CN114385362A - Multi-transfer robot scheduling method based on cloud-edge computing - Google Patents
Multi-transfer robot scheduling method based on cloud-edge computing Download PDFInfo
- Publication number
- CN114385362A CN114385362A CN202210036118.4A CN202210036118A CN114385362A CN 114385362 A CN114385362 A CN 114385362A CN 202210036118 A CN202210036118 A CN 202210036118A CN 114385362 A CN114385362 A CN 114385362A
- Authority
- CN
- China
- Prior art keywords
- robot
- path
- computing server
- algorithm
- edge computing
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 69
- 238000012546 transfer Methods 0.000 title claims abstract description 39
- 238000004040 coloring Methods 0.000 claims abstract description 23
- 238000004891 communication Methods 0.000 claims abstract description 6
- 238000005516 engineering process Methods 0.000 claims abstract description 5
- 230000008569 process Effects 0.000 claims description 30
- 238000010845 search algorithm Methods 0.000 claims description 24
- 238000004364 calculation method Methods 0.000 claims description 15
- 239000003086 colorant Substances 0.000 claims description 12
- 238000005457 optimization Methods 0.000 claims description 7
- 238000001514 detection method Methods 0.000 claims description 6
- 238000003780 insertion Methods 0.000 claims description 3
- 230000037431 insertion Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000004806 packaging method and process Methods 0.000 claims description 2
- 230000008901 benefit Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5011—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals
- G06F9/5016—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals the resource being the memory
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/54—Interprogram communication
- G06F9/544—Buffers; Shared memory; Pipes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
- G06Q10/047—Optimisation of routes or paths, e.g. travelling salesman problem
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
- G06Q10/083—Shipping
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Software Systems (AREA)
- Entrepreneurship & Innovation (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- General Business, Economics & Management (AREA)
- Operations Research (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- General Engineering & Computer Science (AREA)
- Educational Administration (AREA)
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
Abstract
The invention provides a cloud-edge computing-based multi-transfer robot scheduling method, which mainly comprises the following steps: each robot is connected to an edge computing server through a wireless communication technology, and the edge computing server requests a cloud computing server; the method comprises the steps that an edge computing server instantiates a multi-carrier robot task scheduling problem into a coloring traveler problem model, and requests coloring traveler problem solving service of a cloud computing server by using HTTP service, wherein the service is packaged in the cloud computing server by adopting a container; the cloud computing server executes a parallel coloring traveler problem solving algorithm after receiving the scheduling request, and sends a multi-robot task scheduling result to the edge computing server through an HTTP (hyper text transport protocol); and after receiving the task scheduling result of the multiple robots, the edge computing server executes a path planning algorithm of the multiple robots and sends the obtained shortest path to each robot. The invention solves the problems that the dispatching of the multi-carrying robot is long in time consumption and the cluster scale is limited.
Description
Technical Field
The invention relates to the field of multi-transfer robot scheduling, in particular to a multi-transfer robot scheduling method based on cloud-edge computing.
Background
The rapid development of the express delivery industry puts higher requirements on the automation degree and the working efficiency of the warehousing operation. At present, domestic power distribution and logistics companies gradually adopt a multipoint carrying robot integrating taking, placing, storing and transporting functions to replace a lifting type goods shelf to carry out overall carrying for automatic storage transformation. The novel multi-point transfer robot is matched with a good scheduling method, and the warehousing automation degree and the operation efficiency are expected to be further improved.
The problem of scheduling the multi-transfer robots includes two parts, namely job scheduling which determines which robot the job tasks are executed in which order, and path planning which plans a collision-free path for each transfer robot. For multi-robot job scheduling problems, the mainstream methods include market-based methods and optimization-based methods. The method based on the market adopts an auction principle, namely, the robot autonomously carries out bid auction on tasks according to the state of the robot and completes task scheduling in a mode of 'high-priced people'. This approach is not suitable for large-scale robot clusters, and the task scheduling results are not optimized enough. Generally speaking, an optimization-based method can perform more optimal and larger-scale task scheduling, and generally, a specific problem is mathematically modeled and then solved by an intelligent optimization algorithm. The multi-robot path planning method generally adopts an A-star algorithm to plan the path of the multiple robots, and then carries out the solution of path conflict among the multiple robots based on rules. With the development of cloud computing technology, people begin to migrate the computing of the robot body to be located in a cloud computing center for computing so as to reduce the computing load of the robot body. However, cloud computing is not sufficient to support connection of more and more internet of things devices, so edge computing is proposed as a supplement to cloud computing, and application services of part of network center nodes are migrated to the network edge nodes, so that load of a cloud computing center is reduced in a data processing mode on a data production side. Cloud computing and edge computing can support scheduling of larger-scale transfer robot clusters.
