WO2020181729A1 - 一种基于分布式并行计算的路径规划方法及其系统 - Google Patents

一种基于分布式并行计算的路径规划方法及其系统 Download PDF

Info

Publication number
WO2020181729A1
WO2020181729A1 PCT/CN2019/102439 CN2019102439W WO2020181729A1 WO 2020181729 A1 WO2020181729 A1 WO 2020181729A1 CN 2019102439 W CN2019102439 W CN 2019102439W WO 2020181729 A1 WO2020181729 A1 WO 2020181729A1
Authority
WO
WIPO (PCT)
Prior art keywords
message
mobile robot
path
path planning
data
Prior art date
Application number
PCT/CN2019/102439
Other languages
English (en)
French (fr)
Inventor
吕太之
冯茂岩
赵涛
张军
陈勇
孙炯宁
Original Assignee
江苏海事职业技术学院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 江苏海事职业技术学院 filed Critical 江苏海事职业技术学院
Publication of WO2020181729A1 publication Critical patent/WO2020181729A1/zh
Priority to ZA2021/07292A priority Critical patent/ZA202107292B/en

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
    • G01C21/20—Instruments for performing navigational calculations
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
    • G01S17/02—Systems using the reflection of electromagnetic waves other than radio waves

Definitions

  • the invention relates to applications in the field of autonomous navigation of mobile robots, in particular to a path planning method and system based on distributed parallel computing.
  • Robotics is one of the most active areas of high-tech research today. The research of mobile robots involves multiple disciplines and has been widely used in different fields. Autonomous navigation is a basic function that a mobile robot should possess, and the execution efficiency of path planning is a key factor restricting whether a mobile robot can navigate autonomously.
  • the traditional path planning method relies on the calculation of the mobile robot's own computing power, while path planning is a typical computationally intensive task and requires high requirements for airborne equipment.
  • cloud computing technology has begun to be applied in the field of robotics.
  • the intensive calculation of autonomous navigation is executed in the cloud, and the path is planned by distributed parallel computing through the streaming computing big data platform. This can make up for the lack of airborne capacity and overburden of mobile robots during autonomous navigation, and improve mobile robots The efficiency and accuracy of path planning.
  • the execution of path planning in the cloud reduces the performance requirements of mobile robots, and can unload part of the airborne equipment to improve the mobile portability of the robot.
  • the purpose of the present invention is to provide a path planning method and system based on distributed parallel computing for improving the speed of mobile robot path planning and making up for the lack of airborne capacity during path planning of mobile robots.
  • a path planning method based on distributed parallel computing of the present invention includes the following steps:
  • Step 1 Data collection
  • the mobile robot regularly reads data from the lidar sensor through the data acquisition module, then formats it, and stores the data in the local database;
  • Step 2 Message transmission
  • the mobile robot encapsulates the sensor data as a message, and calls the remote interface to submit the message to the cloud;
  • Step 3 Path planning
  • Distributed parallel computing path planning method take out the message from the message queue, group the lidar data, each group generates an environment map in parallel, merges the environment map, constructs the search interval through the starting point connection, and improves the A * algorithm in parallel construction.
  • View and search path encapsulate the combined path as a message and return to the mobile robot;
  • Step 4 Path control
  • the mobile robot obtains the returned message and generates a control sequence according to the planned path to make the mobile robot move toward the target point.
  • step 3 includes the following steps:
  • Step 3a Take out the message from the message queue, group the radar data, and distribute it to multiple computing nodes;
  • Step 3b Parallel environment map generation, fit the grouped data into line segment groups at each computing node, and expand into polygons according to the anti-collision radius of the mobile robot;
  • Step 3c Receive the polygon information transmitted by the last computing node, construct the starting point connection, take out two vertices, and construct the interval path search task;
  • Step 3d At each computing node, use the improved A * algorithm based on the polygon to construct a viewable view and find the optimal path in the interval in parallel;
  • Step 3e Combine paths and take out the global optimal path according to the evaluation function
  • Step 3f Encapsulate the global optimal path as a message and return it to the mobile robot.
  • step 1 includes the following steps:
  • Step 1a After the mobile robot is started, load the lidar driver
  • Step 1b According to the sampling period, call the corresponding driving method to collect data
  • Step 1c Format and store in the local database.
  • step 1 includes the following steps:
  • Step 1a When the mobile robot starts, load the supported sensors and communication protocols and register them in the virtual table, which includes the lidar type, data format, communication method and parameters;
  • Step 1b After the mobile robot is connected to the radar, it recognizes the communication protocol and the lidar type according to the connection information and calls the virtual table registration method to obtain lidar data;
  • Step 1c Format the lidar data and store it in the local database of the mobile robot.
  • step 2 includes the following steps:
  • Step 2a The mobile robot encapsulates the lidar data and the target point into a message, and uploads it to the cloud through the IOT gateway;
  • Step 2b The cloud stores the received message in the message queue.
  • a path planning system based on distributed parallel computing of the present invention includes a mobile robot terminal and a cloud;
  • the mobile robot system runs on an embedded operating system, and the mobile robot terminal includes four parts: a message collection module, a path management module, an IoT gateway, and a message client;
  • the information collected by the message collection module comes from sensors, including lidar and inertial sensors;
  • the cloud includes four parts: message service cluster module, path planning module, distributed storage system and application coordination service module.
  • the mobile robot end includes
  • Lidar data acquisition module used to collect lidar data and format it into a unified form
  • the path management module that converts the received path into a control sequence and controls the mobile robot to move toward the target;
  • IoT gateway used for message encapsulation, decapsulation and forwarding
  • the client component used to push messages to the cloud message queue and get messages from the message queue.
  • the cloud includes
  • a message service cluster module that realizes high-availability and load-balanced message service through a cluster
  • a distributed storage system that realizes the storage of mobile robot message data and map data in a distributed manner
  • Path planning module to realize distributed parallel path planning through distributed parallel computing cluster
  • the application coordination service module that realizes the coordination, monitoring and management between the main control node and N computing nodes in the computing cluster.
  • the path planning speed of the mobile robot of the present invention is high, and the onboard capacity of the mobile robot is sufficient during path planning.
  • Combining an improved path planning algorithm with cloud computing technology, and using distributed parallel processing based on streaming computing technology makes the present invention has the advantages of simplicity, safety, high efficiency, high accuracy, and easy implementation of software and hardware, and can realize mobile robots Real-time path planning.
  • FIG. 1 is a diagram of the path planning method and equipment structure of the distributed parallel computing of the present invention
  • Figure 2 is a flow chart of the path planning method for distributed parallel computing of the present invention
