A method and system for intelligent scheduling of a quadruped robot

Abstract

This application relates to the field of robot scheduling management, and particularly to an intelligent scheduling method and system for quadruped robots. The method includes the steps of: constructing and training a graph neural network; the nodes of the graph neural network represent the coordinates of loading and unloading points, the weight of goods, and the initial power of the robot; the edges represent the routes between the nodes, the input is the adjacency matrix and the node feature matrix; the output is the predicted power consumption value and the predicted completion time value of the task; obtaining the task route, selecting the robots at the nodes directly associated with the nodes passed by the task route and the robots at the nodes passed by the task route as target robots; calculating the out-value of the target robots; obtaining the priority coefficients of each task, using the priority coefficients and the out-value as the edge weights for the matching of tasks and robots, and taking the task allocation plan with the largest accumulated sum of edge weights as the matching plan to perform global matching on multiple robots and multiple tasks. This application has the effect of improving the flexibility of task allocation.

Description

Technical Field

[0001] This application relates to the field of robot scheduling management, and particularly to an intelligent scheduling method and system for quadruped robots. Background Art

[0002] In a factory or warehouse environment, when multiple quadruped robots are performing tasks at multiple loading and unloading points, the task allocation and scheduling management of the robots are important links. For example, in the picking task allocation method and its goods sorting system disclosed in the Chinese patent application document with the publication number CN112967002B, the picking task allocation method includes: determining the priority of the picking task according to the target goods location of the picking task; allocating at least one picking task to the robot according to the current location of the robot and / or the priority of the picking task; calculating the total proportion of the completed picking task indicators in the total picking task indicators; and when the total proportion is greater than a preset first completion threshold, allocating the picking task with the highest priority to the robot.

[0003] In a multi-robot scenario, the prior art solutions do not fully consider the collaborative work among robots. For example, when multiple robots approach high-priority tasks simultaneously, task conflicts or resource waste may occur, and the flexibility of task allocation is insufficient. Summary of the Invention

[0004] To solve the technical problem of low flexibility in task classification, this application provides an intelligent scheduling method and system for quadruped robots.

[0005] In a first aspect, this application provides an intelligent scheduling method and system for quadruped robots, adopting the following technical solutions:

[0006] A method for intelligent scheduling of a quadruped robot, comprising the steps of: constructing and training a graph neural network; the nodes of the graph neural network represent the coordinates of loading and unloading points, the weight of goods, and the initial power of the robot; the edges of the graph neural network represent the routes between nodes, and the input of the graph neural network is the adjacency matrix and the node feature matrix; the output is the predicted power consumption value and the predicted completion time value of the task; obtaining the task route, and selecting the robots at the nodes directly associated with the nodes passed by the task route and the robots at the nodes passed by the task route as target robots; calculating the bid value of the target robots; obtaining the priority coefficients of each task, using the priority coefficients and the bid values as the edge weights for the matching of tasks and robots, and taking the task allocation plan with the largest cumulative sum of edge weights as the matching plan, and performing global matching on multiple robots and multiple tasks according to the KM matching algorithm; the calculation method of the bid value is: according to the predicted power consumption value, the predicted completion time value of each target robot under the same task and a preset constraint function, calculate the bid value of each target robot for the same task, the predicted power consumption value and the bid value are in an inverse relationship, and the predicted completion time value and the bid value are in an inverse relationship; the constraint function expresses that when the predicted power consumption value is greater than the initial power of the target robot, the bid value of this robot is set to 0; in the graph neural network, direct association means that there is a direct edge connection between the nodes in the graph.

[0007] By having each robot bid for each task, the bid value reflects the completion cost (power consumption for completing the task) and completion quality (time for completing the task) of the robot when completing the task. The lower the completion cost and the higher the completion quality, the higher the bid value. At this time, the possibility of the robot matching with this task is higher. At the same time, calculating the bid value in combination with the constraint function can effectively prevent tasks from being assigned to robots that cannot complete the tasks, and make the bid value of robots with insufficient power be 0. Using the bid value and task priority as the edge weights for the matching of tasks and robots, combined with the prediction of power consumption and completion time by the graph neural network, and performing global matching of multiple tasks and multiple robots according to the matching algorithm, the task allocation can be dynamically adjusted according to real-time data, and the flexibility of task allocation is high;

