A robot-based task distribution method and system

Abstract

The present invention provides a robot-based task distribution method and system, the method comprising: receiving a distribution task request, the distribution task request including a distribution destination; when the robot arrives at the level of the distribution destination, detecting the pedestrian posture of each pedestrian in the scene; locking the pedestrian whose pedestrian posture is detected to be a waving posture as a target customer; according to the position of the target customer, controlling the robot to move to the target customer position to distribute the task. In the present invention, when the robot arrives at the destination level, it can identify the posture of the pedestrian in the scene, and when the pedestrian posture is identified as a waving posture, determine the target customer, perceive the position of the target customer in advance, and re-plan the route according to the position of the target customer, thereby increasing the flexibility of the robot's distribution tasks and increasing the interaction between the robot and the customer.

Description

Technical Field

[0001] The present invention relates to the field of robotics technology, and more specifically, to a robot-based task distribution method and system. Background Art

[0002] Intelligent delivery service robots have gradually entered people's daily lives. In many smart buildings, hotels and other places, robots have begun to play the role of delivery attendants. The building manager or food delivery platform initiates a robot request and informs the robot of the delivery destination and the contact information of the delivery customer. When the robot arrives or is about to arrive at the delivery destination, the robot initiates a phone notification request to the cloud to inform the user to pick up the item.

[0003] In current delivery scenarios, when a robot is assigned a task, it follows a fixed route, reaches a pre-set point, waits for the user to pick up the food / items, and then returns to the original standby point. Compared with the interaction between service personnel and customers during the delivery process, this type of delivery service robot that operates along a fixed route is very rigid and wastes transportation capacity. Summary of the invention

[0004] The present invention aims at the technical problems existing in the prior art and provides a robot-based task distribution method and system.

[0005] According to a first aspect of the present invention, a robot-based task delivery method is provided, comprising: receiving a delivery task request, the delivery task request including a delivery destination; when the robot arrives at the level of the delivery destination, detecting the pedestrian posture of each pedestrian in the scene; locking a pedestrian whose pedestrian posture is detected to be a waving posture as a target customer; and according to the target customer position, controlling the robot to move to the target customer position to deliver the task.

[0006] Based on the above technical solution, the present invention can also make the following improvements.

[0007] Optionally, when the robot arrives at the flat floor of the delivery destination, a human posture detection model is turned on to detect the posture of each pedestrian in the scene; the human posture detection model includes a human posture detection sub-model and a human motion recognition sub-model, and the human posture detection sub-model is an anchor-free single-stage detection network Centernet; the picture is input into the human posture detection sub-model to obtain the position of the pedestrian in the picture and multiple human posture key points of the pedestrian; the human motion recognition sub-model recognizes the pedestrian's posture according to the positional relationship between the multiple human posture key points of the pedestrian.

[0008] Optionally, the anchor-free single-stage detection network Centernet includes a backbone network backbone and a detection head head, and the backbone network backbone adopts a lightweight network architecture; the picture is input into the backbone network backbone to obtain high-dimensional feature information in the picture; the high-dimensional feature information is input into the detection head head to obtain the position of pedestrians in the picture and the position of key points of pedestrians.

[0009] Optionally, the key point positions of the pedestrian include left and right shoulder key point positions, left and right elbow key point positions, and left and right hand key point positions.

[0010] Optionally, the human motion recognition sub-model identifies the pedestrian's posture according to the positional relationship between multiple human posture key points of the pedestrian, including: obtaining the key point positions of the pedestrian's left shoulder, left elbow and left hand based on the movement symmetry of the left and right arms; or obtaining the key point positions of the pedestrian's right shoulder, right elbow and right hand; the human motion recognition model determines whether the key point positions of the pedestrian's left shoulder, left elbow and left hand meet both the first action condition and the second action condition; or determines whether the key point positions of the pedestrian's right shoulder, right elbow and right hand meet both the first action condition and the second action condition; if satisfied, the pedestrian's posture is a waving posture.