  • Fig. 3 is a flow chart of real-time acquisition of mobile robot lidar data of the present invention.
  • Figure 5 is an analysis flow chart of the distributed parallel computing path planning method of the present invention.
  • Fig. 6 is a flow chart of the parallel view construction and path finding method based on the improved A * algorithm of the present invention
  • FIG. 7 is a diagram of experimental results of the path planning method of distributed parallel computing in the laboratory corridor of the present invention.
  • Fig. 8 is the display interface of the path planning method of distributed parallel computing of the present invention on the smart phone side.
  • a path planning system based on distributed parallel computing of the present invention the path planning system in autonomous navigation of mobile robots includes a mobile robot end and a cloud;
  • the mobile robot system runs on an embedded operating system, and the mobile robot terminal includes four parts: a message collection module, a path management module, an IoT gateway, and a message client;
  • the information collected by the message collection module comes from sensors, including lidar and inertial sensors;
  • the cloud includes four parts: message service cluster module, path planning module, distributed storage system and application coordination service module.
  • the mobile robot end includes
  • Lidar data acquisition module used to collect lidar data and format it into a unified form
  • the path management module that converts the received path into a control sequence and controls the mobile robot to move toward the target;
  • IoT gateway used for message encapsulation, decapsulation and forwarding
  • the client component used to push messages to the cloud message queue and get messages from the message queue.
  • the cloud includes
  • a message service cluster module that realizes high-availability and load-balanced message service through a cluster
  • a distributed storage system that realizes the storage of mobile robot message data and map data in a distributed manner
  • Path planning module to realize distributed parallel path planning through distributed parallel computing cluster
  • the application coordination service module that realizes the coordination, monitoring and management between the main control node and N computing nodes in the computing cluster.
  • the distributed parallel computing cluster in Figure 1 is implemented on the Apache Storm platform, a large-scale distributed streaming computing system, to provide distributed parallel computing capabilities; Hadoop HDFS provides a distributed file system for path planning to complete the path And the storage of lidar information; application coordination services are implemented through Zookeeper, providing consistent services for path planning distributed computing.
  • the main control node is implemented by Nimbus in the Storm platform.
  • the worker node is responsible for performing computing tasks, running the Supervisor monitoring process on each worker node, accepting the tasks assigned by Nimbus, and starting and stopping the work processes under their own management.
  • the work process will start two types of components, one is Spout, which is a component that generates a source data stream; the other is Bolt, which is a component that receives data and then performs processing.
  • the mobile robot deploys ROS (Robot Operating System), which encapsulates the robot's hardware and provides the same expression for upper-level applications.
  • ROS Robot Operating System
  • a path planning method and system based on distributed parallel computing includes the following steps: data collection, message transmission, distributed parallel path planning and path control.
  • a path planning method and equipment based on distributed parallel computing includes the following steps:
  • Step 1 the mobile robot is configured with a single-line lidar, as shown in Figure 3 to realize lidar data collection.
  • the mobile robot starts, it completes the initialization of the virtual table, and processes different types of lidar data in a unified way.
  • the robot judges the sensor status every 10ms. When the status is available, it reads the lidar data. After reading the lidar data for one scan cycle, the data is formatted as follows and stored in the local data database;
  • Step 2 Message transmission
  • the mobile robot encapsulates the lidar data and the target point as a message, and transmits it to the message client through the IOT gateway; the message client, as shown in Figure 4, sends the message to the cloud.
  • messaging is implemented through Kafka, a high-throughput distributed publish-subscribe messaging system.
  • the cloud message cluster server stores the received message in the message queue.
  • Step 3 Distributed parallel path planning
  • the distributed parallel computing path planning method is shown in Figure 5.
  • the search interval is constructed by connecting lines, and the viewable and search paths are constructed in parallel through the improved A * algorithm. After the paths are combined, they are encapsulated as a message and returned to the mobile robot.
  • Step 3a This step is implemented by the ReadTask task in Figure 5, which is implemented by the Spout component. This task takes out the message from the message queue and traverses the distance array data. When the distance between the neighboring points exceeds the anti-collision radius R of the mobile robot, it is regarded as a new group, and the grouping information is sent to multiple groups in a tuple. calculate node.
  • Step 3b This step is completed by multiple MapTasks in Figure 5 in parallel.
  • the MapTask is implemented by the Bolt component. This task fits the grouped data into a line segment group by the least square method at each node, and each group of line segments expands into a polygon.
  • Steps 3c-3d The task is executed in parallel by multiple SVGATasks in Figure 5 and implemented by the Bolt component.
  • Each computing node realizes a certain interval path search, and uses an improved A * algorithm based on polygons to construct a view and find the optimal path in parallel.
  • the process of improving the A * algorithm search is shown in Figure 6.
  • the evaluation function is defined as follows:
  • gn represents the actual path length from the starting node to the current node
  • hn is the heuristic function, which is the estimation of the shortest path from the current node to the target node.
  • MAX represents the maximum possible distance from the starting point to the target point
  • vc represents the number of visits to the node, which can ensure that the same node will not be visited repeatedly.
  • the second is to explore the path from the current node to the target point. Construct a line from the current node to the target point. If the line is in the visible view or does not conflict with the obstacle, it means that the path has been found, the path is output, and the algorithm ends. If the line conflicts with the obstacle, add the node to the CLOSED table, and add the vertices of the obstacle being crossed to the OPEN table.
  • Step 3f This step is implemented by the CombineTask task, and the optimal path is encapsulated as a message through the kafka interface and returned to the mobile robot.
  • Step 4 Path control
  • the mobile robot obtains the returned message, and generates a control sequence for the planned path, so that the mobile robot moves toward the target point.
  • Figure 7 shows that in the corridor of the laboratory, the robot reaches the destination point from the starting point in the figure.
  • the mobile robot uses a path planning algorithm based on distributed parallel computing to achieve real-time path planning, avoiding obstacles and reaching the destination point safely.
  • Figure 8 shows the display interface on the smart phone, which contains the line segment information scanned by the lidar and the physiological information of the operator.
  • the present invention combines an improved path planning algorithm with cloud computing technology. Based on cloud computing technology, the use of distributed parallel processing enables the present invention to have the advantages of high efficiency, high precision and easy software and hardware implementation, and can realize real-time path planning of mobile robots. .