[0008] Through the graph neural network, robots and tasks can be modeled as nodes and edges in a graph, and node features and edge features are used for task allocation, which can effectively handle the complex relationships in multi-robot task allocation, improve the efficiency and accuracy of task allocation, and can dynamically adjust the task allocation strategy through the graph neural network. For example, when the location of the task point or the state of the robot changes, tasks can be quickly reallocated; moreover, the global information processing ability of the graph neural network can achieve global optimization of multi-robot task allocation;

[0009] Directly associate the nodes passed by the task route with the robots. The purpose of selecting the target robots is to screen the robots and quickly locate the robots directly related to the task. In the case of a large number of robots, this method can significantly reduce the screening time and the computational workload when subsequent robots are matched with the task, reduce the waste of computing resources, and improve the computing efficiency.

[0010] Optionally, the expression for the output value is:

[0011] ; where represents the output value of the th target robot for the task , represents the exponential function with as the base, represents the predicted completion time of the th target robot to complete the task , represents the predicted power consumption of the th target robot to complete the task , represents the preset time adjustment parameter, represents the preset power consumption adjustment parameter; represents the constraint function; represents the initial power of the th target robot.

[0012] For a task, the smaller the predicted completion time of the robot and the smaller the predicted power consumption, the higher the output value of the robot for the task. At this time, the higher the possibility of matching between the robot and the task. When the predicted power consumption of the robot exceeds its initial power, the output value directly becomes zero, ensuring that the robot will not be assigned tasks beyond its capabilities. This not only avoids task failures caused by insufficient power but also improves the reliability of task allocation; setting multiple adjustment parameters can better adapt to different actual working conditions.

[0013] Optionally, the expression for the output value is:

[0014] ; where represents the output value of the th target robot for the task , represents the exponential function with as the base, represents the predicted completion time of the th target robot to complete the task , represents the predicted completion time of the th target robot to complete the task The predicted power consumption value, represents a preset time adjustment parameter, represents a preset power consumption adjustment parameter; represents a constraint function; represents the initial power of the nth target robot; represents a preset safe power;

[0015] In actual working conditions, the factory loading and unloading tasks usually require the robot to complete the tasks efficiently and return to charge safely to avoid downtime due to power exhaustion. The safe power is added in the calculation of the constraint conditions to reserve enough power for the robot to automatically return to the charging position.

[0016] Optionally, the expression for the cumulative sum of edge weights is: ; represents the cumulative sum of edge weights, represents the nth target robot's bid value for task ; represents the priority coefficient of task ; is the total number of target robots; represents the total number of tasks.

[0017] Multiplying the priority coefficient of each task by the bid value of the corresponding robot ensures that high-priority tasks contribute more to the overall edge weight. If a task has a high priority, the bid value of the robot assigned to that task will have a greater impact on the overall edge weight. Considering both the importance of the task and the execution ability of the robot comprehensively. It ensures that the task allocation scheme not only considers the urgency of the task but also the actual ability of the robot. The calculation method of the cumulative sum of edge weights enables the system to evaluate the quality of the task allocation scheme from a global perspective. By maximizing the cumulative sum of edge weights, the optimal task allocation scheme can be found to ensure the highest overall task execution efficiency.

[0018] Optionally, the graph neural network includes a shared feature extraction layer, independent first and second output layers. The shared feature extraction layer is connected to the first output layer and the second output layer respectively. The first output layer is used to output the predicted power consumption value, and the second output layer is used to output the predicted completion time value; The shared feature extraction layer includes: a graph convolutional layer for extracting local features of nodes; a graph attention layer for assigning different weights to different neighboring nodes through an attention mechanism, and a gated graph neural network layer for processing the dynamic transmission between nodes through a gating mechanism; The output of the graph convolutional layer is used as the input of the graph attention layer, and the output of the graph attention layer is used as the input of the gated graph neural network layer.

[0019] The role of the shared feature extraction layer is to extract the common features of all nodes, and these features can be shared and utilized by subsequent different tasks (such as power consumption prediction and completion time prediction). The graph convolutional layer extracts the local features of nodes and can capture the attribute information of the nodes themselves; the graph attention layer assigns different weights to different neighboring nodes through the attention mechanism, enabling the model to pay more attention to the neighboring nodes that are more important to the current node, thereby better utilizing the relationship information between nodes. The gated graph neural network layer processes the dynamic transmission between nodes through the gating mechanism and can better handle the long-term dependencies in the graph, avoiding the loss of information during the transmission process. The combination of the shared feature extraction layer and the independent output layer enables the model to simultaneously output the power consumption prediction value and the completion time prediction value. These two output layers correspond to different task requirements respectively, improving the versatility and adaptability of the model.