[0011] Optionally, the first action condition is that the hand key point position is above the elbow key point position, and the elbow key point position is above the shoulder key point position, which is expressed by the following formula: ey -P sy >gap and P wy -P ey >gap, the pedestrian’s posture is determined to be waving; where P sy is the position coordinate of the shoulder key point, P ey is the position coordinate of the elbow key point, P wy is the position coordinate of the key points of the hand, and gap is the preset threshold.

[0012] Optionally, whether the key point positions of the pedestrian's left shoulder, left elbow and left hand meet the second action condition is determined in the following manner: using the robot's front camera to take a sequence of pictures of the pedestrian at fixed intervals, the picture sequence including multiple pictures; using a human posture detection sub-model to extract the key point positions of the pedestrian in each picture, and obtain human key point sequence data; based on the human key point positions extracted from two adjacent frames of pictures, calculate the movement change of the human key points in each two adjacent frames of pictures, and obtain movement change sequence data; if the movement change sequence data meets the preset conditions, it is determined that the key point positions of the pedestrian's left shoulder, left elbow and left hand meet the second action condition.

[0013] Optionally, the method calculates the movement change of the human body key points in each of the two adjacent frames of pictures according to the positions of the human body key points extracted from the two adjacent frames of pictures to obtain the movement change sequence data, including: obtaining the human body key point positions (Psx1, Pex1, Pwx1) of the previous frame time series t1 and the human body key point positions (Psx0, Pex0, Pwx0) of the current frame time series t0, Psx1, Pex1, Pwx1 are the human body shoulder key point positions, elbow key point positions and hand key point positions of the previous frame time series t1, and Psx0, Pex0, Pwx0 are the human body shoulder key point positions, elbow key point positions and hand key point positions of the current frame time series t0; calculate the movement change of the elbow and hand of the current frame time series t0 (v e0 ,vω0):

[0014] V e =(P ex0 -P sx0 )-(P ex1 -P sx1 );

[0015] V ω =(P ωx0 -P sx0 )-(P ωx1 -P sx1 );

[0016] Calculate the movement change of the key points of the human body in each two adjacent frames and obtain the movement change sequence data (v e ,v ω )={(v e0 ,v ω0 ),(v e1 ,v ω1 ),(v e2 ,v ω2 ),...,(v eN ,v ωN )}.

[0017] Optionally, the preset conditions are as follows:

[0018] mean(Topk(V e ))>MinE;

[0019] And mean(Topk(-V e ))>MinE;

[0020] And mean(Topk(V ω ))>MinE;

[0021] And mean(Topk(-V ω ))>MinE;

[0022] Among them, Topk represents the largest K movement changes in the movement change sequence data, K<N, mean represents the average value, MinE and MinW represent the minimum boundary conditions of the movement change.

[0023] According to a second aspect of the present invention, there is provided a robot-based task delivery system, comprising: a receiving module for receiving a delivery task request, wherein the delivery task request includes a delivery destination; a detection module for detecting the pedestrian posture of each pedestrian in the scene when the robot arrives at the level of the delivery destination; a control module for locking a pedestrian whose pedestrian posture is detected to be a waving posture as a target customer, and controlling the robot to move to the target customer position to deliver the task according to the target customer position.

[0024] According to a third aspect of the present invention, there is provided an electronic device comprising a memory and a processor, wherein the processor is configured to implement the steps of a robot-based task distribution method when executing a computer management program stored in the memory.

[0025] According to a fourth aspect of the present invention, there is provided a computer-readable storage medium having a computer management program stored thereon, wherein the computer management program, when executed by a processor, implements the steps of a robot-based task distribution method.

[0026] The present invention provides a robot-based task distribution method and system. When the robot arrives at the destination level, it can recognize the posture of pedestrians in the scene. When the pedestrian's posture is recognized as a waving posture, the target customer is determined, the position of the target customer is sensed in advance, and the route is replanned according to the position of the target customer, thereby increasing the flexibility of the robot's delivery tasks and increasing the interaction between the robot and the customer. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 A flow chart of a robot-based task distribution method provided by the present invention;

[0028] Figure 2 An overall flow chart of a robot task distribution method provided by the present invention;

[0029] Figure 3 A schematic diagram of the structure of a robot-based task distribution system provided by the present invention;