Landscapes

  • Engineering & Computer Science (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Electromagnetism (AREA)
  • Automation & Control Theory (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
  • Traffic Control Systems (AREA)

Abstract

一种基于分布式并行计算的路径规划方法及其系统，该方法包括如下步骤：移动机器人通过数据采集模块定时从激光雷达传感器读取数据，然后进行格式化处理，数据存储到本地数据库；移动机器人将传感器数据封装为消息，调用远程接口将消息提交到云端；云端执行分布式并行计算路径规划方法，从消息队列中取出消息，将激光雷达数据分组，每个分组并行生成环境地图，通过起始点连线构建搜索区间，通过改进A *算法并行构造可视图和查找路径，合并路径后封装为消息返回给移动机器人；移动机器人获取返回路径，朝目标点移动。该路径规划方法速度快，解决了移动机器人路径规划时自身机载能力不足的问题。

Description

一种基于分布式并行计算的路径规划方法及其系统 技术领域
本发明涉及移动机器人自主导航领域的应用，具体为一种基于分布式并行计算的路径规划方法及其系统。
背景技术
机器人学是当今高技术研究最活跃的领域之一，移动机器人的研究涉及多个学科，在不同的领域得到了广泛的应用。自主导航是移动机器人应具备的基本功能，路径规划的执行效率是制约移动机器人是否能自主导航的关键因素。
传统的路径规划方法是依赖移动机器人自身算力的进行计算，而路径规划是典型的计算密集型任务，对机载设备要求高。受益于网络数据传输速率的快速增长，云计算技术在机器人领域开始得到应用。将自主导航密集计算的部分放到云端执行，通过流式计算大数据平台，采用分布式并行计算规划路径，可以弥补移动机器人自主导航时候机载能力不足和负担过重的短板，提高移动机器人路径规划的效率和精度。同时在云端执行路径规划降低了移动机器人性能要求，可以对部分机载设备进行卸载，提高机器人的移动便携性能。
发明内容
本发明的目的在于提供一种用于提高移动机器人路径规划的速度，弥补移动机器人路径规划时候机载能力不足的基于分布式并行计算的路径规划方法及其系统。
本发明为解决上述技术问题采用以下技术方案：本发明的一种基于分布式并行计算的路径规划方法，包括如下步骤：
步骤1：数据采集，
移动机器人通过数据采集模块定时从激光雷达传感器读取数据，然后进行格式化处理，数据存储到本地数据库；
步骤2：消息传输，
移动机器人将传感器数据封装为消息，调用远程接口将消息提交到云端；
步骤3：路径规划，
分布式并行计算路径规划方法，从消息队列中取出消息，将激光雷达数据分组，每个分组并行生成环境地图，合并环境地图，通过起始点连线构建搜索区间，通过改进A *算法并行构造可视图和查找路径，合并路径后封装为消息返回给移动机器人；
步骤4：路径控制，
移动机器人获取返回的消息，根据规划路径生成控制序列，使得移动机器人朝目标点移动。
进一步地，上述步骤3包括如下步骤：
步骤3a：从消息队列中取出消息，将雷达数据进行分组，分发给多个计算节点；
步骤3b：并行环境地图生成，在每个计算节点将分组数据拟合为线段组，根据移动机器人防碰撞半径膨胀为多边形；
步骤3c：接收上一计算节点传递的多边形信息，构造起始点连线取出两个顶点，构建区间路径搜索任务；
步骤3d：在每个计算节点，基于多边形使用改进A *算法并行构造可视图和 查找区间最优路径；
步骤3e：合并路径，根据评价函数取出全局最优路径；
步骤3f：将全局最优路径封装为消息返回给移动机器人。
进一步地，上述步骤1包括如下步骤：
步骤1a：移动机器人启动后，加载激光雷达驱动；
步骤1b：根据采样周期调用对应驱动方法采集数据；
步骤1c：格式化存储到本地数据库中。
进一步地，上述步骤1包括如下步骤：
步骤1a：当移动机器人启动，加载支持的传感器和通信协议注册到虚拟表中，表中包括激光雷达类型、数据格式、通信方法和参数；
步骤1b：当移动机器人连接到雷达后，根据连接信息识别出通信协议和激光雷达类型调用虚拟表注册的方法获取激光雷达数据；