At present, the domestic multi-transfer robot scheduling technology mainly adopts a serial algorithm, uses a single industrial control computer and completes task scheduling and path planning of the multi-transfer robot by integrated software. The computing architecture fusing cloud computing and edge computing can support more flexible and more optimized multi-robot scheduling, wherein the quality and scale of task scheduling solution can be further improved by the powerful computing capability of the cloud computing in cooperation with a parallel solving algorithm. The robot is connected to the edge to form global state information, and more optimized multi-robot path planning is completed based on the global state information.
Disclosure of Invention
The purpose of the invention is as follows: the invention aims to provide a multi-transfer robot scheduling method based on cloud-edge computing, aiming at the problems that the scheduling of a multi-transfer robot is long in time consumption and the cluster scale of the multi-transfer robot is limited.
The invention is realized by the following technical scheme:
a multi-transfer robot scheduling method based on cloud-edge computing realizes task scheduling and path planning of at least two mobile transfer robots by using a cloud computing server and an edge computing server respectively, and comprises the following specific steps:
step S1: each carrying robot is connected to an edge computing server through a wireless communication technology, and the edge computing server requests a cloud computing server;
step S2: the method comprises the steps that an edge computing server instantiates a multi-carrier robot task scheduling problem into a coloring traveler problem model, an HTTP service is used for requesting coloring traveler problem solving services of a cloud computing server, and the coloring traveler problem solving services are packaged in the cloud computing server by adopting containers;
step S3: the cloud computing server executes a parallel coloring traveler problem solving algorithm after receiving the scheduling request, and sends a multi-robot task scheduling result to the edge computing server through an HTTP (hyper text transport protocol);
step S4: and after receiving the task scheduling result of the multiple robots, the edge computing server executes a path planning algorithm of the multiple robots and sends the obtained shortest path to each transfer robot.
Further, in step S1, the edge computing server and the transfer robot use an HTTP protocol to ensure low-latency communication capability, and complete multi-robot path planning using the computing capability of the edge computing server, and the cloud computing server provides a scheduling solution service and calls through an API;
the cloud computing server adopts OpenFaaS to provide cloud service, the OpenFaaS provides watchdog application, the application can package a general binary program into an OpenFaaS function, a watchdog application process is used as an initial process of the OpenFaaS function and can simultaneously receive a plurality of service calling requests, the OpenFaaS is provided with a lightweight HTTP server, each time the watchdog application receives one HTTP request, a new solving service process is created, the HTTP request is analyzed, and the weight of the request is oriented to be standard input of the solving process; and when the solving process is finished, the watchdog application encapsulates the standard output of the process into an HTTP (hyper text transport protocol) return body and sends the HTTP return body to the service requester.
Further, the method for instantiating the multi-carrier robot task scheduling problem into the coloring traveler problem model and the coloring traveler problem solving service for calling the cloud computing server described in step S2 specifically include the following steps:
step 2-1: inputting a batch of carrying tasks into a cloud computing server, extracting coordinates of each task by the cloud computing server to generate a task position set, and modeling according to a coloring traveler problem theory, wherein each carrying robot is represented as a merchant, the task position corresponds to a city, each merchant has different colors, each city has one to a plurality of colors, the colors represent the matching relationship between the robot and the tasks, and an objective function is that the total travel of all merchants completing the visit of all cities is shortest;
step 2-2: establishing a candidate city set without intersection for each businessman, firstly, placing a single-color city with the same color as each businessman into the candidate city set;
step 2-3: the parallel variable neighborhood search algorithm is adopted, and specifically comprises the following steps:
1) randomly distributing the multicolor cities to a certain trader with the same color according to colors, completing the division of an initial candidate city set, and randomly generating an initial solution of the city sequence of each trader by utilizing each candidate city set;
2) local search is carried out on the obtained city sequence by adopting a 2-opt algorithm, solution disturbance is carried out by adopting a random insertion strategy, a new city access sequence of each merchant is formed, and the process is realized in parallel by using a thread pool mode;
3) and (3) repeating the step (1) and the step (2) until the set iteration times are met, and outputting the optimal city access sequence of each merchant, namely the optimal task scheduling solution of each corresponding transfer robot.
