A fast assignment scheduling method for multi-task scenario of multi-robot system

Abstract

The application discloses a kind of multi-arm system's quick distribution scheduling method for multi-task scene, it is related to multi-arm system field, according to the workspace model of mechanical arm, task set is prepackaged offline;To be distributed mechanical arm, carry out online distribution scheduling.The application combines offline grouping and online deployment, reduces the number of iteration traversal when online distribution scheduling by offline prepackaging, so as to reduce distribution time, so that the quick distribution scheduling problem of multiple mechanical arms in multi-task scene can be solved, compared with offline algorithm, the single scheduling time required by this task is shorter, meet the control period of mechanical arm, can be quickly deployed and applied in some time requirement industrial scene, if there is unknown task increase, it can be quickly distributed to corresponding group by offline prepackaging algorithm, then respond through online distribution scheduling process.

Description

Technical Field

[0001] This invention relates to the field of multi-robotic arm systems, and more particularly to a rapid allocation and scheduling method for multi-robotic arm systems in multi-task scenarios. Background Technology

[0002] Traditional single-robot systems often suffer from structural limitations and a small workspace, making it difficult to handle multiple tasks simultaneously. Multi-robot systems, on the other hand, offer significant advantages in processing complex tasks, such as strong adaptability, high execution efficiency, and high fault tolerance. Based on these characteristics, research and applications of multi-robot systems are becoming increasingly sophisticated.

[0003] In multi-arm robotic systems, the workspaces of the robotic arms often involve some interaction. In such cases, due to the numerous tasks the system performs, collisions with other robots are common. Therefore, the system itself needs to allocate and schedule the robotic arms in time and space to achieve more coordinated and efficient movement. Furthermore, if a new task is suddenly added during system operation, the system needs to respond quickly to the new task without requiring system replanning.

[0004] Currently, most existing allocation and scheduling methods for multi-robotic arm systems focus on handling and responding to problems where the set of task objectives is completely known. These methods involve pre-establishing a global optimization model for the entire system and then using extensive computation to iteratively solve the model to achieve the objective. This often requires that the global information of the system is known, meaning that all tasks in the multi-robotic arm system are known. In other words, if an unknown task objective is suddenly added during execution, these offline methods cannot respond quickly. In such cases, the system often needs to rebuild the optimization model and solve it again, which is very time-consuming. Furthermore, for scenarios with high flexibility and real-time requirements, this offline method is completely unsuitable.

[0005] To increase the efficiency of scheduling and allocation, this invention develops a method for rapid task allocation and scheduling of multi-robot systems in multi-task scenarios. This method is more suitable for scenarios with high flexibility and strong real-time requirements. It can quickly allocate and schedule the next target for each robot arm online, and can also quickly process and respond to unknown tasks added during system operation. Summary of the Invention

[0006] In view of the above-mentioned deficiencies of the prior art, the technical problem to be solved by the present invention is how to improve the running speed of multi-robotic arm task allocation and scheduling, and how to quickly respond to new and unknown tasks added during operation.

[0007] To achieve the above objectives, this invention provides a rapid allocation and scheduling method for multi-robot systems in multi-task scenarios, comprising the following steps:

[0008] The task set is pre-grouped offline based on the workspace model of the robotic arm:

[0009] The workspace model is established based on the installation position and kinematic model of the robotic arm. Based on the workspace model, it is determined whether the robotic arm can complete the tasks in the task set. For tasks that can be completed by at least one robotic arm, they are assigned to the shared task group of the corresponding robotic arm. For tasks that can be completed by only one robotic arm, they are assigned to the independent task group of the corresponding robotic arm. For tasks that cannot be completed by any robotic arm, an error is reported and the process is exited.

[0010] Online allocation and scheduling of robotic arms to be assigned:

[0011] Obtain the previous task pose information and the current target task pose information of each robotic arm other than the robotic arm to be assigned, and obtain the task pose information set An; the motion planning algorithm is a consistent and measurable point-to-point motion planning algorithm.

[0012] Using the task pose information set An and the motion planning algorithm, a motion envelope is established for each robotic arm other than the robotic arm to be assigned, resulting in a motion envelope set EAn.