步骤1c：将激光雷达数据格式化，存储到移动机器人的本地数据库中。
更进一步地，上述步骤2包括如下步骤：
步骤2a：移动机器人将激光雷达数据和目标点封装为消息，通过IOT网关上传到云端；
步骤2b：云端将接收到的消息存储到消息队列中。
本发明的一种基于分布式并行计算的路径规划系统，所述的移动机器人自主导航中的路径规划系统包括移动机器人端和云端；
所述移动机器人系统运行在嵌入式操作系统上，移动机器人端包括四个部分：消息采集模块、路径管理模块、IoT网关和消息客户端；
消息采集模块采集的信息来源于传感器，包括激光雷达和惯性传感器；
云端包括四个部分：消息服务集群模块、路径规划模块、分布式存储系统和 应用协调服务模块。
进一步地，所述的移动机器人端包括
用于采集激光雷达数据，将其格式化为统一的形式的激光雷达数据采集模块；
将收到的路径转换为控制序列，控制移动机器人朝目标移动的路径管理模块；
用于消息的封装、解封以及转发的IoT网关；
用于将消息推送到云端消息队列中以及从消息队列中获取消息的客户端组件。
进一步地，所述的云端包括
通过集群方式实现高可用和负载均衡的消息服务的消息服务集群模块；
通过分布式方式实现移动机器人消息数据和地图数据的存储的分布式存储系统；
通过分布式并行计算集群实现分布式并行路径规划的路径规划模块；
实现计算集群中主控节点、N个计算节点之间的协调、监控和管理的应用协调服务模块。
有益效果：本发明移动机器人路径规划的速度高，移动机器人路径规划时候机载能力足。将一种改进的路径规划算法和云计算技术相结合，基于流式计算技术使用分布式并行处理使得本发明具有简便安全、高效的优点、精度高、易于软硬件实现的优点，可以实现移动机器人实时路径规划。
附图说明
图1为本发明的分布式并行计算的路径规划方法及其设备架构图；
图2为本发明的分布式并行计算的路径规划方法流程图；
图3为本发明的移动机器人激光雷达数据实时采集流程图；
图4为本发明的移动机器人激光雷达信息实时上传流程图；
图5为本发明分布式并行计算路径规划方法分析流程图；
图6为本发明基于改进A *算法的并行可视图构造和路径查找方法流程图；
图7为本发明分布式并行计算的路径规划方法在实验室走廊实验结果图；
图8为本发明分布式并行计算的路径规划方法在智能手机端的显示界面。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
如图1所示，本发明的一种基于分布式并行计算的路径规划系统，所述的移动机器人自主导航中的路径规划系统包括移动机器人端和云端；
所述移动机器人系统运行在嵌入式操作系统上，移动机器人端包括四个部分：消息采集模块、路径管理模块、IoT网关和消息客户端；
消息采集模块采集的信息来源于传感器，包括激光雷达和惯性传感器；
云端包括四个部分：消息服务集群模块、路径规划模块、分布式存储系统和应用协调服务模块。
所述的移动机器人端包括
用于采集激光雷达数据，将其格式化为统一的形式的激光雷达数据采集模块；
将收到的路径转换为控制序列，控制移动机器人朝目标移动的路径管理模块；
用于消息的封装、解封以及转发的IoT网关；
用于将消息推送到云端消息队列中以及从消息队列中获取消息的客户端组件。
所述的云端包括
通过集群方式实现高可用和负载均衡的消息服务的消息服务集群模块；
通过分布式方式实现移动机器人消息数据和地图数据的存储的分布式存储系统；
通过分布式并行计算集群实现分布式并行路径规划的路径规划模块；
实现计算集群中主控节点、N个计算节点之间的协调、监控和管理的应用协调服务模块。
实施例1
在本实施例中，图1中的分布式并行计算集群通过大规模的分布式流式计算系统Apache Storm平台实现，提供分布式并行计算能力；Hadoop HDFS为路径规划提供分布式文件系统，完成路径和激光雷达信息的存储；应用程序协调服务通过Zookeeper实现，为路径规划分布式计算提供一致性服务。主控节点在Storm平台中由Nimbus实现。工作节点负责执行计算任务，在每个工作节点上运行Supervisor监控进程，负责接受Nimbus分配的任务，启动和停止属于自己管理的工作进程。工作进程会启动两种类型的组件，一类是Spout，是产生源数据流的组件；一类是Bolt，是接受数据然后执行处理的组件。移动机器人部署ROS(Robot Operating System)，封装了机器人的硬件，对上层应用提供了相同的表达方式。