Further, the thread pool manner described in step 2-3 refers to that m parallel variable neighborhood search algorithms are executed in a manner of each independent thread, and the specific thread allocation and execution manner is as follows:
the method comprises the steps that when a thread pool is initialized, s threads and a task queue are generated according to the capacity s of the thread pool, and the task queue stores computing tasks solved by m merchant city access sequences; when the task queue is not empty, an idle thread is awakened and used for taking out a calculation task in the task queue and calling a variable neighborhood searching algorithm to execute the calculation task; the thread pool capacity s is min (c, m), wherein c is the number of processor cores; when c is less than m, executing m calculation tasks in batches, and executing the calculation tasks of the city sequence optimization solution of no more than s traders each time until all m calculation tasks are executed; and when c > -m, executing the calculation task of the city sequence optimization solution of the m travelers.
Further, the method for packaging by using a container described in step S2 specifically includes:
1) the container uses Docker to pull OpenFaaS watch and Alpine Linux as basic mirror images;
2) installing a compiling tool g + + and a cmake in the container, copying a source code of the parallel variable neighborhood search algorithm to the container, and compiling;
3) setting the fprocess environment variable as a binary code path to be executed, executing the process pointed by the fprocess environment variable by the watchdog application when receiving a call request every time, and compiling a solving algorithm into a container mirror image by the Docker calling a compiling command according to a Dockerfile of solving service.
Further, the multi-robot path planning algorithm in step S4 adopts a parallelized collision-based search algorithm, which is divided into two layers, namely a top layer for performing path collision detection of the multi-carrier robot and a bottom layer for performing collision-free path planning of each carrier robot, wherein the top layer for performing path collision detection of the multi-carrier robot searches for collision-free paths by tree expansion using a constraint tree data structure, and each constraint tree node is a multi-way tree and has the following three attributes:
each constraint in the set belongs to one robot, the constraint set of a root node of the constraint tree is empty, and each child node inherits the constraint set of a parent node and adds a new constraint;
solving the node, namely solving a path obtained by planning the path at the bottom layer of the calling algorithm under the constraint of the current node constraint set by all the robots;
the total cost is the sum of the obtained path costs of all the robots;
specifically, a conflict-based search algorithm needs to detect whether paths of multiple robots conflict or not on each constraint tree node, when a conflict is detected, multiple constraint tree child nodes are generated according to the conflict, then path planning of each conflict robot on the bottom layer is called for multiple times, and execution is performed in a multi-thread mode;
the method specifically comprises the following steps:
step 4-1, inputting an electronic grid map of a given operation scene, mapping a task scheduling result into a sequence of an initial position and a target position of each transfer robot, and executing a plurality of space-time A-x algorithm flows in parallel in a thread pool mode to obtain the shortest path of each robot;
the space-time A algorithm is a path planning algorithm which runs on the weighted graph, and the weighted graph, the starting point and the target point are given, and the goal of the space-time A is to find a path with the minimum cost from the starting point to the target point;
the search process of the space-time A algorithm comprises the following steps: starting from the starting vertex, executing edge search to expand to other vertexes until a target vertex is found;
the space-time A algorithm adopts a heuristic rule to select the vertex for expansion, namely, the vertex v which is not visited and is the minimum is selected for expansion each time the vertex is expanded, and the calculation formula of f (v) is as follows:
f(v)＝g(v)+h(v)
where g (v) is the path cost from the departure point to v, given in the data set, and h (v) is a heuristic function for estimating the cost of the vertex v to the destination point; if the value of h (v) is not greater than the actual cost from the vertex v to the target point, h (v) is called as an allowable heuristic function, and the space-time A-x algorithm using the allowable heuristic function can ensure that the path cost from the starting point to the target point obtained in the solving process is minimum; in the two-dimensional grid map, the space-time A-star algorithm uses different allowable heuristic functions, including a heuristic function for estimating the cost of Manhattan distance and diagonal distance; let the coordinate of the current node v be expressed as (x)v,yv) The target node is d, and its coordinates are expressed as (x)d,yd) Then, the heuristic function of the manhattan distance is adopted as follows:
|xd-xv|+|yd-yv|
the heuristic function using diagonal distance is:
step 4-2, detecting conflicts among paths according to the obtained shortest paths of different robots, namely, overlapping parts of the two paths in the same time period;
and 4-3, when the path conflict of the robots is detected, converting the path conflict between the robots into the path constraint of the conflicting robots, namely converting the spatial conflict into the waiting constraint on time, then executing multi-robot path re-planning under multi-constraint in parallel by adopting a thread pool mode and executing the step 4-2 again until no path conflict exists, and outputting the optimal path without conflict of each robot at the moment.
Has the advantages that: the invention can execute task scheduling and path planning under the cloud side environment, and solves the problems that the scheduling of the multi-transfer robots is long in time consumption and the cluster scale of the multi-transfer robots is limited.