[0030] Figure 4 A schematic diagram of the hardware structure of a possible electronic device provided by the present invention;

[0031] Figure 5 A schematic diagram of the hardware structure of a possible computer-readable storage medium provided by the present invention. DETAILED DESCRIPTION

[0032] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0033] Figure 1 A flow chart of a robot-based task distribution method provided by the present invention, such as Figure 1 As shown, the method includes: 101. receiving a delivery task request, wherein the delivery task request includes a delivery destination; 102. when the robot arrives at the level of the delivery destination, detecting the pedestrian posture of each pedestrian in the scene; 103. locking a pedestrian whose pedestrian posture is detected to be a waving posture as a target customer; 104. according to the target customer position, controlling the robot to move to the target customer position to perform task delivery.

[0034] It can be understood that, in contrast to the traditional robot task distribution method, the present invention provides an interactive method of customer gesture recognition. The customer waves to the robot, and the robot detects the gestures of pedestrians in the scene, confirms the target customer, and senses the target customer's location in advance, thereby improving service quality and saving service resources.

[0035] Specifically, when the platform sends a task delivery command to the robot, the robot receives the delivery task request, which includes the delivery destination and the customer's contact information. When the robot receives the task delivery command, it starts delivering. When the robot leaves the elevator and reaches the floor, it turns on the human posture detection model. When it detects a pedestrian waving in the environment, it locks the pedestrian as the target customer and obtains the target customer's spatial position. After arriving near the target customer, it waits for the customer to pick up the meal. After picking up the meal, it returns to the original standby point.

[0036] In the present invention, when the robot arrives at the destination level, it can recognize the posture of pedestrians in the scene. When the posture of the pedestrian is recognized as a waving posture, the target customer is determined, the position of the target customer is sensed in advance, and the route is replanned according to the position of the target customer, which increases the flexibility of the robot's delivery task and also increases the interaction between the robot and the customer.

[0037] In a possible implementation manner, when the robot arrives at the flat floor of the delivery destination, a human posture detection model is turned on to detect the posture of each pedestrian in the scene; the human posture detection model includes a human posture detection sub-model and a human motion recognition sub-model, and the human posture detection sub-model is an anchor-free single-stage detection network Centernet; an image is input into the human posture detection sub-model to obtain the position of the pedestrian in the image and multiple human posture key points of the pedestrian; the human motion recognition sub-model recognizes the pedestrian's posture according to the positional relationship between the multiple human posture key points of the pedestrian.

[0038] It is understandable that when the robot leaves the elevator and reaches a floor, it turns on the front camera, takes pictures of pedestrians in the scene on the floor, obtains pictures including pedestrians, and detects the positions and postures of the pedestrians in the pictures.

[0039] In the embodiment of the present invention, a human posture detection model is used to detect the position and posture of pedestrians in a picture. The human posture detection model mainly includes a human posture detection sub-model and a human action recognition sub-model.

[0040] In a possible implementation manner, the anchor-free single-stage detection network Centernet includes a backbone network backbone and a detection head head, and the backbone network backbone adopts a lightweight network architecture; the image is input into the backbone network backbone to obtain high-dimensional feature information in the image; the high-dimensional feature information is input into the detection head head to obtain the position of pedestrians in the image and the key point positions of pedestrians.

[0041] It is understandable that the human posture detection sub-model uses the anchor-free single-stage detection network Centernet to obtain the position of pedestrians in the picture and the key points of human posture from the input picture end to end. The anchor-free single-stage detection network Centernet consists of the backbone network backbone and the detection head head. The backbone network backbone uses a lightweight network architecture Shufflenet, which is easy to deploy on edge devices such as robots to extract high-dimensional feature information in the picture. The network inputs the high-dimensional features extracted by the backbone network backbone into the detection head head to obtain the position of pedestrians in the picture and the key point positions of pedestrians KeyPoints.

[0042] In a possible implementation manner, the key point positions of the pedestrian include left and right shoulder key point positions, left and right elbow key point positions, and left and right hand key point positions.