如图2所示，一种基于分布式并行计算的路径规划方法及其系统，处理步骤分为以下四个流程：数据采集、消息传输、分布式并行路径规划和路径控制。一种基于分布式并行计算的路径规划方法及其设备，包括以下步骤：
步骤1：本实施例中，移动机器人配置的是单线激光雷达，如图3所示实现激光雷达数据采集。移动机器人在启动的时候，完成虚拟表初始化，将不同类型的激光雷达数据以统一的方式处理。机器人每隔10ms判断传感器状态，当状态可用时，读取激光雷达数据，完成一个扫描周期的激光雷达数据读取后，将数据格式化如下的形式，并存储到本地数据数据库；
Figure PCTCN2019102439-appb-000001
步骤2：消息传输，
移动机器人将激光雷达数据和目标点封装为消息，通过IOT网关传递给消息客户端；消息客户端如图4所示，将消息发送给云端。这里消息传递通过高吞吐量的分布式发布订阅消息系统Kafka实现。云端消息集群服务器将接收到的消息存储到消息队列中。
步骤3：分布式并行路径规划，
分布式并行计算路径规划方法如图5所示，首先从消息队列中取出消息，将激光雷达数据分割为若干组，每个分组并行生成包含多边形顶点信息的环境地图，合并环境地图，通过起始点连线构建搜索区间，通过改进A *算法并行构造可视图和查找路径，合并路径后封装为消息返回给移动机器人。
步骤3a：该步骤由图5中的ReadTask任务实现，该任务通过Spout组件实现。该任务从消息队列中取出消息，遍历距离数组数据，当临近点的距离超过移动机器人防碰撞半径R，将其作为一个新的分组，将分组信息以元组(tuple)的方式发送到多个计算节点。
步骤3b：该步骤由图5中的多个MapTask并行完成，MapTask通过Bolt组件实现，该任务在每个节点将分组数据通过最小二乘法拟合为线段组，每组线段膨胀为一个多边形。
步骤3c-3d：任务由图5中的多个SVGATask并行执行，通过Bolt组件实现。每个计算节点实现某区间路径搜索，基于多边形使用改进A *算法并行构造可视图和查找最优的路径。改进A *算法搜索的流程如图6所示。
改进A *算法搜索的流程如下：
(1)初始化。将目标点放入到OPEN表中，起始点存放到CLOSED表，障碍物中相邻的边存入到可视图中。OPEN表用来存储待扩展的节点，CLOSED表用来提高搜索效率和存储路径。
(2)取出评价最小节点扩展。评价函数定义如下：
f(x)＝gn+hn+vc×MAX
其中gn表示起始节点到当前节点的实际路径长度，hn是启发函数，是当前节点到目标节点最短路径的估计。MAX表示从起始点到目标点可能的最大距离，vc表示该节点的访问次数，可以确保不会重复访问相同的节点。
(3)根据节点状态查找路径，分为2个分支执行。
一是探索上一节点到当前节点的路径，构造上一节点到当前节点的连线。如果该线已存在于可视图中或与障碍物无冲突，更新节点状态，如果该连线线不在可视图中，将其作为可视边添加到可视图中。如果连线与障碍物有冲突，取出被穿越障碍物两个方向上距离连线最远的顶点添加到OPEN表中。
二是探索当前节点到目标点的路径。构造当前节点到目标点的连线。如果该线已在可视图中或者与障碍物没有冲突，表示已查找到路径，输出路径，算法结束。如果该线与障碍物有冲突，将该节点添加到CLOSED表中，添加被穿越障碍物的顶点到OPEN表中。
步骤3f：该步骤由CombineTask任务实现，通过kafka接口将最优路径封装为消息返回给移动机器人。
步骤4：路径控制，
移动机器人获取返回的消息，将规划路径生成控制序列，使得移动机器人朝目标点移动。
图7显示在实验室的走廊，机器人从图中的起点到达目的点。移动机器人通过一种基于分布式并行计算的路径规划算法实现实时的路径规划避开障碍物安全到达目的点。图8显示了在智能手机上的显示界面，包含了激光雷达扫描的线段信息，操作者的生理信息等。
本发明将一种改进的路径规划算法和云计算技术相结合，基于云计算技术，使用分布式并行处理使得本发明具有高效、精度高、易于软硬件实现的优点，可以实现移动机器人实时路径规划。
以上显示和描述了本发明的基本原理、主要特征和本发明的优点。本行业的技术人员应该了解，本发明不受上述实施例的限制，上述实施例和说明书中 描述的只是说明本发明的原理，在不脱离本发明精神和范围的前提下，本发明还会有各种变化和改进，本发明要求保护范围由所附的权利要求书、说明书及其等效物界定。