Drawings
FIG. 1 is a software flow diagram of a method in accordance with the present invention;
FIG. 2 is a flowchart of a task scheduling method for a multi-carrier robot according to the present invention;
FIG. 3 is a flow chart of a multi-transfer robot path planning method of the present invention;
fig. 4 is a graph comparing cloud execution and local execution in the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
A multi-transfer robot scheduling method based on cloud-edge computing realizes task scheduling and path planning of at least two mobile transfer robots by using a cloud computing server and an edge computing server respectively, and comprises the following specific steps
Step 1: sending a scheduling task request to a cloud computing server;
step 2: the cloud computing server executes a parallel coloring traveler problem solving algorithm after receiving the scheduling request, and sends a multi-robot task scheduling result to the edge computing server through an HTTP (hyper text transport protocol);
and step 3: and the edge computing server executes a multi-robot path planning algorithm after receiving the multi-robot task scheduling result and sends the obtained shortest path to each transfer robot.
In step S1, the edge computing server and the transfer robot use an HTTP protocol to ensure low-latency communication capability, and complete multi-robot path planning using the computing capability of the edge computing server, and the cloud computing server provides a scheduling solution service and calls through an API;
the cloud computing server adopts OpenFaaS to provide cloud service, the OpenFaaS provides watchdog application, the application can package a general binary program into an OpenFaaS function, a watchdog application process is used as an initial process of the OpenFaaS function and can simultaneously receive a plurality of service calling requests, the OpenFaaS is provided with a lightweight HTTP server, each time the watchdog application receives one HTTP request, a new solving service process is created, the HTTP request is analyzed, and the weight of the request is oriented to be standard input of the solving process; and when the solving process is finished, the watchdog application encapsulates the standard output of the process into an HTTP (hyper text transport protocol) return body and sends the HTTP return body to the service requester.
Further, the method for instantiating the multi-carrier robot task scheduling problem into the coloring traveler problem model and the coloring traveler problem solving service for calling the cloud computing server described in step S2 specifically include the following steps:
further, the method for instantiating the multi-carrier robot task scheduling problem into the coloring traveler problem model and the coloring traveler problem solving service for calling the cloud computing server described in step S2 specifically include the following steps:
step 2-1: inputting a batch of carrying tasks into a cloud computing server, extracting coordinates of each task by the cloud computing server to generate a task position set, and modeling according to a coloring traveler problem theory, wherein each carrying robot is represented as a merchant, the task position corresponds to a city, each merchant has different colors, each city has one to a plurality of colors, the colors represent the matching relationship between the robot and the tasks, and an objective function is that the total travel of all merchants completing the visit of all cities is shortest;
step 2-2: establishing a candidate city set without intersection for each businessman, firstly, placing a single-color city with the same color as each businessman into the candidate city set;
step 2-3: the parallel variable neighborhood search algorithm is adopted, and specifically comprises the following steps: the flow chart of the algorithm is shown in fig. 2.
Step 1, randomly distributing multicolor cities to traders with the same color according to colors, completing the division of an initial candidate city set, and randomly generating an initial solution of a city sequence of each trader by using each candidate city set;
step 2, local search is carried out on the obtained city sequence by adopting a 2-opt algorithm, solution disturbance is carried out by adopting a random insertion strategy, a new city access sequence of each merchant is formed, and the process is realized in parallel by using a thread pool mode;
and 3, repeating the step 1 and the step 2 until the set iteration times are met, and outputting the optimal city access sequence of each merchant, namely the optimal task scheduling solution of each corresponding carrying robot.