[0020] Optionally, the loss function for graph neural network training is:

[0021] + ; In the formula, represents the loss function, represents the predicted completion time value for the th target robot to complete task , represents the true completion time value for the th target robot to complete task ; represents the total number of target robots; represents the predicted power consumption value for the th target robot to complete task , represents the true power consumption value for the th target robot.

[0022] This loss function simultaneously considers the differences between the predicted completion time value and the true value as well as the differences between the predicted power consumption value and the true value. In actual working conditions, accurately predicting the time and power consumption for a robot to complete a task is crucial for task allocation. By minimizing this loss function, the graph neural network can more accurately predict these two key indicators, thereby improving the reliability of task allocation.

[0023] Optionally, the task route is a sequence of nodes, and the sequence of nodes represents the loading and unloading points that the robot needs to visit in sequence.

[0024] Representing the task route as a sequence of nodes can clearly define the order that the robot needs to visit.

[0025] Optionally, the shortest path between nodes is calculated using a path planning algorithm as the route between nodes.

[0026] Calculating the shortest path through a path planning algorithm can ensure that the robot takes the shortest route when completing tasks. In actual working conditions, the factory environment is complex, with various obstacles and restrictions. Shortest path planning can reduce the movement time and energy consumption of the robot during task execution, improving task execution efficiency.

[0027] Optionally, the path planning algorithm is the A-star algorithm or the Dijkstra algorithm.

[0028] In a second aspect, the present application provides a quadruped robot intelligent scheduling system, adopting the following technical solution:

[0029] A quadruped robot intelligent scheduling system includes: a processor and a memory, where the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned quadruped robot intelligent scheduling method is implemented.

[0030] Generate a computer program for the above-mentioned quadruped robot intelligent scheduling method and store it in the memory to be loaded and executed by the processor. Thus, a system is made according to the memory and the processor, which is convenient to use.

[0031] The present application has the following technical effects:

[0032] 1. By having each robot bid on each task, the bid value reflects the completion cost (power consumption for completing the task) and completion quality (time for completing the task) of the robot when completing the task. The lower the completion cost and the higher the completion quality, the higher the bid value. At this time, the possibility of the robot matching the task is higher. At the same time, combining the constraint function to calculate the bid value can effectively prevent tasks from being assigned to robots that cannot complete the tasks, making the bid value of robots with insufficient power zero. Using the bid value and task priority as the edge weights for task-robot matching, combined with the graph neural network to predict power consumption and completion time, and performing global matching of multiple tasks and multiple robots according to the matching algorithm, the task allocation can be dynamically adjusted according to real-time data, and the flexibility of task allocation is high.

[0033] 2. Through the graph neural network, robots and tasks can be modeled as nodes and edges in a graph, and node features and edge features are used for task allocation, which can effectively handle complex relationships in multi-robot task allocation, improving the efficiency and accuracy of task allocation. Through the graph neural network, the task allocation strategy can be dynamically adjusted. For example, when the task point location or robot status changes, tasks can be quickly reallocated; and the global information processing ability of the graph neural network can achieve global optimization of multi-robot task allocation.

[0034] 3. Directly associate the nodes passed by the task route with the robots. The purpose of selecting the target robot is to screen the robots and quickly locate the robots directly related to the task. In the case of a large number of robots, this method can significantly reduce the screening time and the computational workload when subsequent robots are matched with the task, reduce the waste of computing resources, and improve the computing efficiency. Brief Description of the Drawings

[0035] Figure 1 is a flowchart of a method for intelligent scheduling of a quadruped robot according to an embodiment of the present application. Detailed Embodiments

[0036] An embodiment of the present application discloses a method for intelligent scheduling of a quadruped robot. Referring to Figure 1 , it includes steps S1 - S4, specifically as follows:

[0037] S1: Construct and train a graph neural network; the nodes of the graph neural network represent the coordinates of loading and unloading points, the weight of goods, and the initial power of the robot; the edges of the graph neural network represent the routes between the nodes, and the input of the graph neural network is the adjacency matrix and the node feature matrix; the output is the predicted value of power consumption and the predicted value of the completion time of the task.