Claims (8)

  1. 一种基于分布式并行计算的路径规划方法，其特征在于包括如下步骤：
    步骤1：数据采集，
    移动机器人通过数据采集模块定时从激光雷达传感器读取数据，然后进行格式化处理，数据存储到本地数据库；
    步骤2：消息传输，
    移动机器人将传感器数据封装为消息，调用远程接口将消息提交到云端；
    步骤3：路径规划，
    分布式并行计算路径规划方法，从消息队列中取出消息，将激光雷达数据分组，每个分组并行生成环境地图，合并环境地图，通过起始点连线构建搜索区间，通过改进A *算法并行构造可视图和查找路径，合并路径后封装为消息返回给移动机器人；
    步骤4：路径控制，
    移动机器人获取返回的消息，根据规划路径生成控制序列，使得移动机器人朝目标点移动。
  2. 如权利要求1所述的基于分布式并行计算的路径规划方法，其特征在于：上述步骤3包括如下步骤：
    步骤3a：从消息队列中取出消息，将雷达数据进行分组，分发给多个计算节点；
    步骤3b：并行环境地图生成，在每个计算节点将分组数据拟合为线段组，根据移动机器人防碰撞半径膨胀为多边形；
    步骤3c：接收上一计算节点传递的多边形信息，构造起始点连线取出两个顶点，构建区间路径搜索任务；
    步骤3d：在每个计算节点，基于多边形使用改进A *算法并行构造可视图和查找区间最优路径；
    步骤3e：合并路径，根据评价函数取出全局最优路径；
    步骤3f：将全局最优路径封装为消息返回给移动机器人。
  3. 如权利要求1所述的基于分布式并行计算的路径规划方法，其特征在于：上述步骤1包括如下步骤：
    步骤1a：移动机器人启动后，加载激光雷达驱动；
    步骤1b：根据采样周期调用对应驱动方法采集数据；
    步骤1c：格式化存储到本地数据库中。
  4. 如权利要求3所述的基于分布式并行计算的路径规划方法，其特征在于：上述步骤1包括如下步骤：
    步骤1a：当移动机器人启动，加载支持的传感器和通信协议注册到虚拟表中，表中包括激光雷达类型、数据格式、通信方法和参数；
    步骤1b：当移动机器人连接到雷达后，根据连接信息识别出通信协议和激光雷达类型调用虚拟表注册的方法获取激光雷达数据；
    步骤1c：将激光雷达数据格式化，存储到移动机器人的本地数据库中。
  5. 如权利要求1所述的基于分布式并行计算的路径规划方法，其特征在于：上述步骤2包括如下步骤：
    步骤2a：移动机器人将激光雷达数据和目标点封装为消息，通过IOT网关上传到云端；
    步骤2b：云端将接收到的消息存储到消息队列中。
  6. 一种基于分布式并行计算的路径规划系统，其特征在于：所述的移动机器人自主导航中的路径规划系统包括移动机器人端和云端；
    所述移动机器人系统运行在嵌入式操作系统上，移动机器人端包括四个部分：消息采集模块、路径管理模块、IoT网关和消息客户端；
    消息采集模块采集的信息来源于传感器，包括激光雷达和惯性传感器；
    云端包括四个部分：消息服务集群模块、路径规划模块、分布式存储系统和应用协调服务模块。
  7. 根据权利要求6所述的基于分布式并行计算的路径规划系统，其特征在于：所述的移动机器人端包括
    用于采集激光雷达数据，将其格式化为统一的形式的激光雷达数据采集模块；
    将收到的路径转换为控制序列，控制移动机器人朝目标移动的路径管理模块；
    用于消息的封装、解封以及转发的IoT网关；
    用于将消息推送到云端消息队列中以及从消息队列中获取客户端消息的消息客户端。
  8. 根据权利要求6所述的基于分布式并行计算的路径规划系统，其特征在于：所述的云端包括
    通过集群方式实现高可用和负载均衡的消息服务的消息服务集群模块；
    通过分布式方式实现移动机器人消息数据和地图数据的存储的分布式存储系统；
    通过分布式并行计算集群实现分布式并行路径规划的路径规划模块；
    实现计算集群中主控节点、N个计算节点之间的协调、监控和管理的应用协调服务模块。
PCT/CN2019/102439 2019-03-08 2019-08-26 一种基于分布式并行计算的路径规划方法及其系统 WO2020181729A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
ZA2021/07292A ZA202107292B (en) 2019-03-08 2021-09-28 Path planning method and system based on distributed parallel computing

