限制开放CC BY-NC-SA
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[蔡奇昂]'s Articles
Baidu academic
Similar articles in Baidu academic
[蔡奇昂]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[蔡奇昂]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.