[0038] In the graph neural network, each loading and unloading point is a node, and the edge represents the connection relationship between the nodes, which can specifically represent the actual path of the robot from one loading and unloading point to another, and also represents the order of the loading and unloading points that the robot needs to visit in sequence. Use a path planning algorithm to calculate the shortest path between the nodes as the route between the nodes. The path planning algorithm can be the A-star algorithm or the Dijkstra algorithm, and the prior art will not be elaborated here.

[0039] The robots in the present application are robots of the same model and the same parameters, aiming to make the parameters of the robots in all aspects, such as the load limit, the total power, etc., consistent. The robots selected in the present application are quadruped robots that can move and transport goods.

[0040] In one embodiment, the input of the graph neural network is the adjacency matrix and the node feature matrix. The graph neural adjacency matrix represents the connection relationship between the nodes in the graph and is used to describe which nodes have a direct path between them. The node feature matrix contains the features of each node, such as the coordinates of the loading and unloading points, the weight of the goods, the initial power of the robot, etc. The output is the predicted value of power consumption and the predicted value of the completion time of the task.

[0041] The neural network includes a shared feature extraction layer, independent first and second output layers. The shared feature extraction layer is respectively connected to the first and second output layers. The first output layer is used to output the predicted value of power consumption, and the second output layer is used to output the predicted value of the completion time.

[0042] The shared feature extraction layer includes: a graph convolutional layer for extracting local features of nodes; a graph attention layer for assigning different weights to different neighboring nodes through an attention mechanism; and a gated graph neural network layer for processing the dynamic transmission between nodes through a gating mechanism. The output of the graph convolutional layer serves as the input to the graph attention layer, and the output of the graph attention layer serves as the input to the gated graph neural network layer.

[0043] During the training of the graph neural network, the stopping condition for model training is set to reach a preset number of training times (e.g., 1000 times) or the loss function is less than a preset loss function threshold (e.g., 0.01). The prior art is not elaborated here. The loss function for the training of the graph neural network is:

[0044] + ; where represents the loss function, represents the predicted completion time of the th target robot to complete the task, and represents the actual completion time of the th target robot to complete the task; represents the total number of target robots; represents the predicted power consumption of the th target robot to complete the task, and represents the actual power consumption of the th target robot.

[0045] S2: Obtain the task route, and select the robots at the nodes directly associated with the nodes passed by the task route and the robots at the nodes passed by the task route as target robots.

[0046] In the graph neural network, direct association means that there is a direct edge connection between nodes in the graph. In this case, information transmission occurs through directly connected nodes. Indirect association means that the information transmission between nodes is not direct but is relayed through intermediate nodes. The concepts of direct association and indirect association are prior art in the graph neural network and will not be elaborated further.

[0047] The purpose of selecting target robots is to screen the robots and quickly locate the robots directly related to the task. In the case of a large number of robots, this method can significantly reduce the screening time and the computational amount during the subsequent matching of robots with tasks, reduce the waste of computing resources, and improve the computing efficiency.

[0048] S3: Calculate the bid value of the target robot.

[0049] ​The calculation method of the output value is as follows: According to the predicted power consumption values, predicted completion time values of each target robot under the same task, and a preset constraint function, calculate the output value of each target robot for the same task. The predicted power consumption value and the output value are in an inverse relationship, and the predicted completion time value and the output value are in an inverse relationship. The constraint function means that when the predicted power consumption value is greater than the initial power of the target robot, the output value of this robot is set to 0.

[0050] In one embodiment, the expression of the output value is:

[0051] ; where represents the output value of the th target robot for task , represents the exponential function with as the base, represents the predicted completion time value of the th target robot to complete task , represents the predicted power consumption value of the th target robot to complete task , represents a preset time adjustment parameter, represents a preset power consumption adjustment parameter; represents the constraint function; represents the th target robot's initial power.

[0052] By having each robot bid for each task, the output value reflects the completion cost (power consumption for completing the task) and completion quality (time for completing the task) of the robot. The lower the completion cost and the higher the completion quality, the higher the output value. At this time, the higher the possibility of the robot matching this task. Conversely, the lower the output value, the lower the possibility of the robot matching this task.

