An environment-driven task allocation method for heterogeneous spatial crowdsourcing

Abstract

The present invention relates to an environment-driven task allocation method for heterogeneous space crowdsourcing. Taking into account the spatiotemporal constraints of participants and tasks, different perceptual constraints, and the dynamics of heterogeneous space crowdsourcing environments, the large-scale global task allocation problem is transformed into a task allocation problem based on multi-region optimization. The specific process is as follows: construct available task allocation regions according to the spatiotemporal characteristics and matching degree of tasks and participants, and set agents for each region; secondly, train regional agents based on the discretized soft actor-critic algorithm and the centralized training decentralized execution framework, so that regional agents can select appropriate participants, and continuously learn and optimize task allocation results to adapt to the dynamics of heterogeneous space crowdsourcing environments.

Description

Technical Field

[0001] The present invention relates to the fields of spatial crowdsourcing and multi-agent deep reinforcement learning, and in particular to a task allocation method in heterogeneous spatial crowdsourcing that is affected by dynamic environmental changes. Background Art

[0002] In recent years, the rapid development of mobile networks and the popularity of mobile devices have promoted a new perception paradigm, namely spatial crowdsourcing (SC). Different from traditional crowdsourcing work, spatial crowdsourcing tasks are related to locations, and participants must reach the location specified by the task to complete the task. Spatial crowdsourcing (SC) has applications in many aspects, such as environmental monitoring, social networks, and academic research. Generally, spatial crowdsourcing has three components: 1) tasks; 2) workers; 3) platform (Spatial Crowdsourcing Server, SC-Server). The spatial crowdsourcing platform assigns tasks to nearby workers, that is, task allocation. Existing research on spatial crowdsourcing task allocation involves many aspects, such as: from single task allocation to multi-task allocation, from offline task allocation to online task allocation, from opportunistic perception to participatory perception, etc.

[0003] With the development of the Internet of Things and edge computing technologies, some environmental sensors with sensing capabilities have become popular. These sensors can collect various types of data (such as photos and videos) near their locations. Heterogeneous crowd sensing integrates multiple sensing resources in the environment. It has the characteristics of heterogeneous participant perception and collaborative perception. Inspired by heterogeneous crowd sensing, the present invention proposes an environment-driven task allocation method for heterogeneous space crowdsourcing that takes into account both human participants and environmental sensors. Summary of the invention

[0004] Technical issues to be solved

[0005] In order to avoid the shortcomings of the prior art, the present invention proposes an environment-driven task allocation method for heterogeneous space crowdsourcing.

[0006] Technical Solution

[0007] An environment-driven task allocation method for heterogeneous space crowdsourcing, characterized by the following steps:

[0008] Step 1: Obtain the basic information of each participant registered on the crowdsourcing platform: the participant's current location, perception radius, expected execution time, movement speed, perception data type, and perception quality; as well as the information of each task to be assigned: the location of the task, the start time of the task, the deadline of the task, the perception radius of the task, the execution time required for the task, and the perception data type required for the task; the task set is T, the participant P consists of workers W and environmental sensors S, and the perception data type set is

[0009] Step 2: According to the spatiotemporal characteristics of tasks and participants, the maximum available participant set MP for each task is divided into {mp1,…,mp n The maximum available set of participants should satisfy the following two conditions:

[0010]

[0011]

[0012] in, is the task to be assigned t i The start time of It is the task i location, It is the task i The deadline, It is the task i The perception radius, is a participant k location, is a participant k The moving speed, is a participant k The expected execution time, Sensors j The perception radius, Sensors j location;

[0013] The selection of the maximum available participant set MP is based on a common purpose: selecting participants who are closest to the task is more likely to complete the task before the deadline; because workers who are farther away will spend more time moving, and sensors that are farther away cannot cover the scope of the current task; selecting the maximum available participant set based on the spatiotemporal characteristics of participants and tasks can avoid the problem of uneven distribution of tasks and participants in equal grid perception areas; in addition, in the selection of the maximum available participant set, it is also necessary to minimize the similarity of the perception data contributed by the participants, while maximizing the matching degree of the perception data types between the task and the participants; the perception similarity between participants is calculated by introducing the Jaccard parameter; the perception data type of the participant is expressed as Perceptual similarity is defined as:

[0014]

[0015] Step 3: For task t i Constructs its available allocation area, which is determined by its sensing range and the region consisting of the largest set of participants Composition, that is, RT=(rt1,rt2,……,rt n ),in

[0016] Step 4: Construct the perceived quality of different participants. In order to maximize the execution of the task, the temporal and spatial characteristics and perceived similarity of the participants are considered in the construction of the region. However, the size of the perceived data provided by different participants is different. The perceived quality is set by the amount of data provided by the participants. In order to uniformly measure the perceived data gain contributed by different participants, the data amount is set to The perceptual data gain is defined as

[0017] Step 5: Assign tasks to the region. In step 4, the amount of data contributed by the participant is used as the perceived quality of the participant. By taking the overall perceived quality of the participant and the matching degree between the participant and the task as the optimization goal, the task is assigned to the region rt. i To this end, the discretized soft actor-critic algorithm SAC and the centralized training decentralized execution framework CTDE are used. The algorithm selects the appropriate participants according to the region rt generated in step 3. i Set up regional agents and assign tasks within the region; the actor network makes decisions based on the heterogeneous crowdsourcing environment in which it is located, uses the centralized training decentralized execution framework CTDE to expand multi-regional agents, and uses the information of other regional agents to train the critic network to ensure that the heterogeneous space crowdsourcing environment of each regional agent is stable.

[0018] A computer system, characterized in that it comprises: one or more processors, and a computer-readable storage medium for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned method.

[0019] A computer-readable storage medium, characterized in that it stores computer-executable instructions, which are used to implement the above method when executed.

[0020] A computer program, characterized by comprising computer executable instructions, which are used to implement the above method when executed.

[0021] Beneficial Effects

[0022] This paper proposes an environment-driven task allocation method for heterogeneous space crowdsourcing, which takes into account the spatiotemporal constraints of participants and tasks, different perceptual constraints, and the dynamics of heterogeneous space crowdsourcing environments, and transforms the large-scale global task allocation problem into a task allocation problem based on multi-region optimization. Available task allocation regions are constructed according to the spatiotemporal characteristics and matching degree of tasks and participants, and agents are set for each region; secondly, regional agents are trained based on the discretized soft behavior criticism algorithm and the decentralized execution centralized training framework, so that regional agents can select appropriate participants, and continuously learn and optimize task allocation results to adapt to the dynamics of heterogeneous space crowdsourcing environments.

[0023] 1. Construct the maximum available participant set based on the spatiotemporal constraints of tasks and participants and the perceived similarity between participants, and construct the available allocation area for tasks based on the maximum available participant set;

[0024] 2. A task allocation algorithm based on the discrete Soft Actor-Critic (SAC) algorithm and the Centralized Training with Decentralized Execution (CTDE) framework is designed for multi-region optimization. Taking the total perceived quality of participants and the matching degree between participants and tasks as dual optimization goals, the regional agent can adapt to the dynamic changes of the heterogeneous spatial crowdsourcing environment, continuously interact with the heterogeneous spatial crowdsourcing environment, and select appropriate participants to complete the task. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The drawings are only for the purpose of illustrating particular embodiments and are not to be considered limiting of the present invention. Like reference symbols denote like components throughout the drawings.

[0026] Figure 1 A flowchart of an environment-driven task allocation method for heterogeneous space crowdsourcing in an example of the present invention;

[0027] Figure 2 The process of constructing the maximum available set of participants is illustrated;

[0028] Figure 3 This is a schematic diagram of how the perception data gain changes with the amount of perception data contributed by participants. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0030] There are 3 drawings in this invention, please refer to Figure 1 As shown, the specific steps of the present invention are as follows:

[0031] Step 1: Obtain the basic information of each participant registered on the crowdsourcing platform: the participant’s current location, perception radius, expected execution time, movement speed, perception data type, and perception quality; as well as the information of each task to be assigned: the location of the task, the start time of the task, the deadline of the task, the perception radius of the task, the execution time required for the task, and the perception data type required for the task. The task set is T, the participant P consists of workers W and environmental sensors S, and the perception data type set is

[0032] Step 2: Divide the maximum available participant set MP for each task according to the spatiotemporal characteristics of tasks and participants. The maximum available participant set should meet the following two conditions:

[0033]

[0034]

[0035] in, is the task to be assigned t i The start time of It is the task i location, It is the task i The deadline, It is the task i The perception radius, is a participant k location, is a participant k The moving speed, is a participant k The expected execution time, Sensors j The perception radius, Sensors j location.

[0036] The selection of the maximum available participant set MP is based on a common purpose: selecting participants who are closest to the task is more likely to complete the task before the deadline. Because workers who are farther away will spend more time moving, and sensors that are farther away cannot cover the scope of the current task. Selecting the maximum available participant set based on the spatiotemporal characteristics of participants and tasks can avoid the problem of uneven distribution of tasks and participants in equal grid perception areas. In addition, in the selection of the maximum available participant set, it is also necessary to minimize the similarity of the perception data contributed by the participants, while maximizing the matching degree of the perception data types between the task and the participants. The perceptual similarity between participants is calculated by introducing the Jaccard parameter. The perception data type of the participant is expressed as Perceptual similarity can be defined as:

[0037]

[0038] Step 3: For task t i Constructs its available allocation area, which is determined by its sensing range ( It is the task i The perception radius) and the area consisting of the largest set of participants Composition, that is, RT=(rt1,rt2,……,rt n ),in

[0039] Step 4: Construct the perceived quality of different participants. In order to maximize the execution of the task, the spatiotemporal characteristics and perceived similarity of the participants are considered in the construction of the region. However, the size of the perceived data provided by different participants is different. For example, the amount of data available to workers is limited by the storage space of the mobile devices they carry. Different amounts of data will also produce different data gains. Therefore, the perceived quality of the participants is set by the amount of data provided by them. In order to uniformly measure the perceived data gain contributed by different participants, the amount of data is set to The perceptual data gain is defined as In order to verify the correctness of the model, the error between professional equipment and mobile phone measurement noise was calculated by using 130 noise data. The experiment shows that the log(·) function can better reflect the change of the perceived data gain contributed by the participants with the amount of perceived data provided by them, and also reflects that the data gain contributed by the participants has a certain data marginal benefit.

[0040] Step 5: Assign tasks to regions. In step 4, the amount of data contributed by the participant is used as the perceived quality of the participant. By taking the overall perceived quality of the participant and the matching degree between the participant and the task as the optimization target, the task is assigned to the region rt. iTo this end, the discretized soft actor-critic algorithm (SoftActor-Critic, SAC) and the centralized training decentralized execution framework (Centralized Training with DecentralizedExecution, CTDE) are used. The algorithm is based on the region rt generated in step 3. i Set up regional agents and assign tasks within the region. The actor network makes decisions based on the heterogeneous crowdsourcing environment in which it is located, uses the Centralized Training with Decentralized Execution (CTDE) framework to expand multi-regional agents, and uses the information of other regional agents to train the critic network to ensure that the heterogeneous space crowdsourcing environment of each regional agent is stable.

[0041] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should be included in the protection scope of the present invention.

Claims

1. An environment-driven task allocation method for heterogeneous space crowdsourcing, characterized in that Here are the steps: Step 1: Obtain the basic information of each participant registered on the crowdsourcing platform: the participant's current location, perception radius, expected execution time, movement speed, perception data type, and perception quality; as well as the information of each task to be assigned: the location of the task, the start time of the task, the deadline of the task, the perception radius of the task, the execution time required for the task, and the perception data type required for the task; the task set is T, the participant P consists of workers W and environmental sensors S, and the perception data type set is Step 2: According to the spatiotemporal characteristics of tasks and participants, the maximum available participant set MP = {mp1, ..., mp n }; The maximum available participant set should satisfy the following two conditions: (1) (2) in, is the task to be assigned t i The start time of It is the task i location, It is the task i The deadline, It is the task i The perception radius, is a participant k location, is a participant k The moving speed, is a participant k The expected execution time, Sensors j The perception radius, Sensors j location; The selection of the maximum available participant set MP is based on a common purpose: selecting participants who are closest to the task is more likely to complete the task before the deadline; because workers who are farther away will spend more time moving, and sensors that are farther away cannot cover the scope of the current task; selecting the maximum available participant set based on the spatiotemporal characteristics of participants and tasks can avoid the problem of uneven distribution of tasks and participants in equal grid perception areas; in addition, in the selection of the maximum available participant set, it is also necessary to minimize the similarity of the perception data contributed by the participants, while maximizing the matching degree of the perception data types between the task and the participants; the perception similarity between participants is calculated by introducing the Jaccard parameter; the perception data type of the participant is expressed as Perceptual similarity is defined as: Step 3: For task t i Constructs its available allocation area, which is determined by its sensing range and the region consisting of the largest set of participants Composition, that is, RT=(rt1,rt2,……,rt n ),in Step 4: Construct the perceived quality of different participants. In order to maximize the execution of the task, the temporal and spatial characteristics and perceived similarity of the participants are considered in the construction of the region. However, the size of the perceived data provided by different participants is different. The perceived quality is set by the amount of data provided by the participants. In order to uniformly measure the perceived data gain contributed by different participants, the data amount is set to The perceptual data gain is defined as Step 5: Assign tasks to the region. In step 4, the amount of data contributed by the participant is used as the perceived quality of the participant. By taking the overall perceived quality of the participant and the matching degree between the participant and the task as the optimization goal, the task is assigned to the region rt. i To this end, the discretized soft actor-critic algorithm SAC and the centralized training decentralized execution framework CTDE are used. The algorithm selects the appropriate participants according to the region rt generated in step 3. i Set up regional agents and assign tasks within the region; the actor network makes decisions based on the heterogeneous crowdsourcing environment in which it is located, uses the centralized training decentralized execution framework CTDE to expand multi-regional agents, and uses the information of other regional agents to train the critic network to ensure that the heterogeneous space crowdsourcing environment of each regional agent is stable.

2. A computer system, characterized in that include: One or more processors, and a computer-readable storage medium for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors are enabled to implement the method of claim 1.

3. A computer-readable storage medium, characterized in that Computer executable instructions are stored, and when the instructions are executed, they are used to implement the method of claim 1.

4. A computer program product, characterized in that The invention comprises computer executable instructions, which are used to implement the method of claim 1 when being executed.

Images