[0043] It can be understood that, in the embodiment of the present invention, the interactive action between the robot and the customer is defined as an action similar to waving, and the key points of the human body related to the action are six key points: the left and right shoulders, elbows, and wrists. The picture captured by the robot camera is input into the human posture detection sub-model, so that the position and posture of the pedestrian in the picture can be detected.

[0044] In one possible implementation manner, the human motion recognition sub-model identifies the pedestrian's posture according to the positional relationship between multiple human posture key points of the pedestrian, including: based on the movement symmetry of the left and right arms, obtaining the key point positions of the pedestrian's left shoulder, left elbow and left hand; or, obtaining the key point positions of the pedestrian's right shoulder, right elbow and right hand; the human motion recognition model determines whether the key point positions of the pedestrian's left shoulder, left elbow and left hand meet the first action condition and the second action condition at the same time; or determines whether the key point positions of the pedestrian's right shoulder, right elbow and right hand meet the first action condition and the second action condition at the same time; if satisfied, the pedestrian's posture is a waving posture.

[0045] It can be understood that the above embodiment detects the key points of the human body in the image through the human posture detection sub-model, and uses (Psx, Psy) (pointshoulder), (Pex, Pey) (pointelbow), and (Pwx, Pwy) (pointwrist) to represent the coordinates of the key points of the left and right shoulders, the key points of the left and right elbows, and the key points of the left and right hands, respectively. Taking into account the symmetry of the movements of the left and right arms, a waving movement of any unilateral arm indicates that the customer has made an interactive action. The embodiment of the present invention obtains the key point position of the right arm.

[0046] When judging whether the pedestrian's posture is a waving posture, two conditions are mainly set, namely waving action condition one and waving action condition two (also referred to as the first action condition and the second action condition in the embodiment of the present invention). When the positional relationship between the shoulder key point position, the elbow key point position and the hand key point position of the pedestrian's right arm simultaneously satisfies the waving action condition one and the waving action condition two, the pedestrian's posture is judged to be a waving posture.

[0047] In a possible implementation manner, the first action condition is that the hand key point position is above the elbow key point position, and the elbow key point position is above the shoulder key point position, which is expressed by the following formula:

[0048] When P ey -P sy >gap and P wy -P ey >gap, the pedestrian’s posture is determined to be waving;

[0049] Among them, P sy is the position coordinate of the shoulder key point, P ey is the position coordinate of the elbow key point, P wy is the position coordinate of the key points of the hand, gap is a preset threshold, and in the embodiment of the present invention, gap=10, which is related to the size of the picture.

[0050] In a possible implementation manner, whether the key point positions of the pedestrian's left shoulder, left elbow and left hand meet the second action condition is determined in the following manner: a sequence of pictures of the pedestrian is taken at fixed intervals using the robot's front camera, wherein the picture sequence includes multiple pictures; the key point positions of the pedestrian in each picture are extracted using a human posture detection submodel to obtain human key point sequence data; based on the human key point positions extracted from two adjacent frames of pictures, the movement change of the human key points in each of two adjacent frames of pictures is calculated to obtain movement change sequence data; if the movement change sequence data meets the preset conditions, it is determined that the key point positions of the pedestrian's left shoulder, left elbow and left hand meet the second action condition.

[0051] It can be understood that when the robot arrives at the delivery destination, it turns on the front camera, samples the environmental image data once every 50ms, and takes multiple pictures. For each picture, the human body key points are extracted through the human posture detection sub-model to obtain the human body key point sequence data.

[0052] Get the human body key point positions (Psx1, Pex1, Pwx1) of the previous frame time series t1 and the human body key point positions (Psx0, Pex0, Pwx0) of the current frame time series t0, where Psx1, Pex1, and Pwx1 are the human body shoulder key point positions, elbow key point positions, and hand key point positions of the previous frame time series t1, respectively; Psx0, Pex0, and Pwx0 are the human body shoulder key point positions, elbow key point positions, and hand key point positions of the current frame time series t0.

[0053] Calculate the movement change of the elbow and hand in the current frame time series t0 (v e0 ,vω0):

[0054] V e =(P ex0 -P sx0 )-(P ex1 -Psx1 );

[0055] V ω =(P ωx0 -P sx0 )-(P ωx1 -P sx1 ).