[0053] Calculating the output value using the constraint function can effectively prevent tasks from being assigned to robots that cannot complete the tasks, and make the output value of robots with insufficient power be 0. When the predicted power consumption value of the th target robot to complete task is less than or equal to the initial power of the th target robot, it indicates that the power of the robot at this time is not enough or just enough to support the robot to complete the task. The constraint function takes 0, and the output value takes 0. The lower the output value, the lower the probability of this robot matching this task in subsequent matching. Conversely, the constraint function takes 1.

[0054] The time adjustment parameter is used to adjust the influence degree of the task completion time on the output value. When the time adjustment parameter increases, the attenuation effect of time on the output value is enhanced, that is, the longer the completion time, the faster the output value decreases; conversely, the attenuation effect of time on the output value is weakened. The power consumption adjustment parameter is used to adjust the influence degree of the power consumption on the output value. When the power consumption adjustment parameter increases, the attenuation effect of the power consumption on the output value is enhanced, that is, the greater the power consumption, the faster the output value decreases; conversely, the attenuation effect of the power consumption on the output value is weakened. Exemplarily, , , the values of the time adjustment parameter and the power consumption adjustment parameter can be adjusted according to the actual application scenario.

[0055] In other embodiments, the constraint function can also be:

[0056] ; represents the constraint function; represents the initial power of the th target robot; represents the preset safe power; represents the preset power adjustment parameter.

[0057] In the actual working condition, the factory loading and unloading task usually requires the robot to complete the task efficiently and return to charge safely to avoid downtime due to power exhaustion. The safe power is added in the calculation of the constraint condition, which is the power reserved for the robot to automatically return to the charging position. The power adjustment parameter is used to adjust the weight of the safe power, so as to control the allowable power consumption range of the robot when performing tasks. Exemplarily, , the value of can be adjusted according to the actual application scenario.

[0058] S4: Obtain the priority coefficients of each task, use the priority coefficient and the output value as the edge weights for the matching of tasks and robots, and use the task assignment scheme with the largest sum of edge weights as the matching scheme, and perform global matching on multiple robots and multiple tasks according to the KM matching algorithm.

[0059] The KM matching algorithm is an algorithm used to solve the maximum weight matching problem of a weighted bipartite graph. In the KM matching algorithm, the edge weight refers to the weight of each edge in the bipartite graph, which reflects the "value" or "importance" of the edge. The KM matching algorithm and the edge weights are both prior arts and will not be elaborated.

[0060] The expression of the sum of edge weights is: ; represents the sum of edge weights, represents the th target robot's output value for task ; Represents the priority coefficient of the task ; is the total number of target robots; Represents the task total number.

[0061] Obtain the tasks input by the user and the urgency of the tasks. Different levels of urgency can be set with different priority coefficients. The larger the priority coefficient, the more urgent the task. Exemplarily, the tasks are divided into two categories, urgent tasks and normal tasks. The priority coefficient of urgent tasks is 1.2, and the priority coefficient of normal tasks is 1.

[0062] Multiplying the priority coefficient of each task by the bid price of the corresponding robot ensures that high-priority tasks contribute more to the overall edge weight. If a task has a high priority, then the bid price of the robot assigned to this task will have a greater impact on the overall edge weight.

[0063] Comprehensively consider the importance of the task and the execution ability of the robot. Ensure that the task allocation scheme not only considers the urgency of the task, but also considers the actual ability of the robot. The calculation method of the cumulative sum of edge weights enables the system to evaluate the quality of the task allocation scheme from a global perspective. By maximizing the cumulative sum of edge weights, the optimal task allocation scheme can be found to ensure the highest overall task execution efficiency.

[0064] The embodiment of the present application also discloses a quadruped robot intelligent scheduling system, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the quadruped robot intelligent scheduling method according to the present application is implemented.

[0065] The above system also includes other components well known to those skilled in the art such as a communication bus and a communication interface. Their settings and functions are known in the art, so they will not be described in detail here.

[0066] In the present application, the aforementioned memory can be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, device, or device. For example, a computer-readable storage medium can be any suitable magnetic storage medium or magneto-optical storage medium, such as, a resistive random access memory, a dynamic random access memory, a static random access memory, etc., or any other medium that can be used to store the required information and can be accessed by an application program, module, or both. Any such computer storage medium can be part of the device or accessible or connectable to the device.

[0067] The above are all preferred embodiments of the present application. The protection scope of the present application is not limited by this. Therefore, all equivalent changes made according to the structure, shape, and principle of the present application should be covered within the protection scope of the present application.