A flow chart of a parallel collision-based search algorithm of the method to which the present invention relates is shown in fig. 3. The algorithm is divided into two layers, namely a top layer for executing path conflict detection of the multi-handling robot and a bottom layer for executing conflict-free path planning of each handling robot, wherein the top layer for executing the path conflict detection of the multi-handling robot searches conflict-free paths by using a constraint tree data structure to perform tree expansion, the constraint tree is a multi-branch tree, and each constraint tree node has the following three attributes:
each constraint in the set belongs to one robot, the constraint set of a root node of the constraint tree is empty, and each child node inherits the constraint set of a parent node and adds a new constraint;
solving the node, namely solving a path obtained by planning the path at the bottom layer of the calling algorithm under the constraint of the current node constraint set by all the robots;
the total cost is the sum of the obtained path costs of all the robots;
specifically, a conflict-based search algorithm needs to detect whether paths of multiple robots conflict or not on each constraint tree node, when a conflict is detected, multiple constraint tree child nodes are generated according to the conflict, then path planning of each conflict robot on the bottom layer is called for multiple times, and execution is performed in a multi-thread mode;
the method specifically comprises the following steps:
step 4-1, inputting an electronic grid map of a given operation scene, mapping a task scheduling result into a sequence of an initial position and a target position of each transfer robot, and executing a plurality of space-time A-x algorithm flows in parallel in a thread pool mode to obtain the shortest path of each robot;
the space-time A algorithm is a path planning algorithm which runs on the weighted graph, and the weighted graph, the starting point and the target point are given, and the goal of the space-time A is to find a path with the minimum cost from the starting point to the target point;
the search process of the space-time A algorithm comprises the following steps: starting from the starting vertex, executing edge search to expand to other vertexes until a target vertex is found;
the space-time A algorithm adopts a heuristic rule to select the vertex for expansion, namely, the vertex v which is not visited and is the minimum is selected for expansion each time the vertex is expanded, and the calculation formula of f (v) is as follows:
f(v)＝g(v)+h(v)
where g (v) is the path cost from the departure point to v, given in the data set, and h (v) is a heuristic function for estimating the cost of the vertex v to the destination point; if the value of h (v) is not greater than the actual cost from the vertex v to the target point, h (v) is called as an allowable heuristic function, and the space-time A-x algorithm using the allowable heuristic function can ensure that the path cost from the starting point to the target point obtained in the solving process is minimum; in a two-dimensional grid map, the spatio-temporal A-Algorithm uses different admissible heuristic functions, including heuristics for cost estimation of Manhattan distance and diagonal distanceA function; let the coordinate of the current node v be expressed as (x)v,yv) The target node is d, and its coordinates are expressed as (x)d,yd) Then, the heuristic function of the manhattan distance is adopted as follows:
|xd-xv|+|yd-yv|
the heuristic function using diagonal distance is:
step 4-2, detecting conflicts among paths according to the obtained shortest paths of different robots, namely, overlapping parts of the two paths in the same time period;
and 4-3, when the path conflict of the robots is detected, converting the path conflict between the robots into the path constraint of the conflicting robots, namely converting the spatial conflict into the waiting constraint on time, then executing multi-robot path re-planning under multi-constraint in parallel by adopting a thread pool mode and executing the step 4-2 again until no path conflict exists, and outputting the optimal path without conflict of each robot at the moment.
Table 1 shows the comparison of the results of the variable neighborhood search algorithm and the parallel variable neighborhood search algorithm running on a regular data set. As can be seen from the table, on the regular data set, the minimum cost value and the average value of the solution obtained by the parallel variable neighborhood searching algorithm executed in parallel are obviously superior to those of the variable neighborhood searching algorithm executed in series. This is because parallel computing enables more iterative solutions than serial computing in the same time. Therefore, the parallel variable neighborhood search algorithm can explore a larger search space, so that a better solution can be found.
TABLE 1 comparison of results of IVNS and PIVNS algorithms run on a regular data set
In order to verify the effectiveness of the parallel variable neighborhood searching algorithm, the solution effect of the parallel variable neighborhood searching algorithm and the variable neighborhood searching algorithm in the same time is compared. The algorithm runs independently 10 times on each data set, and the thread pool size takes the minimum of the number of travelers m and the number of processor cores. Since the core number of the device processor running the experiment is 40, the thread pool size takes min (m, 40).
The results of the series-parallel conflict-based search algorithm are shown in table 2.
TABLE 2 run time consuming comparison of serial and parallel CBS algorithms on different datasets
As can be seen from table 2, in the same problem, since the parallel conflict-based search algorithm uses multi-thread operation, the time consumption for operating the parallel conflict-based search algorithm is shorter than that of the conflict-based search algorithm, and as the number of robots in the problem increases, the advantage of the parallel conflict-based search algorithm is more obvious. When the number of robots is less than 30, the increase in the time consumption for solving is not significant. However, when the number of robots exceeds 30, the time required for solution increases significantly. This occurs because when the number of robots increases to a certain extent, the probability of collision of paths between the robots increases significantly, and thus more searches are required to find a solution. When the top-level node expansion is carried out, a plurality of child nodes are generated according to the path conflict by the parallel conflict-based search algorithm, and the process is executed in parallel. Thus, when the number of top level expansion nodes of the algorithm is the same, the parallel collision-based search algorithm runs less time-consuming than the collision-based search algorithm and has greater advantages as the number of top level expansion nodes increases.
Finally, it should be noted that: the above embodiments are only used for illustrating the technical solutions of the present application and not for limiting the scope of protection thereof, and although the present application is described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: numerous variations, modifications, and equivalents will occur to those skilled in the art upon reading the present application and are within the scope of the claims appended hereto.