[0056] Based on the above formula, the key points of each part in each adjacent frame of the picture, such as the movement change of the elbow key point and the hand key point, are calculated. Taking into account the time information, the time series data (v e ,v ω )={(v e0 ,v ω0 ),(v e1 ,v ω1 ),(v e2 ,v ω2 ),...,(v eN ,v ωN For example, in the embodiment of the present invention, since the sampling interval is 50ms, the most recent 20 frames of data (v e ,v ω )={(v e0 ,v ω0 ),(v e1 ,v ω1 ),(v e2 ,v ω2 ),...,(v e20 ,v ω20 )}, judge the characteristics of 20 groups of data, and judge that the pedestrian is waving if the preset conditions are met.

[0057] In a possible implementation manner, the preset conditions are as follows:

[0058] mean(Topk(V e ))>MinE;

[0059] And mean(Topk(-V e ))>MinE;

[0060] And mean(Topk(V ω ))>MinE;

[0061] And mean(Topk(-V ω ))>MinE;

[0062] Among them, Topk represents the largest K movement changes in the movement change sequence data, K<N, mean represents the average value, MinE and MinW represent the minimum boundary conditions of the movement change. In the embodiment of the present invention, K is 5, MinE and MinW are 3 and 6 respectively.

[0063] When the positional relationship between the key points of various parts of the detected pedestrian satisfies the first action condition and the second action condition at the same time, the pedestrian's posture is a waving posture; when one of the action conditions is not met, the pedestrian's posture is not a waving posture, and the key point position data of the pedestrian is retained in the historical cache data pool.

[0064] See also Figure 2 , is an overall flow chart of a robot-based task distribution method. When a distribution task is required, the distribution platform issues a distribution task to the robot. When the robot receives the task distribution command, it performs the task distribution. When the robot reaches the target level, it turns on the front camera and takes pictures of pedestrians in the target level scene to obtain pictures including pedestrians. The taken pictures are input into the human posture detection sub-model, and the positions of the key points of the human body are input. According to the positions of the key points of the human body, the human action recognition sub-model recognizes the human posture. When the pedestrian's posture is identified as a waving posture, the pedestrian is the target customer. The position of the pedestrian is obtained, and the robot is controlled to change the original route and move to the position of the target customer.

[0065] Figure 3 A structure diagram of a robot-based task distribution system provided by an embodiment of the present invention, such as Figure 3 As shown, a robot-based task distribution system includes a receiving module 301, a detection module 302 and a control module 303, wherein:

[0066] The receiving module 301 is used to receive a delivery task request, wherein the delivery task request includes a delivery destination; the detection module 302 is used to detect the pedestrian posture of each pedestrian in the scene when the robot arrives at the level of the delivery destination; the control module 303 is used to lock a pedestrian whose pedestrian posture is detected to be a waving posture as a target customer, and control the robot to move to the target customer position to perform task delivery according to the target customer position.

[0067] It can be understood that the robot-based task distribution system provided by the present invention corresponds to the robot-based task distribution method provided by the aforementioned embodiments. The relevant technical features of the robot-based task distribution system can refer to the relevant technical features of the robot-based task distribution method, which will not be repeated here.

[0068] See also Figure 4 , Figure 4 Schematic diagram of an electronic device provided by an embodiment of the present invention. Figure 4As shown, an embodiment of the present invention provides an electronic device, including a memory 410, a processor 420, and a computer program 411 stored in the memory 410 and executable on the processor 420. When the processor 420 executes the computer program 411, the following steps are implemented: receiving a delivery task request, wherein the delivery task request includes a delivery destination; when the robot arrives at the level of the delivery destination, detecting the posture of each pedestrian in the scene; locking a pedestrian whose posture is detected to be a waving posture as a target customer; and according to the target customer's position, controlling the robot to move to the target customer's position to deliver the task.