Claims

1. An intelligent scheduling method for a quadruped robot, characterized in that, Including the steps: Construct and train a graph neural network; the nodes of the graph neural network represent the coordinates of loading and unloading points, the weight of goods, and the initial power of the robot; the edges of the graph neural network represent the routes between nodes, and the input of the graph neural network is the adjacency matrix and the node feature matrix; The output is the predicted power consumption value and the predicted completion time value of the task; obtain the task route, and select the robots at the nodes directly associated with the nodes passed by the task route and the robots at the nodes passed by the task route as the target robots; Calculate the bid value of the target robot; obtain the priority coefficient of each task, use the priority coefficient and the bid value as the edge weights for the matching of tasks and robots, and use the task allocation plan with the largest cumulative sum of edge weights as the matching plan, and perform global matching on multiple robots and multiple tasks according to the KM matching algorithm; The calculation method of the bid value is: according to the predicted power consumption value, the predicted completion time value of each target robot under the same task and a preset constraint function, calculate the bid value of each target robot for the same task. The predicted power consumption value and the bid value are in an inverse relationship, and the predicted completion time value and the bid value are in an inverse relationship; the constraint function means that when the predicted power consumption value is greater than the initial power of the target robot, the bid value of this robot is set to 0; In the graph neural network, direct association means that there is a direct edge connection between the nodes in the graph.

2. The intelligent scheduling method for a quadruped robot according to claim 1, wherein The expression of the bid value is: ； Among them, represents the th bid value of the target robot for the task . represents the exponential function with as the base. represents the predicted completion time of the th target robot for completing the task . represents the predicted power consumption of the th target robot for completing the task . represents the preset time adjustment parameter, represents the preset power consumption adjustment parameter; represents the constraint function; represents the th initial battery level of the target robot.

3. The intelligent scheduling method for a quadruped robot according to claim 1, characterized in that, The expression of the bid value is: ； Among them, represents the th bid value of the target robot for the task , represents the exponential function with as the base, represents the predicted completion time of the th target robot to complete the task , represents the predicted power consumption of the th target robot to complete the task , represents the preset time adjustment parameter, represents the preset power consumption adjustment parameter; represents the constraint function; represents the initial power of the th target robot; represents the preset safety power; represents the preset power adjustment parameter.

4. The intelligent scheduling method for a quadruped robot according to claim 1, wherein, The expression for the cumulative sum of edge weights is: ; represents the cumulative sum of edge weights, represents the bid value of the th target robot for task ; represents the priority coefficient of task ; is the total number of target robots; represents the total number of tasks.

5. The intelligent scheduling method for a quadruped robot according to claim 1, wherein, The graph neural network includes a shared feature extraction layer, independent first and second output layers. The shared feature extraction layer is respectively connected to the first output layer and the second output layer. The first output layer is used to output the predicted power consumption value, and the second output layer is used to output the predicted completion time value; The shared feature extraction layer includes: a graph convolutional layer for extracting local features of nodes; a graph attention layer for assigning different weights to different neighboring nodes through an attention mechanism, and a gated graph neural network layer for processing the dynamic transmission between nodes through a gating mechanism; the output of the graph convolutional layer is used as the input of the graph attention layer, and the output of the graph attention layer is used as the input of the gated graph neural network layer.

6. The intelligent scheduling method for a quadruped robot according to claim 1, wherein The loss function for training the graph neural network is: + ; Wherein, represents the loss function, represents the predicted completion time of the th target robot to complete the task ; represents the true completion time of the th target robot to complete the task ; represents the total number of target robots; represents the predicted power consumption of the th target robot to complete the task ; represents the true power consumption of the th target robot.

7. The intelligent scheduling method for a quadruped robot according to claim 1, wherein, The task route is a node sequence, and the node sequence represents the loading and unloading points that the robot needs to visit in sequence.

8. The intelligent scheduling method for a quadruped robot according to claim 1, wherein, Use a path planning algorithm to calculate the shortest path between nodes as the route between nodes.

9. The intelligent scheduling method for a quadruped robot according to claim 8, wherein The path planning algorithm is the A-star algorithm or the Dijkstra algorithm.

10. A quadruped robot intelligent scheduling system, characterized in that, Including: A processor and a memory, the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the intelligent scheduling method of the quadruped robot according to any one of claims 1-9 is implemented.

Images