Claims (6)
1. A multi-transfer robot scheduling method based on cloud-edge computing is characterized in that task scheduling and path planning of at least two mobile transfer robots are achieved by using one cloud computing server and one edge computing server respectively, and the method comprises the following specific steps:
step S1: each carrying robot is connected to an edge computing server through a wireless communication technology, and the edge computing server requests a cloud computing server;
step S2: the method comprises the steps that an edge computing server instantiates a multi-carrier robot task scheduling problem into a coloring traveler problem model, an HTTP service is used for requesting coloring traveler problem solving services of a cloud computing server, and the coloring traveler problem solving services are packaged in the cloud computing server by adopting containers;
step S3: the cloud computing server executes a parallel coloring traveler problem solving algorithm after receiving the scheduling request, and sends a multi-robot task scheduling result to the edge computing server through an HTTP (hyper text transport protocol);
step S4: and after receiving the task scheduling result of the multiple robots, the edge computing server executes a path planning algorithm of the multiple robots and sends the obtained shortest path to each transfer robot.
2. The cloud-edge computing based multi-transfer robot scheduling method of claim 1, wherein: in step S1, the edge computing server and the transfer robot use an HTTP protocol to ensure low-latency communication capability, and complete multi-robot path planning using the computing capability of the edge computing server, and the cloud computing server provides a scheduling solution service and calls through an API;
the cloud computing server adopts OpenFaaS to provide cloud service, the OpenFaaS provides watchdog application, the application can package a general binary program into an OpenFaaS function, a watchdog application process is used as an initial process of the OpenFaaS function and can simultaneously receive a plurality of service calling requests, the OpenFaaS is provided with a lightweight HTTP server, each time the watchdog application receives one HTTP request, a new solving service process is created, the HTTP request is analyzed, and the weight of the request is oriented to be standard input of the solving process; and when the solving process is finished, the watchdog application encapsulates the standard output of the process into an HTTP (hyper text transport protocol) return body and sends the HTTP return body to the service requester.
3. The cloud-edge computing based multi-transfer robot scheduling method of claim 1, wherein: the method for instantiating the multi-carrier robot task scheduling problem into the coloring traveler problem model and the coloring traveler problem solving service for calling the cloud computing server, which are described in the step S2, specifically include the following steps:
step 2-1: inputting a batch of carrying tasks into a cloud computing server, extracting coordinates of each task by the cloud computing server to generate a task position set, and modeling according to a coloring traveler problem theory, wherein each carrying robot is represented as a merchant, the task position corresponds to a city, each merchant has different colors, each city has one to a plurality of colors, the colors represent the matching relationship between the robot and the tasks, and an objective function is that the total travel of all merchants completing the visit of all cities is shortest;
step 2-2: establishing a candidate city set without intersection for each businessman, firstly, placing a single-color city with the same color as each businessman into the candidate city set;
step 2-3: the parallel variable neighborhood search algorithm is adopted, and specifically comprises the following steps:
1) randomly distributing the multicolor cities to a certain trader with the same color according to colors, completing the division of an initial candidate city set, and randomly generating an initial solution of the city sequence of each trader by utilizing each candidate city set;
2) local search is carried out on the obtained city sequence by adopting a 2-ept algorithm, solution disturbance is carried out by adopting a random insertion strategy, a new city access sequence of each merchant is formed, and the process is realized in parallel by using a thread pool mode;
3) and (3) repeating the step (1) and the step (2) until the set iteration times are met, and outputting the optimal city access sequence of each merchant, namely the optimal task scheduling solution of each corresponding transfer robot.
4. The cloud-edge computing based multi-transfer robot scheduling method of claim 3, wherein: the thread pool mode in step 2-3 refers to executing m parallel variable neighborhood search algorithms in a mode of each independent thread, and the specific thread allocation and execution mode is as follows:
the method comprises the steps that when a thread pool is initialized, s threads and a task queue are generated according to the capacity s of the thread pool, and the task queue stores computing tasks solved by m merchant city access sequences; when the task queue is not empty, an idle thread is awakened and used for taking out a calculation task in the task queue and calling a variable neighborhood searching algorithm to execute the calculation task; the thread pool capacity s is min (c, m), wherein c is the number of processor cores; when c is less than m, executing m calculation tasks in batches, and executing the calculation tasks of the city sequence optimization solution of no more than s traders each time until all m calculation tasks are executed; and when c is larger than m, executing the calculation task of the city sequence optimization solution of m travelers.
5. The cloud-edge computing based multi-transfer robot scheduling method of claim 1, wherein: the method for packaging a container in step S2 includes:
1) the container uses Docker to pull OpenFaaS watch and Alpine Linux as basic mirror images;
2) installing a compiling tool g + + and a cmake in the container, copying a source code of the parallel variable neighborhood search algorithm to the container, and compiling;
3) setting the fprocess environment variable as a binary code path to be executed, executing the process pointed by the fprocess environment variable by the watchdog application when receiving a call request every time, and compiling a solving algorithm into a container mirror image by the Docker calling a compiling command according to a Dockerfile of solving service.