[0069] See also Figure 5 , Figure 5 Schematic diagram of an embodiment of a computer-readable storage medium provided by the present invention. Figure 5 As shown, this embodiment provides a computer-readable storage medium 500 on which a computer program 511 is stored. When the computer program 511 is executed by a processor, the following steps are implemented: receiving a delivery task request, wherein the delivery task request includes a delivery destination; when the robot arrives at the level of the delivery destination, detecting the pedestrian posture of each pedestrian in the scene; locking a pedestrian whose pedestrian posture is detected to be a waving posture as a target customer; and according to the target customer's location, controlling the robot to move to the target customer's location to deliver the task.

[0070] An embodiment of the present invention provides a robot-based task delivery method and system. When the robot arrives at the destination level, it can recognize the posture of pedestrians in the scene. When the pedestrian's posture is recognized as a waving posture, the target customer is determined, the location of the target customer is sensed in advance, and the route is replanned according to the location of the target customer, thereby increasing the flexibility of the robot's delivery tasks and increasing the interaction between the robot and the customer.

[0071] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0072] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0073] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0074] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0075] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0076] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0077] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A robot-based task distribution method, characterized in that: include: receiving a delivery task request, wherein the delivery task request includes a delivery destination; When the robot reaches the flat floor of the delivery destination, it detects the pedestrian posture of each pedestrian in the scene; A pedestrian whose gesture is detected as waving is locked as a target customer; According to the target customer's location, control the robot to move to the target customer's location to deliver the task; When the robot reaches the flat floor of the delivery destination, the human posture detection model is turned on to detect the posture of each pedestrian in the scene; The human posture detection model includes a human posture detection sub-model and a human action recognition sub-model, and the human posture detection sub-model is an anchor-free single-stage detection network Centernet; Input the image into the human posture detection sub-model to obtain the position of the pedestrian in the image and multiple human posture key points of the pedestrian; The human action recognition sub-model recognizes pedestrian postures based on the positional relationship between multiple human posture key points of pedestrians; The anchor-free single-stage detection network Centernet includes a backbone network backbone and a detection head head, and the backbone network backbone adopts a lightweight network architecture; Input the image into the backbone network to obtain high-dimensional feature information in the image; Input the high-dimensional feature information into the detection head to obtain the position of the pedestrian in the image and the key point position of the pedestrian; The key point positions of the pedestrian include left and right shoulder key point positions, left and right elbow key point positions, and left and right hand key point positions; The human motion recognition sub-model recognizes the pedestrian's posture according to the positional relationship between multiple human posture key points of the pedestrian, including: based on the movement symmetry of the left and right arms, obtaining the key point positions of the pedestrian's left shoulder, left elbow and left hand; or obtaining the key point positions of the pedestrian's right shoulder, right elbow and right hand; the human motion recognition model determines whether the key point positions of the pedestrian's left shoulder, left elbow and left hand meet the first action condition and the second action condition at the same time; or determines whether the key point positions of the pedestrian's right shoulder, right elbow and right hand meet the first action condition and the second action condition at the same time; if they meet, the pedestrian's posture is a waving posture; the first action condition is that the hand key point position is above the elbow key point position, and the elbow key point position is above the shoulder key point position, which is expressed by the following formula: when P ey -P sy >gap and P wy -P ey >gap, the pedestrian’s posture is determined to be waving; where P sy is the position coordinate of the shoulder key point, P ey is the position coordinate of the elbow key point, P wy is the position coordinate of the key points of the hand, and gap is the preset threshold.

2. The task distribution method according to claim 1, characterized in that: Determine whether the key point positions of the left shoulder, the left elbow, and the left hand of the pedestrian meet the second action condition by: using a front camera of the robot to take a sequence of pictures of the pedestrian at fixed intervals, the sequence of pictures including a plurality of pictures; The human posture detection sub-model is used to extract the key point positions of pedestrians in each picture to obtain the human key point sequence data; according to the human key point positions extracted from two adjacent frames of pictures, the movement change of the human key points in each of the two adjacent frames of pictures is calculated to obtain the movement change sequence data; If the movement variation sequence data meets the preset conditions, it is determined that the key point positions of the pedestrian's left shoulder, left elbow and left hand meet the second action condition.