[0013] Traverse the set Bn of all task groups to be executed corresponding to the robotic arm to be assigned. The set Bn of task groups to be executed consists of all independent task groups and shared task groups to be executed corresponding to the robotic arm to be assigned.

[0014] For each task Bi to be executed in the set of tasks to be executed Bn, obtain the pose information of the previous task of the robotic arm to be assigned and the pose information of the task Bi to be executed to obtain the task pose information set BAI. Through the task pose information set BAI and the motion planning algorithm, establish the motion envelope EBi to be detected of the robotic arm to be assigned.

[0015] Select tasks Bi that do not intersect with the motion envelope EBi to be detected and the motion envelope set EAn, and add them to the feasible task group of the robotic arm to be assigned.

[0016] In the feasible task group, a greedy algorithm is used to optimize the task with the minimum motion cost, which is then selected as the next target task for the robotic arm to be assigned and removed from the set of tasks to be executed, Bn.

[0017] The motion cost is the motion cost of the robotic arm to be assigned, measured by a motion planning algorithm, plus the actual execution cost of the robotic arm to be assigned for the task.

[0018] Furthermore, when the task set includes obstacle avoidance scenarios, the workspace model is established based on the installation position of the robotic arm, the kinematic model, and the position and shape of the obstacles.

[0019] Further, determining whether the robotic arm can complete the tasks in the task set includes: inputting the Cartesian length between the base of the robotic arm and the task, and the angle between the quaternion of the posture of the base of the robotic arm and the posture of the task, into the inverse solver of the robotic arm; if the inverse solver has an inverse solution, it is determined that the robotic arm can complete the task; otherwise, it is determined that the robotic arm cannot complete the task.

[0020] Furthermore, the consistent and measurable point-to-point motion planning algorithm is either a gradient-based motion planning algorithm or a stochastic trajectory planning algorithm.

[0021] Furthermore, the motion envelope is established using convex hull fitting.

[0022] Furthermore, the cost of motion is one or more of motion time, motion energy, and motion torque.

[0023] Furthermore, when the robotic arm is in an unassigned state, its previous task pose and current target task pose are both the initial preparation pose.

[0024] Furthermore, when the robotic arm to be assigned completes all of its corresponding sets of tasks to be executed, Bn, the pose of the next target task is set as the termination pose.

[0025] Furthermore, when the feasible task group is empty, the pose of the next target task of the robotic arm to be assigned is set as the initial preparation pose.

[0026] Furthermore, before the online allocation and scheduling begins, it is detected whether there are any tasks to be executed. If there are no tasks to be executed, the system terminates; if there are tasks to be executed, the system enters the online allocation and scheduling process.

[0027] The technical effects of this invention are as follows:

[0028] 1. The task set is pre-grouped offline based on the workspace of the robotic arm. The offline pre-grouping process reduces the number of iterations during online allocation and scheduling, thereby reducing allocation time. Furthermore, offline pre-grouping can preserve the integrity of tasks that can be executed by multiple robotic arms.

[0029] 2. During the online allocation process, convex bodies are used to describe the motion envelope, enabling rapid collision detection;

[0030] 3. By calculating the motion cost in the feasible task group through motion planning algorithm, and then combining it with greedy algorithm to optimize the task with the minimum cost, the motion process can be optimized to achieve local optimum of overall motion.

[0031] 4. This invention can solve the problem of rapid allocation and scheduling of multiple robotic arms in multi-task scenarios. Compared with offline algorithms, the single scheduling time required for this task is shorter, which meets the control cycle of the robotic arm and can be quickly deployed and applied in some time-sensitive industrial scenarios.

[0032] 5. If an unknown task is added during system operation, it can be quickly assigned to the corresponding group using an offline pre-grouping algorithm, and then responded through an online allocation and scheduling process. This allows the system to add tasks during operation, which is something that some offline algorithms and online algorithms that rely on the initial process cannot achieve. Therefore, this invention can meet the needs of some industrial operation scenarios that require rapid response and task addition.

[0033] The following will further explain the concept, specific structure, and technical effects of the present invention in conjunction with the accompanying drawings, so as to fully understand the purpose, features, and effects of the present invention. Attached Figure Description

[0034] Figure 1 This is a flowchart of an offline pre-grouping process according to a preferred embodiment of the present invention;

[0035] Figure 2 This is a flowchart of an online allocation and scheduling process according to a preferred embodiment of the present invention;

[0036] Figure 3 This is a preferred embodiment of the present invention, which describes offline pre-grouping and online allocation scheduling. Detailed Implementation

[0037] The following description, with reference to the accompanying drawings, illustrates several preferred embodiments of the present invention to make its technical content clearer and easier to understand. The present invention can be embodied in many different forms, and the scope of protection of the present invention is not limited to the embodiments mentioned herein.

[0038] In the accompanying drawings, components with the same structure are indicated by the same numerical designation, and components with similar structures or functions are indicated by similar numerical designations. The dimensions and thicknesses of each component shown in the drawings are arbitrary, and the present invention does not limit the dimensions and thicknesses of each component. To make the illustrations clearer, the thickness of some components has been appropriately exaggerated in the drawings.

[0039] like Figure 1The diagram shows the offline pre-grouping flowchart. Before the system runs, the existing task set is pre-grouped offline based on the workspace model of the robotic arm. First, the distance between each task and each robotic arm base in the task set is calculated. This distance includes the length of the Cartesian axis and the angle formed by the attitude quaternion. These two values ​​are input into the inverse kinematics of the corresponding robotic arm. The existence of an inverse kinematics quickly determines whether the task can be completed. If no robotic arm can complete the task, meaning there is no inverse kinematics for any robotic arm, then all robotic arms are unreachable for this task and cannot execute it. The program then reports an error and exits. If at least one inverse kinematics exists, the robotic arm is considered capable of completing the task. If multiple robotic arms can complete the task, the task is added to a shared task group for the corresponding multiple robotic arms. If only one robotic arm can complete the task, the task is added to an independent task group for that robotic arm, thus completing the task allocation.

[0040] Figure 2 This is an online task allocation and scheduling flowchart. When the system runs, it first checks whether there are any tasks to be processed.

[0041] If all tasks in the system are completed, the system terminates, ceasing allocation and scheduling.

[0042] If there are tasks to be performed, the system operates normally and monitors the movement status of the robotic arm.

[0043] When a robotic arm completes a task and needs to be assigned the next task, there are robotic arms waiting to be assigned. At the current moment, a motion envelope is established for the other robotic arms.

[0044] When the current independent task group and shared task group involved in the robotic arm are empty, set the termination pose to the target task pose of the robotic arm;

[0045] If the independent task group and shared task group involved by the current robotic arm are not empty, establish a motion envelope to be detected from the current pose to the pose of each task to be executed; add the tasks corresponding to the motion envelopes that do not intersect with the motion envelopes formed by other robotic arms to the feasible task group.

[0046] When the feasible task group is empty, set the initial ready pose to the target task pose of the robotic arm.

[0047] If the feasible task group is not empty, calculate the running cost of the tasks in the feasible task group; take the task with the minimum running cost as the next target and delete the task from the corresponding independent task group or shared task group.

[0048] When an unknown new task is added to the system during the process, a process such as... Figure 1The offline pre-grouping process shown adds the tasks to the corresponding robotic arm's independent or shared task groups. Then, it proceeds to... Figure 2 The online allocation and scheduling process is shown.

[0049] like Figure 3 The diagram shown illustrates offline pre-grouping and online allocation scheduling, where m x m z m y This represents three robotic arms, with P1-P10 representing 10 tasks, such as... Figure 3 (a) shows a schematic diagram of offline pre-grouping, based on... Figure 1 The flowchart shows that the existing task set of 10 tasks (P1-P10) will be assigned to 3 independent task groups (Px, Py, Pz) and 4 shared task groups (Pxy, Pyz, Pxz, and Pxyz), where,

[0050] Px indicates that it can only be generated by the robotic arm m. x Completed independent task groups,

[0051] Py indicates that it can only be generated by the robotic arm m. y Completed independent task groups,

[0052] Pz indicates that it can only be generated by the robotic arm m. z Completed independent task groups,

[0053] Pxy represents the ability to be generated by the robotic arm m x or m y Completed shared task group,

[0054] Pyz indicates that it can be generated by the robotic arm m y or m z Completed shared task group,

[0055] Pxz indicates that it can be generated by the robotic arm m x or m z Completed shared task group,

[0056] Pxyz indicates that it can be generated by the robotic arm m x or m z or m y Completed shared task group.

[0057] The 10 tasks in the diagram are finally grouped as follows:

[0058] Tasks P1, P2, and P3 have been added to Px.

[0059] Tasks P7 and P8 have been added to Python.

[0060] Mission P10 has been added to Pz.

[0061] Task P9 has been added to PyZ.

[0062] Tasks P5 and P6 have been added to Pxz.

[0063] Mission P4 has been added to Pxyz.

[0064] In this example, Pxy is an empty set.

[0065] Figure 3 (b)-(d) illustrate the operation diagrams of the overall system during online allocation and scheduling, using a scenario with three robotic arms as an example. Figure 3 As shown in (b), when the robotic arm mx needs to be assigned the next task, the other robotic arms establish motion envelopes Ey and Ez based on the pose of the previous task and the pose of the current target.

[0066] Then, for the independent and shared task groups of the robotic arm mx, detectable motion envelopes Ex1, Ex2, and Ex3 are established from the pose of the current task P2 to the poses of each task to be executed, P1, P3, and P5. If the detectable motion envelope does not intersect with the motion envelopes of other robotic arms, the corresponding task is added to the feasible task group Fx at the current time. As shown in the figure, Ex1 intersects with Ez, while Ex2 and Ex3 do not intersect with Ey and Ez. Therefore, only P1 and P3 are added to the feasible task group Fx at the current time.

[0067] Then, for feasible task groups, the motion costs C1 and C2 are evaluated in conjunction with the task itself, such as... Figure 3 As shown in (c). Finally, among all tasks, the task with the shortest cost is obtained using a greedy algorithm and is selected as the next target for the currently assigned robotic arm. This task is then removed from the original independent task group or shared task group. Figure 3 (d) shows the final allocation results for the robotic arm mx.

[0068] In some embodiments, the specific process of the fast allocation and scheduling method for a multi-robotic arm system in a multi-task scenario is as follows:

[0069] Offline pre-grouping preparation:

[0070] P1: To model the system's working scenario, the following information is needed: the kinematic and collision models of the robotic arm, and the installation position of the robotic arm. If the system involves obstacle avoidance in complex scenarios, the position and shape information of obstacles in the system should also be obtained.

[0071] When there is a priority for the type of robotic arm, the robotic arms are numbered according to the priority. Robotic arms with higher priority numbers are assigned to schedule first. If there is no specific priority for the type of robotic arm, the number only represents the order and has no special meaning.

[0072] P2: A workspace model for the robotic arm is pre-established based on its installation location and kinematic model. Then, the existing task set of the system is pre-grouped offline based on this workspace model. For tasks that exist only in a specific robotic arm's workspace model (i.e., only that robotic arm can complete the task), they are added to the corresponding robotic arm's independent task group. For tasks that exist in a shared workspace model of multiple robotic arms (i.e., all of these robotic arms can complete the task), they are added to the corresponding shared task group of multiple robotic arms to preserve task completeness (completeness here refers to the probability that a task will be assigned to multiple robotic arms; during pre-grouping, the task may be assigned to any of the relevant robotic arms, which is probabilistically complete).

[0073] P3: Predetermine a consistent and measurable point-to-point motion planning algorithm. Consistency means that for a given start and end point, the algorithm must converge to a unique solution. Some sampling-based planning algorithms lack consistency, cannot be measured, and their running cost cannot be determined. Therefore, optimization-based algorithms are needed here, such as the gradient-based motion planning algorithm CHOMP (Covariant Hamilton Optimization Motion Planning) and the stochastic trajectory planning algorithm STOMP (Stochastic Trajectory Optimization Motion Planning). If obstacle avoidance in complex environments is considered, this obstacle avoidance process needs to be incorporated into the motion planning algorithm.

[0074] Online allocation and scheduling officially commenced operation:

[0075] W1: During the first allocation, since all robotic arms are in a state of waiting to be allocated, and the system executes serially, the allocation program is called according to the number of the robotic arm. The robotic arm with the first number runs directly, and the subsequent robotic arms are called in sequence according to the allocation program, i.e., steps W2-W8.

[0076] W2: At the start of the allocation process, the system first reads historical allocation data to obtain the state information of other robotic arms besides the currently assigned one. This state information includes its previous task pose and current target task pose. When a robotic arm is unassigned, its previous task pose and current target task pose are set as the initial preparation pose. A motion envelope is then established for each of the other robotic arms using these two task poses and the corresponding motion planning algorithm. Here, the motion envelope refers to the spatial set swept by the robotic arm during its movement from the previous pose to the target pose. This is a concave body, which is not conducive to collision detection. In actual calculations, a larger convex hull is used to fit the motion envelope to accelerate collision detection.

[0077] W3: For the robotic arm currently awaiting assignment, traverse its associated task groups, including its independent task groups and all shared task groups. Then, for each task, construct the motion envelope to be detected from the current pose to the pose of the task to be executed.

[0078] W4: Determine if a collision has occurred based on whether the motion envelope of the target robot intersects with the motion envelopes of other robotic arms, thus determining the feasibility of the task. Add the feasible task to the feasible task group of the currently assigned robotic arm.

[0079] W5: In the feasible task group, the motion cost of the robotic arm is measured using the motion planning algorithm determined by P3, and then the actual execution cost of the task is considered (the actual execution cost of the task is specified when the task is defined, or the cost of executing the task alone). The two costs are added together to obtain the final cost of the task. The motion cost here generally refers to the motion time. In a typical multi-robotic arm system, the overall motion time is often the primary consideration, but other indicators can also be used as the cost, such as minimizing the robotic arm's motion (energy cost) or minimizing the output torque (power cost), etc.

[0080] W6: Based on the final cost of the tasks in the feasible task group, a greedy algorithm is used to optimize and select the task with the minimum cost, which becomes the next task for the currently assigned robotic arm, thus completing the assignment process. By obtaining the current optimal solution for each robotic arm, the algorithm eventually converges to the global-local optimal solution of the overall system.

[0081] W7: For a given robotic arm to be tested, if the current feasible task group is empty, the next target task pose is designated as the initial preparation pose to avoid interlocking. Interlocking here refers to the situation where, if the robotic arm remains stationary without a designated target task, it will occupy space. If all currently feasible tasks are occupied by another robotic arm, and the arm itself is also occupying tasks from other robotic arms, the system will stall; this is called interlocking.

[0082] W8: If a robotic arm completes all related tasks, then specify its target task pose as the termination pose to complete the allocation and scheduling for that robotic arm.

[0083] W9: The system stops when all robotic arms have completed their tasks.

[0084] New unknown mission added:

[0085] Because the overall system is allocated and scheduled online, and the motion trajectory is calculated online, it has good real-time performance. Therefore, for unknown new tasks, the system can quickly assign tasks to the corresponding groups by calling an offline pre-grouping algorithm, and then respond through the online allocation and scheduling process. This ensures coordinated and rapid movement of the entire system without collisions and allows for rapid response to new tasks.

[0086] This invention proposes a rapid task allocation and scheduling method for multiple robotic arms in multi-task scenarios. The allocation steps are W2 to W8. The proposed architecture can quickly assign the next task target to the robotic arm based on current information. Furthermore, since the method performs online allocation and scheduling, it does not rely on the initial trajectory. Therefore, the system can accept newly added tasks online and can quickly allocate unknown tasks. When an unknown task joins the system during movement, an offline pre-grouping process is performed to add it to the corresponding task group. Then, it is allocated equally with existing tasks. This method can meet the needs of industrial operation scenarios requiring rapid response and task addition. The motion scheduling process of this invention is more coordinated, faster, and more efficient, while also enabling rapid response to new tasks.

[0087] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A fast allocation and scheduling method for multi-robot systems in multi-task scenarios, characterized in that, Includes the following steps: The task set is pre-grouped offline based on the workspace model of the robotic arm: The workspace model is established based on the installation position and kinematic model of the robotic arm. Based on the workspace model, it is determined whether the robotic arm can complete the tasks in the task set. For tasks that can be completed by at least one robotic arm, they are assigned to the shared task group of the corresponding robotic arm. For tasks that can be completed by only one robotic arm, they are assigned to the independent task group of the corresponding robotic arm. For tasks that cannot be completed by any robotic arm, an error is reported and the process is exited. Online allocation and scheduling of robotic arms to be assigned: Obtain the previous task pose information and the current target task pose information of each robotic arm other than the robotic arm to be assigned, and obtain the task pose information set An; Using the task pose information set An and the motion planning algorithm, a motion envelope is established for each robotic arm other than the robotic arm to be assigned, resulting in a motion envelope set EAn; the motion planning algorithm is a consistent and measurable point-to-point motion planning algorithm. Traverse the set Bn of all task groups to be executed corresponding to the robotic arm to be assigned. The set Bn of task groups to be executed consists of all independent task groups and shared task groups to be executed corresponding to the robotic arm to be assigned. For each task Bi to be executed in the set of tasks to be executed Bn, the current task pose information of the robot arm to be assigned and the pose information of the task Bi to be executed are obtained to obtain the task pose information set BAI. Through the task pose information set BAI and the motion planning algorithm, the motion envelope EBi to be detected of the robot arm to be assigned is established. Select tasks Bi that do not intersect with the motion envelope EBi to be detected and the motion envelope set EAn, and add them to the feasible task group of the robotic arm to be assigned. In the feasible task group, a greedy algorithm is used to optimize the task with the minimum motion cost, which is then selected as the next target task for the robotic arm to be assigned and removed from the set of tasks to be executed, Bn. The motion cost is the motion cost of the robotic arm to be assigned, measured by a motion planning algorithm, plus the actual execution cost of the robotic arm to be assigned for the task. 2.The fast assignment scheduling method for multi-task-scene oriented multi-arm system of claim 1, wherein, When the task set includes obstacle avoidance scenarios, the workspace model is established based on the installation position of the robotic arm, the kinematic model, and the position and shape of the obstacles. 3.The fast assignment scheduling method for multi-task-scene oriented multi-arm system of claim 1, wherein, Determining whether a robotic arm can complete a task in the task set includes: inputting the Cartesian length between the base of the robotic arm and the task, and the angle between the quaternion of the posture of the base of the robotic arm and the posture of the task, into the inverse kinematics of the robotic arm. If the inverse kinematics has an inverse solution, it is determined that the robotic arm can complete the task; otherwise, it is determined that the robotic arm cannot complete the task. 4.The fast assignment scheduling method for multi-task-scene oriented multi-arm system of claim 1, wherein, The consistent and measurable point-to-point motion planning algorithm is either a gradient-based motion planning algorithm or a stochastic trajectory planning algorithm.

5. The rapid allocation and scheduling method for multi-robot systems in multi-task scenarios as described in claim 1, characterized in that, The motion envelope is established using convex hull fitting.

6. The rapid allocation and scheduling method for a multi-robot system in a multi-task scenario as described in claim 1, characterized in that, The cost of motion is one or more of motion time, motion energy, and motion torque.

7. The fast assignment scheduling method for multi-task oriented multi-robot system of claim 1, wherein, When the robotic arm is unassigned, both the previous task pose and the current target task pose of the robotic arm are set to the starting preparation pose.

8. The rapid allocation and scheduling method for a multi-robot system in a multi-task scenario as described in claim 1, characterized in that, When the robotic arm to be assigned completes all of its corresponding task groups Bn, the next target task pose of the robotic arm to be assigned is set as the termination pose. 9.The fast assignment scheduling method for multi-task oriented multi-manipulator scene of claim 1, wherein, When the feasible task group is empty, the next target task pose of the robotic arm to be assigned is set as the initial preparation pose.

10. The fast assignment scheduling method for multi-task oriented multi-robot system of claim 1, wherein, Before the online allocation and scheduling begins, it is checked whether there are any tasks to be executed. If there are no tasks to be executed, the system terminates. If there are tasks to be executed, the system enters the online allocation and scheduling process.

Images