6. The cloud-edge computing based multi-transfer robot scheduling method of claim 1, wherein: the multi-robot path planning algorithm described in step S4 adopts a parallelized conflict-based search algorithm, which is divided into two layers, namely a top layer for performing path conflict detection of the multi-carrier robot and a bottom layer for performing conflict-free path planning of each carrier robot, the top layer for performing path conflict detection of the multi-carrier robot searches for a conflict-free path by performing tree expansion using a constraint tree data structure, the constraint tree is a multi-way tree, and each constraint tree node has the following three attributes:
each constraint in the set belongs to one robot, the constraint set of a root node of the constraint tree is empty, and each child node inherits the constraint set of a parent node and adds a new constraint;
solving the node, namely solving a path obtained by planning the path at the bottom layer of the calling algorithm under the constraint of the current node constraint set by all the robots;
the total cost is the sum of the obtained path costs of all the robots;
specifically, a conflict-based search algorithm needs to detect whether paths of multiple robots conflict or not on each constraint tree node, when a conflict is detected, multiple constraint tree child nodes are generated according to the conflict, then path planning of each conflict robot on the bottom layer is called for multiple times, and execution is performed in a multi-thread mode;
the method specifically comprises the following steps:
step 4-1, inputting an electronic grid map of a given operation scene, mapping a task scheduling result into a sequence of an initial position and a target position of each transfer robot, and executing a plurality of space-time A-x algorithm flows in parallel in a thread pool mode to obtain the shortest path of each robot;
the space-time A algorithm is a path planning algorithm which runs on the weighted graph, and the weighted graph, the starting point and the target point are given, and the goal of the space-time A is to find a path with the minimum cost from the starting point to the target point;
the search process of the space-time A algorithm comprises the following steps: starting from the starting vertex, executing edge search to expand to other vertexes until a target vertex is found;
the space-time A algorithm adopts a heuristic rule to select the vertex for expansion, namely, the vertex v which is not visited and is the minimum is selected for expansion each time the vertex is expanded, and the calculation formula of f (v) is as follows:
f(v)＝g(v)+h(v)
where g (v) is the path cost from the departure point to v, given in the data set, and h (v) is a heuristic function for estimating the cost of the vertex v to the destination point; if the value of h (v) is not greater than the actual cost from the vertex v to the target point, h (v) is called as an allowable heuristic function, and the space-time A-x algorithm using the allowable heuristic function can ensure that the path cost from the starting point to the target point obtained in the solving process is minimum; in the two-dimensional grid map, the space-time A-star algorithm uses different allowable heuristic functions, including a heuristic function for estimating the cost of Manhattan distance and diagonal distance; let the coordinate of the current node v be expressed as (x)v，yv) The target node is d, and its coordinates are expressed as (x)d，yd) Then, the heuristic function of the manhattan distance is adopted as follows:
|xd-xv|+|yd-yv|
the heuristic function using diagonal distance is:
step 4-2, detecting conflicts among paths according to the obtained shortest paths of different robots, namely, overlapping parts of the two paths in the same time period;
and 4-3, when the path conflict of the robots is detected, converting the path conflict between the robots into the path constraint of the conflicting robots, namely converting the spatial conflict into the waiting constraint on time, then executing multi-robot path re-planning under multi-constraint in parallel by adopting a thread pool mode and executing the step 4-2 again until no path conflict exists, and outputting the optimal path without conflict of each robot at the moment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210036118.4A CN114385362A (en) | 2022-01-12 | 2022-01-12 | Multi-transfer robot scheduling method based on cloud-edge computing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210036118.4A CN114385362A (en) | 2022-01-12 | 2022-01-12 | Multi-transfer robot scheduling method based on cloud-edge computing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114385362A true CN114385362A (en) | 2022-04-22 |
Family
ID=81202755