3. The task distribution method according to claim 2, characterized in that: The method calculates the movement variation of the human key points in each of the two adjacent frames of pictures according to the positions of the human key points extracted from the two adjacent frames of pictures, and obtains the movement variation sequence data, including: obtaining the human key point positions (Psx1, Pex1, Pwx1) of the previous frame time sequence t1 and the human key point positions (Psx0, Pex0, Pwx0) of the current frame time sequence t0, where Psx1, Pex1, and Pwx1 are the human shoulder key point positions, elbow key point positions, and hand key point positions of the previous frame time sequence t1, respectively, and Psx0, Pex0, and Pwx0 are the human shoulder key point positions, elbow key point positions, and hand key point positions of the current frame time sequence t0; and calculating the movement variation (v e0 ,v ω0 ): IN e =(P ex0 -Ps x0 )-(P ex1 -P sx1 ); V ω =(P ωx0 -P sx0 )-(P ωx1 -P sx1 ); Calculate the movement change of the key points of the human body in each two adjacent frames and obtain the movement change sequence data (v e ,v ω )={(v e0 ,v ω0 ),(v e1 ,v ω1 ),(v e2 ,v ω2 ),...,(v eN ,v ωN )}.

4. The task distribution method according to claim 3, characterized in that: The pre-conditions are as follows: mean(Topk(V e ))>MinE; And mean(Topk(-V e ))>MinE; And mean(Topk(V ω ))>MinE; And mean(Topk(-V ω ))>MinE; Among them, Topk represents the largest K movement changes in the movement change sequence data, K<N, mean represents the average value, MinE and MinW represent the minimum boundary conditions of the movement change.

5. A robot-based task distribution system, characterized in that: include: A receiving module, used for receiving a delivery task request, wherein the delivery task request includes a delivery destination; The detection module is used to detect the pedestrian posture of each pedestrian in the scene when the robot arrives at the flat floor of the delivery destination; A control module is used to lock a pedestrian whose gesture is a waving gesture as a target customer, and control the robot to move to the target customer's location to perform task delivery according to the target customer's location; The detection module is specifically used to start the human posture detection model to detect the posture of each pedestrian in the scene when the robot arrives at the flat floor of the delivery destination; The human posture detection model includes a human posture detection sub-model and a human action recognition sub-model, and the human posture detection sub-model is an anchor-free single-stage detection network Centernet; Input the image into the human posture detection sub-model to obtain the position of the pedestrian in the image and multiple human posture key points of the pedestrian; The human action recognition sub-model is used to recognize pedestrian postures based on the positional relationship between multiple human posture key points of pedestrians; The anchor-free single-stage detection network Centernet includes a backbone network backbone and a detection head head, and the backbone network backbone adopts a lightweight network architecture; Input the image into the backbone network to obtain high-dimensional feature information in the image; Input the high-dimensional feature information into the detection head to obtain the position of the pedestrian in the image and the key point position of the pedestrian; The key point positions of the pedestrian include left and right shoulder key point positions, left and right elbow key point positions, and left and right hand key point positions; The human action recognition sub-model is used to recognize the pedestrian's posture according to the positional relationship between the pedestrian's multiple human posture key points, specifically: based on the movement symmetry of the left and right arms, obtain the pedestrian's left shoulder key point position, left elbow key point position and left hand key point position; or obtain the pedestrian's right shoulder key point position, right elbow key point position and right hand key point position; The human action recognition model is used to determine whether the key point positions of the pedestrian's left shoulder, left elbow and left hand meet the first action condition and the second action condition at the same time; or determining whether the key point position of the right shoulder, the key point position of the right elbow and the key point position of the right hand of the pedestrian simultaneously satisfies the first action condition and the second action condition; If it is satisfied, the pedestrian posture is a waving posture; the first action condition is that the position of the hand key point is above the position of the elbow key point, and the position of the elbow key point is above the position of the shoulder key point, which is expressed by the following formula: When P ey -P sy >gap and P wy -P ey >gap, the pedestrian’s posture is determined to be waving; Among them, P sy is the position coordinate of the shoulder key point, P ey is the position coordinate of the elbow key point, P wy is the position coordinate of the key points of the hand, and gap is the preset threshold.

Images