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910174544.2 2019-03-08
CN201910174544.2A CN110057360A (zh) 2019-03-08 2019-03-08 一种基于分布式并行计算的路径规划方法及其系统

Publications (1)

Publication Number Publication Date
WO2020181729A1 true WO2020181729A1 (zh) 2020-09-17

Family

ID=67316764

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/102439 WO2020181729A1 (zh) 2019-03-08 2019-08-26 一种基于分布式并行计算的路径规划方法及其系统

Country Status (3)

Country Link
CN (1) CN110057360A (zh)
WO (1) WO2020181729A1 (zh)
ZA (1) ZA202107292B (zh)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112833905A (zh) * 2021-01-08 2021-05-25 北京大学 基于改进a*算法的分布式多agv无碰撞路径规划方法
CN112925307A (zh) * 2021-01-20 2021-06-08 中国科学院重庆绿色智能技术研究院 一种用于智能仓储机器人系统的分布式多机器人路径规划方法
CN113218383A (zh) * 2021-05-10 2021-08-06 山东新一代信息产业技术研究院有限公司 一种云端机器人地图构建方法及系统
CN113255870A (zh) * 2021-04-19 2021-08-13 北京市遥感信息研究所 区域重访路径优化方法、装置、设备及介质
CN113503884A (zh) * 2021-04-13 2021-10-15 上海擎朗智能科技有限公司 路径规划方法、装置、设备和存储介质

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110057360A (zh) * 2019-03-08 2019-07-26 江苏海事职业技术学院 一种基于分布式并行计算的路径规划方法及其系统
CN110763245A (zh) * 2019-10-25 2020-02-07 江苏海事职业技术学院 一种基于流式计算的地图创建方法及其系统
CN110879594B (zh) * 2019-11-25 2020-11-17 广西科技师范学院 一种基于大数据的机器人路径规划的数据管理系统
CN111885550B (zh) * 2020-06-06 2023-09-19 浙江中力机械股份有限公司 一种分布式自主移动机器人调度系统
CN112666942A (zh) * 2020-12-15 2021-04-16 美智纵横科技有限责任公司 自移动机器人及其路径规划方法、装置、设备和存储介质
CN112698944A (zh) * 2020-12-29 2021-04-23 乐陵欧曼电子科技有限公司 基于人脑模拟的分布式云计算系统及方法
CN114029944B (zh) * 2021-10-08 2024-07-19 智动时代(北京)科技有限公司 一种三维空间粒晶向位定位方法

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107515002A (zh) * 2016-06-17 2017-12-26 趣之科技（深圳）有限公司 一种基于LiDAR和云计算实现机器人实时室内地图构建和定位导航的系统方法和装置
CN108981695A (zh) * 2018-07-26 2018-12-11 浙江大学 一种基于ros的四轮车导航系统
CN110057360A (zh) * 2019-03-08 2019-07-26 江苏海事职业技术学院 一种基于分布式并行计算的路径规划方法及其系统

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107515002A (zh) * 2016-06-17 2017-12-26 趣之科技（深圳）有限公司 一种基于LiDAR和云计算实现机器人实时室内地图构建和定位导航的系统方法和装置
CN108981695A (zh) * 2018-07-26 2018-12-11 浙江大学 一种基于ros的四轮车导航系统
CN110057360A (zh) * 2019-03-08 2019-07-26 江苏海事职业技术学院 一种基于分布式并行计算的路径规划方法及其系统

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
CAO, QIWU: "Research on Path Planning and Real-time Positioning of Agricultural Robot Based on Cloud Computing", JOURNAL OF AGRICULTURAL MECHANIZATION RESEARCH, no. 2, 28 February 2019 (2019-02-28), pages 211 - 213, ISSN: 1003-188X, DOI: 20191115114124 *
LV, TAIZHI: "Non-official translation: Research on Mobile Robot Path Planning and Map Creation", ELECTRONIC TECHNOLOGY & INFORMATION SCIENCE, CHINA DOCTORAL DISSERTATIONS FULL-TEXT DATABASE, no. 7, 15 July 2018 (2018-07-15), ISSN: 1674-022X, DOI: 27 *
TRAN, N. ET AL.: "Global Path Planning for Autonomous Robots using Modified Visibility-graph", ICCAIS 2013, 28 November 2013 (2013-11-28), pages 317 - 321, XP032559408, DOI: 20191115114750 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112833905A (zh) * 2021-01-08 2021-05-25 北京大学 基于改进a*算法的分布式多agv无碰撞路径规划方法
CN112925307A (zh) * 2021-01-20 2021-06-08 中国科学院重庆绿色智能技术研究院 一种用于智能仓储机器人系统的分布式多机器人路径规划方法
CN112925307B (zh) * 2021-01-20 2023-03-24 中国科学院重庆绿色智能技术研究院 一种用于智能仓储机器人系统的分布式多机器人路径规划方法
CN113503884A (zh) * 2021-04-13 2021-10-15 上海擎朗智能科技有限公司 路径规划方法、装置、设备和存储介质
CN113503884B (zh) * 2021-04-13 2023-09-15 上海擎朗智能科技有限公司 路径规划方法、装置、设备和存储介质
CN113255870A (zh) * 2021-04-19 2021-08-13 北京市遥感信息研究所 区域重访路径优化方法、装置、设备及介质
CN113255870B (zh) * 2021-04-19 2023-08-04 北京市遥感信息研究所 区域重访路径优化方法、装置、设备及介质
CN113218383A (zh) * 2021-05-10 2021-08-06 山东新一代信息产业技术研究院有限公司 一种云端机器人地图构建方法及系统

Also Published As

Publication number Publication date
CN110057360A (zh) 2019-07-26
ZA202107292B (en) 2022-02-23

Similar Documents

Publication Publication Date Title
WO2020181729A1 (zh) 一种基于分布式并行计算的路径规划方法及其系统
Chen et al. A study of robotic cooperation in cloud robotics: Architecture and challenges
Ahmed et al. Cyber Physical System: Architecture, applications and research challenges
CN109479065B (zh) 边缘服务器及其操作方法
Cui et al. Offloading autonomous driving services via edge computing
CN111273892A (zh) 一种基于云端技术和边缘计算实现智能机器人的方法
WO2021197353A1 (zh) 数据分流方法、装置、设备及介质
CN105897509B (zh) 一种基于云计算的机器人软件运行及仿真系统
CN104049575A (zh) 在过程控制系统中收集并且向大数据机器递送数据
CN110658759A (zh) 一种可动态部署ai模型的工业智能控制器
KR102438214B1 (ko) Ict 서비스 제공 방법 및 시스템
CN113701742B (zh) 一种基于云端与边端融合计算的移动机器人slam方法
US11658937B2 (en) Method and apparatus of control for robotic swarms
CN114047760B (zh) 路径规划方法、装置、电子设备及自动驾驶车辆
El Shenawy et al. HDec-POSMDPs MRS exploration and fire searching based on IoT cloud robotics
Sossalla et al. DynNetSLAM: Dynamic visual SLAM network offloading
Martins et al. MRGS: A multi-robot SLAM framework for ROS with efficient information sharing
Badawy et al. New approach to enhancing the performance of cloud-based vision system of mobile robots
CN117406725A (zh) 路径规划方法、装置、设备及存储介质
CN116626700A (zh) 一种机器人定位方法、装置、电子设备和存储介质
Rao et al. Fog computing environment in flying ad-hoc networks: concept, framework, challenges, and applications
CN115841197A (zh) 一种路径规划方法、装置、设备及存储介质
Cao et al. Scaling up edge-assisted real-time collaborative visual slam applications
CN112688984B (zh) 一种向网络节点下发及执行指令的方法、设备及介质
Berndt et al. Unified Multi-Modal Data Aggregation for Complementary Sensor Networks Applied for Localization

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19919480

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19919480

Country of ref document: EP

Kind code of ref document: A1