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210036118.4A Pending CN114385362A (en) | 2022-01-12 | 2022-01-12 | Multi-transfer robot scheduling method based on cloud-edge computing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114385362A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116362652A (en) * | 2023-06-01 | 2023-06-30 | 上海仙工智能科技有限公司 | Transport allocation task scheduling method and system and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017173990A1 (en) * | 2016-04-07 | 2017-10-12 | 北京进化者机器人科技有限公司 | Method for planning shortest path in robot obstacle avoidance |
| CN108549285A (en) * | 2018-05-07 | 2018-09-18 | 安徽恒诺机电科技有限公司 | A kind of pipe gallery automatic tour inspection system |
| CN109819032A (en) * | 2019-01-24 | 2019-05-28 | 中山大学 | A kind of base station selected cloud robot task distribution method with computation migration of joint consideration |
| CN111522666A (en) * | 2020-04-27 | 2020-08-11 | 西安工业大学 | Cloud robot edge computing unloading model and unloading method thereof |
-
2022
- 2022-01-12 CN CN202210036118.4A patent/CN114385362A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017173990A1 (en) * | 2016-04-07 | 2017-10-12 | 北京进化者机器人科技有限公司 | Method for planning shortest path in robot obstacle avoidance |
| CN108549285A (en) * | 2018-05-07 | 2018-09-18 | 安徽恒诺机电科技有限公司 | A kind of pipe gallery automatic tour inspection system |
| CN109819032A (en) * | 2019-01-24 | 2019-05-28 | 中山大学 | A kind of base station selected cloud robot task distribution method with computation migration of joint consideration |
| CN111522666A (en) * | 2020-04-27 | 2020-08-11 | 西安工业大学 | Cloud robot edge computing unloading model and unloading method thereof |
Non-Patent Citations (2)
| Title |
|---|
| PENGHUI DING等: "A Parallel Algorithm for Solving General Colored Traveling Salesman Problems", 《2021 36TH YOUTH ACADEMIC ANNUAL CONFERENCE OF CHINESE ASSOCIATION OF AUTOMATION (YAC)》, 26 July 2021 (2021-07-26), pages 1 - 6 * |
| ZHICHENG LIN等: "Task Scheduling and Path Planning of Multiple AGVs via Cloud and Edge Computing", 《2021 IEEE INTERNATIONAL CONFERENCE ON NETWORKING, SENSING AND CONTROL (ICNSC) 》, 5 December 2021 (2021-12-05), pages 1 - 6, XP034083030, DOI: 10.1109/ICNSC52481.2021.9702191 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116362652A (en) * | 2023-06-01 | 2023-06-30 | 上海仙工智能科技有限公司 | Transport allocation task scheduling method and system and storage medium |
| CN116362652B (en) * | 2023-06-01 | 2023-10-31 | 上海仙工智能科技有限公司 | Transport allocation task scheduling method and system and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Fragapane et al. | Planning and control of autonomous mobile robots for intralogistics: Literature review and research agenda | |
| Bean | Genetic algorithms and random keys for sequencing and optimization | |
| CN115237135A (en) | Mobile robot path planning method and system based on conflict | |
| CN109067834A (en) | Discrete particle cluster dispatching algorithm based on oscillatory type inertia weight | |
| Xue et al. | A hybrid many-objective competitive swarm optimization algorithm for large-scale multirobot task allocation problem | |
| CN111798097B (en) | Autonomous mobile robot task allocation processing method based on market mechanism | |
| Jing et al. | Multi-agent reinforcement learning based on graph convolutional network for flexible job shop scheduling | |
| CN114385362A (en) | Multi-transfer robot scheduling method based on cloud-edge computing | |
| CN109726841B (en) | AGV path calculation method based on unmanned cabin and AGV driving path control method | |
| Rinciog et al. | Fabricatio-rl: a reinforcement learning simulation framework for production scheduling | |
| Shriyam et al. | Task assignment and scheduling for mobile robot teams | |
| US11346669B2 (en) | Systems and methods for utilizing modeling to automatically generate paths for indoor navigation | |
| Xia et al. | A multi-AGV optimal scheduling algorithm based on particle swarm optimization | |
| CN111667060B (en) | Deep learning algorithm compiling method and device and related products | |
| Wang et al. | Research on hybrid real-time picking routing optimization based on multiple picking stations | |
| Jarrah et al. | The optimisation of travelling salesman problem based on parallel ant colony algorithm | |
| Zhen et al. | How to deploy robotic mobile fulfillment systems | |
| Ru et al. | Research on a hybrid neural network task assignment algorithm for solving multi-constraint heterogeneous autonomous underwater robot swarms | |
| Teck et al. | An efficient multi-agent approach to order picking and robot scheduling in a robotic mobile fulfillment system | |
| JP7466147B2 (en) | Information processing system and information processing method | |
| Mei et al. | A Multi-robot Task Allocation and Path Planning Method for Warehouse System | |
| Zhao | Dynamic Path Planning Analysis of Warehouse Handling Robot | |
| Oliveira et al. | Efficient multi-constrained task allocation in smart warehouses | |
| Zhou et al. | An improved bi-objective salp swarm algorithm based on decomposition for green scheduling in flexible manufacturing cellular environments with multiple automated guided vehicles | |
| Zhao et al. | An efficient scheduling and navigation approach for warehouse multi-mobile robots |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |