Double-arm collaborative disassembly planning and control method and system for retired power battery pack

Abstract

The invention relates to a double-arm collaborative disassembly planning and control method and system for a retired power battery pack, and relates to the technical field of robot control, and the method comprises the steps: collecting task trajectory data when an operator guides a robot mechanical arm to complete a battery disassembly task demonstration, and constructing a double-arm collaborative execution strategy. Then, a self-adaptive weight distribution strategy based on a task stage is introduced, a hierarchical Gaussian process model is constructed, space-time dependence between sub-targets is analyzed in a high layer, and a trajectory segment is analyzed in a low layer; variable impedance control is adopted, a double-arm coupling Cartesian impedance control strategy is established, and smooth control over the pose is achieved through rigidity and damping. Finally, in the double-arm cooperative control process, the rigidity is dynamically adjusted by fusing the uncertainty of a high layer and a bottom layer, and the motion speed and interaction force of the robot are limited by controlling a constraint strategy in combination with the pose and maximum force constraint, so that the adaptability, operation safety and cooperation efficiency of the robot are remarkably improved in complex industrial scenes such as battery disassembly.

Description

Technical Field

[0001] This application relates to the field of robot control technology, and in particular to a dual-arm collaborative dismantling planning and control method and system for retired power battery packs. Background Technology

[0003] Currently, the dismantling technology for retired power batteries still relies mainly on manual or semi-automated methods, which have significant technical limitations. In terms of manual dismantling, the complex three-tiered assembly structure of battery packs (cell-module-pack) and the significant differences in fastening methods and adhesive processes between different battery models lead to low dismantling efficiency, high labor intensity, and serious safety hazards. Improper operation can easily cause accidents such as short circuits and thermal runaway. More seriously, manual dismantling is prone to damaging battery separators and mixing electrode materials, resulting in a decrease in the purity of subsequently recycled metals and directly affecting the resource recycling value.

[0004] To address the problems caused by reliance on manual disassembly in battery disassembly technology, existing technologies have introduced robot-assisted disassembly, such as single-arm or dual-arm robots. Specifically, while single-arm robot automated disassembly systems improve efficiency to some extent, they show significant shortcomings when dealing with the collaborative operation requirements of multiple components within the battery pack, especially in delicate processes such as disassembling copper connectors and disconnecting signal lines. Furthermore, the required precision is difficult to meet in confined spaces. Dual-arm robots, although capable of collaborative operation, rely on manual programming by experts, using predefined trajectories and force control parameters. This rigid control strategy cannot adapt to the diversity of component sizes, positions, and connection methods, nor can it cope with dynamic environments such as changes in object position and sudden disturbances. In addition, impedance control is involved when robots perform disassembly tasks, and existing impedance control methods lack dynamic stiffness adjustment for task uncertainties, which can easily cause component damage and poses a risk of robot overload.

[0005] It is evident that existing technologies have operational shortcomings when robots perform battery disassembly tasks, such as inability to adapt to complex operations, low operational precision, and insufficient operational safety. Summary of the Invention

[0006] This application provides a planning and control method and system for the collaborative dismantling of retired power battery packs using dual arms. First, a hierarchical Gaussian process model based on adaptive weights of task stages is used to decouple the task level from the motion level. Then, coupling impedance control ensures the accuracy of dual-arm collaboration. Dynamic stiffness adjustment balances the flexibility and accuracy of human-machine interaction. This significantly improves task adaptability, operational safety, and human-machine collaboration efficiency in complex industrial scenarios such as battery dismantling, solving the problems of low operational adaptability and insufficient accuracy of robots performing battery dismantling tasks in existing technologies.

[0007] Firstly, this application provides a method for planning and controlling the collaborative dismantling of retired power battery packs using a dual-arm system, including:

[0008] During the teaching process of the robot to perform the task of disassembling batteries, the robot's time-driven demonstration motion trajectory data in the task space is acquired, and a dual-arm collaborative execution strategy is constructed based on the temporal relationship of the demonstration motion trajectory data.

[0009] Based on the execution strategy, a hierarchical Gaussian process model is performed. Combined with the introduced adaptive weights, the high-level task planning and low-level trajectory generation in complex battery disassembly are analyzed to obtain trajectory parameters for dual-arm trajectory planning in complex task structures.

[0010] For each robotic arm, a Cartesian impedance control strategy is adopted. Based on trajectory parameter analysis, single-arm compliant control and dual-arm cooperative coupling control are obtained to obtain compliant control parameters and relative attitude stability parameters.

[0011] A layered uncertainty-driven adaptive stiffness adjustment strategy is introduced. Based on compliance control parameters and relative attitude stability parameters, the uncertainties of the upper and lower layers are analyzed by fusion. Adjustment control and constraint control are performed according to the total fusion uncertainty to obtain an adaptive control constraint strategy.

[0012] Based on the aforementioned control constraint strategy, adaptive control is performed in space and task state when the robot performs the task of battery disassembly, and the robot's pose and force are constrained.

[0013] Optionally, during the teaching process of guiding the robot to perform the task of disassembling the battery, time-driven demonstration motion trajectory data of the robot in the task space is acquired, and based on the temporal relationship of the demonstration motion trajectory data, a dual-arm collaborative execution strategy is constructed, including:

[0014] When the robot is guided to disassemble the battery, motion data is collected during each teaching task through sensors installed on the robot.

[0015] Preprocessing is performed based on the motion data, according to... Acquire time-driven demonstration motion trajectory data in the robot's task space;

[0016] Based on the demonstrated motion trajectory data, the temporal relationships are analyzed, and a dual-arm coordinated execution strategy is constructed based on these temporal relationships. ;

[0017] in, This represents the robot's left arm. This represents the robot's left arm. Indicates the first Second demonstration trajectory Location at any given moment Indicates the first Second demonstration trajectory The posture of the moment Indicates the first Second demonstration trajectory The speed of time, Indicates the first Second demonstration trajectory Force / torque at any moment This indicates the amount of data for each demonstrated trajectory after the robot's left and right arms are aligned. This represents the number of demonstrations.

[0018] Optionally, based on the execution strategy, a hierarchical Gaussian process model is performed. Combined with introduced adaptive weights, the high-level task planning and low-level trajectory generation in complex battery disassembly are analyzed to obtain trajectory parameters for dual-arm trajectory planning in complex task structures, including:

[0019] The battery disassembly task is divided into high-level task planning and low-level trajectory generation, and the battery disassembly task is serialized into a task group of sub-stages with dual-arm collaboration based on the high-level task planning.

[0020] Using the execution strategy as input data and logical basis, in high-level task planning, the spatiotemporal dependencies between different tasks in the task group are analyzed based on Gaussian process modeling to obtain the task trajectory parameters of the two-arm collaboration. In low-level trajectory generation, the local trajectory segments of the left and right arms corresponding to each task group are determined through Gaussian process modeling.

[0021] The priority weights assigned to each sub-stage in the task group are determined, and the current execution task stage is dynamically adjusted based on the task trajectory parameters, the local trajectory segments, and the priority weights to obtain the trajectory parameters used for dual-arm trajectory planning.

[0022] Optionally, using the execution strategy as input data and logical basis, in high-level task planning, the spatiotemporal dependencies between different tasks in the task group are analyzed based on Gaussian process modeling to obtain the task trajectory parameters for dual-arm collaboration. Furthermore, in low-level trajectory generation, the local trajectory segments of the left and right arms corresponding to each task group are determined through Gaussian process modeling, including:

[0023] In high-level mission planning, task groups are defined for sub-phases. ,according to By modeling the spatiotemporal dependencies between different tasks using Gaussian processes, the trajectory parameters for dual-arm cooperative tasks are obtained. ;

[0024] In low-level trajectory generation, according to Gaussian process modeling is used to determine each sub-objective. Local trajectory segments corresponding to the left and right arms ;

[0025] in, Indicates the current battery pack. Indicates the distribution trend of battery packs. This represents the covariance matrix defined by the kernel function. It is based on the mean function obtained by modeling the trajectory statistics using a Gaussian process. It is a kernel function in the form of spatiotemporal decomposition that generates covariance by measuring the smoothness and correlation of trajectories in spatial location and time dimension respectively.

[0026] Optionally, the priority weights assigned to each sub-stage in the task group are determined, and dynamically adjusted based on the task trajectory parameters, the local trajectory segments, and the priority weights according to the currently executing task stage detected in real time, to obtain trajectory parameters for dual-arm trajectory planning, including:

[0027] According to the task order corresponding to the sub-stage, Assign fixed priority weights to each sub-stage. ;

[0028] Based on the matching degree between the end pose of each robotic arm and the preset sub-task target position detected in real time, the current task stage is determined and designated as the current stage.

[0029] Based on the current stage, according to By analyzing the correlation between the current stage and each sub-stage, dynamic attention weights are obtained. ;

[0030] Attention weight By integrating the prediction results of each sub-model in Gaussian process modeling, based on Determine the final trajectory output mean , and, according to Determine the final trajectory output variance ;

[0031] Based on trajectory output mean and trajectory output variance Determine the trajectory parameters used for dual-arm trajectory planning;

[0032] in, The total number of tasks. Number the current task phase. To control the weight decay rate of tasks that are not in the current stage.

[0033] Optionally, for each robotic arm, a Cartesian impedance control strategy is adopted. Based on trajectory parameter analysis, single-arm compliant control and dual-arm cooperative coupling control are obtained to obtain compliant control parameters and relative attitude stability parameters, including:

[0034] Based on the trajectory parameters, the robot's body characteristic parameters, dynamic interaction parameters, and motion state parameters are determined, and preset cooperative constraint parameters are obtained.

[0035] Based on the body characteristic parameters and the dynamic interaction parameters, the stiffness and damping parameters are analyzed using a Cartesian impedance control strategy to determine the acceleration command of the robot end effector, which serves as the compliance control parameter.

[0036] Based on the motion state parameters and the cooperative constraint parameters, a coupling force control mechanism is introduced to analyze the relative attitude stability parameters that maintain the relative attitude of each robotic arm from deviating from the desired target.

[0037] The body characteristic parameters include the joint space inertia matrix and Jacobian matrix in the robot's inherent physical characteristic parameters. The dynamic interaction parameters are real-time parameters related to motion control and environmental interaction. The real-time parameters include the current target pose, actual pose, and external forces. The body characteristic parameters include the actual pose and actual velocity of each robotic arm. The cooperative constraint parameters include the desired relative pose, coupling stiffness, and coupling damping.

[0038] Optionally, based on the body characteristic parameters and the dynamic interaction parameters, a Cartesian impedance control strategy is used to analyze the stiffness and damping parameters to determine the acceleration command of the robot's end effector, which serves as the compliance control parameter, including:

[0039] Using the aforementioned body characteristic parameters and the aforementioned dynamic interaction parameters as inputs, a Cartesian impedance control strategy is employed, based on... Calculate the acceleration commands for the robot's end effector;

[0040] Specifically, based on the motion state parameters and the cooperative constraint parameters, a coupling force control mechanism is introduced to analyze the relative attitude stability parameters that maintain the relative attitude of each robotic arm from deviating from the desired target. This includes: using the motion state parameters and the cooperative constraint parameters as inputs, introducing a coupling force control mechanism, and based on... Calculate the relative attitude stability parameters;

[0041] For robot joint space inertial matrix And Jacobi matrix The calculated mission space inertia matrix, Here is the stiffness matrix. Here is the damping matrix. This represents the pose error between the current target position generated by the trajectory planning layer and the actual position of the current end effector. Indicates external force. For the preset desired relative pose, Indicates coupling stiffness, It is a coupling damping.

[0042] Optionally, a layered uncertainty-driven stiffness adaptive adjustment strategy is introduced, based on compliance control parameters and relative attitude stability parameters, to fuse the uncertainties of the upper and lower layers, including:

[0043] Based on compliant control parameters and relative attitude stability parameters, we analyze the high-level uncertainties when the robot encounters risk areas during task execution. and low-level uncertainty ;

[0044] Based on high-level uncertainty and low-level uncertainty The total uncertainty is obtained through dynamic weighted fusion. .

[0045] Optionally, adjustment and constraint control are performed based on the total uncertainty of the fusion to obtain an adaptive control constraint strategy, including:

[0046] With total uncertainty Based on, according to Calculate the adjustment and control stiffness ;

[0047] Based on adjusting control stiffness Introducing displacement saturation constraints The analysis focuses on the robot's end-effector displacement within a safe range, and introduces a maximum stiffness constraint. Analyze the stiffness matrix Stiffness values ​​in each principal direction;

[0048] Based on adjusting control stiffness Based on the end displacement and stiffness values, an adaptive control constraint strategy is determined.

[0049] in, For position The diagonal stiffness matrix at the position, This is the matrix representing the maximum stiffness value. For adjustment coefficients, The stiffness eigenvector matrix, The maximum safe speed allowed for the mission. This represents the stiffness values ​​of the stiffness matrix in each principal direction. This is the maximum safe contact force allowed by the robotic arm.

[0050] Secondly, this application provides a dual-arm collaborative dismantling planning and control system for retired power battery packs, including:

[0051] The demonstration motion trajectory analysis module is used to acquire time-driven demonstration motion trajectory data of the robot in the task space during the teaching process of the robot performing the task of disassembling batteries under guidance, and to construct a dual-arm collaborative execution strategy based on the temporal relationship of the demonstration motion trajectory data.

[0052] The Gaussian process modeling module is used to perform hierarchical Gaussian process modeling based on the execution strategy. Combined with the introduced adaptive weights, it analyzes the high-level task planning and low-level trajectory generation in complex battery disassembly to obtain trajectory parameters for dual-arm trajectory planning in complex task structures.

[0053] The control and adjustment module is used to control each robotic arm of the robot by adopting a Cartesian impedance control strategy. Based on trajectory parameter analysis, it analyzes the single-arm compliant control and the dual-arm cooperative coupling control to obtain compliant control parameters and relative attitude stability parameters.

[0054] The control constraint strategy generation module is used to introduce a stiffness adaptive adjustment strategy driven by hierarchical uncertainty. Based on compliance control parameters and relative attitude stability parameters, it integrates the analysis of uncertainties in the high and low layers, and performs adjustment control and constraint control according to the total fused uncertainty to obtain an adaptive control constraint strategy.

[0055] The control constraint module is used to perform adaptive control in space and task state when the robot performs the battery disassembly task based on the control constraint strategy, and to constrain the robot's pose and force.

[0056] In summary, this application's embodiments collect task trajectory data during operator-guided teaching of a robotic arm to complete a battery disassembly task, and construct a dual-arm collaborative execution strategy. Then, an adaptive weight allocation strategy based on task stages is introduced to construct a hierarchical Gaussian process model. The higher-level model analyzes the spatiotemporal dependencies between sub-objectives, while the lower-level model analyzes trajectory segments. Variable impedance control is employed, establishing a dual-arm coupled Cartesian impedance control strategy, achieving compliant pose control through stiffness and damping. Finally, during the dual-arm collaborative control process, the stiffness is dynamically adjusted by integrating uncertainties from both the higher and lower levels. Combined with pose and maximum force constraints, a control constraint strategy is used to limit the robot's motion speed and interaction force, significantly improving the robot's adaptability, operational safety, and collaborative efficiency in complex industrial scenarios such as battery disassembly. Thus, this application achieves dynamic stiffness adjustment to balance human-robot interaction flexibility and precision, significantly improving task adaptability, operational safety, and human-robot collaboration efficiency in complex industrial scenarios such as battery disassembly, and addressing the operational shortcomings of existing technologies for robots performing battery disassembly tasks. Attached Figure Description

[0057] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0058] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a flowchart illustrating a dual-arm collaborative dismantling planning and control method for a retired power battery pack, as provided in an embodiment of this application.

[0060] Figure 2 This is a flowchart illustrating a dual-arm collaborative dismantling planning and control method for a retired power battery pack, provided in one embodiment of this application.

[0061] Figure 3 This application provides an optional example of a collaborative disassembly planning and control method and system flowchart for a dual-arm robot;

[0062] Figure 4 This is a block diagram of a dual-arm collaborative dismantling planning and control system for retired power battery packs provided in an embodiment of this application. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0064] To facilitate understanding of the embodiments of this application, further explanations and descriptions will be provided below in conjunction with the accompanying drawings and specific embodiments. These embodiments do not constitute a limitation on the embodiments of this application.

[0065] Figure 1 This is a flowchart illustrating a dual-arm collaborative dismantling planning and control method for a retired power battery pack, provided as an embodiment of this application. Figure 1 As shown in the embodiments of this application, the dual-arm collaborative dismantling planning and control method for retired power battery packs may specifically include the following steps:

[0066] Step 110: During the teaching process of the robot performing the task of disassembling the battery, the robot acquires time-driven demonstration motion trajectory data in the task space, and constructs a dual-arm collaborative execution strategy based on the temporal relationship of the demonstration motion trajectory data.

[0067] For example, the task teaching process for disassembling a retired battery typically includes tasks such as: removing the top cover, removing the insulating cap, removing the copper sheet connector, removing the battery management system, and removing structural components.

[0068] In this embodiment, initially, a demonstration-based teaching method is used, where an operator physically guides the robot's robotic arms (in the case of a dual-arm robot, the robotic arms include a left arm and a right arm) to complete the battery disassembly task. While the robot's robotic arms are performing battery disassembly, sensors mounted on the arms collect demonstration trajectory data for each demonstration, including but not limited to: position (the three-dimensional position of the robotic arm's end effector), posture (the posture of the robotic arm's end effector), velocity (the velocity of a segment of the robotic arm), and force / torque (the force and torque received by the robotic arm's end effector). Then, the demonstration trajectory data is preprocessed, including frequency synchronization and outlier removal, to obtain demonstration motion trajectory data, thus acquiring the multiple demonstration motion trajectories within the robot's task space.

[0069] In practical implementation, the collected demonstration trajectory data can be understood as the data collected by the robot during its disassembly task. The data is related to the time of the event, and the subtasks in the decomposed task also have a certain order. Therefore, there is a temporal relationship between the demonstration trajectory data. The demonstration trajectory data can be aligned using the motion time warping algorithm to determine the temporal relationship between the data, and a dual-arm collaborative execution strategy can be constructed based on the temporal relationship.

[0070] Step 120: Perform hierarchical Gaussian process modeling based on the execution strategy, and combine the introduced adaptive weights to analyze the high-level task planning and low-level trajectory generation in complex battery disassembly, so as to obtain trajectory parameters for dual-arm trajectory planning in complex task structures.

[0071] In this embodiment, using the demonstrated motion trajectory as a reference, a hierarchical Gaussian model based on task-stage adaptive weights is introduced for modeling. Specifically, the complex battery disassembly task is divided into high-level task planning (i.e., the high-level task layer, which can be simply described as high-level) and low-level trajectory generation (i.e., the low-level trajectory layer, which can be simply described as low-level). Gaussian process modeling is used to build models for both the high-level and low-level layers. The execution strategy is used as input data to train different sub-models as sample data.

[0072] In the high-level model, Gaussian process modeling is used to determine the mission trajectory parameters for dual-arm coordination. In the low-level model, Gaussian process modeling is also used to determine the local trajectory segments of the left and right arms corresponding to different mission objectives.

[0073] To ensure that high-priority tasks dominate trajectory planning and avoid task overlap and conflict, this embodiment introduces dynamic weights for dynamic adjustment of weights and integrates the results to obtain trajectory parameters for dual-arm trajectory planning. These trajectory parameters can be used to implement a dual-arm trajectory planning method that integrates multiple stages and strategies for complex task structures.

[0074] Step 130: For each robotic arm of the robot, a Cartesian impedance control strategy is adopted. Based on trajectory parameter analysis, single-arm compliant control and dual-arm cooperative coupling control are used to obtain compliant control parameters and relative attitude stability parameters.

[0075] To achieve compliance and synchronization in complex tasks for dual-arm robots, a Cartesian impedance control framework with dual-arm coupling is employed for impedance control. This impedance control comprises two parts: single-arm compliance control and collaborative coupling control between the two arms. Specifically, using trajectory parameters as a reference, the classic Cartesian impedance control strategy is used to analyze the compliance control parameters in single-arm compliance control. In the collaborative coupling control of the two arms, a coupling force control mechanism is introduced to analyze the relative attitude stability parameters that ensure the left and right robotic arms maintain their relative poses without deviating from the desired target.

[0076] Step 140: Introduce a stiffness adaptive adjustment strategy driven by hierarchical uncertainty. Based on compliance control parameters and relative attitude stability parameters, fuse the uncertainties of the upper and lower layers, and perform adjustment control and constraint control according to the total fused uncertainty to obtain an adaptive control constraint strategy.

[0077] To ensure operational safety and flexibility during human-robot collaboration, a hierarchical uncertainty-driven adaptive stiffness adjustment strategy is introduced. This strategy determines the control stiffness by analyzing the uncertainties encountered by the robot during task execution. Furthermore, to ensure that displacement response remains within a safe range under low stiffness and to prevent excessive force due to high stiffness, this embodiment introduces constraint control, such as displacement saturation constraints and stiffness constraints. Through uncertainty analysis and constraint control, an adaptive control constraint strategy is output.

[0078] Step 150: Based on the control constraint strategy, adaptive control is performed in space and task state when the robot performs the battery disassembly task, and the robot's pose and force are constrained.

[0079] This embodiment, through the synergistic integration of the above multi-level methods, outputs a strategy for adaptive control. When the dual-arm robot performs complex tasks, it uses real-time collected robotic arm data to generate control constraint strategies, thereby adaptively controlling and constraining the robot's pose and forces. This enables precise trajectory generation, stable collaborative control, and safe human-machine interaction, providing a reliable solution for applications such as battery disassembly.

[0080] It is evident that existing technologies for robotic battery dismantling tasks suffer from operational limitations, failing to address the challenges of collaborative control and human-robot interaction safety in the flexible dismantling of retired power batteries by single / dual-arm robots. This embodiment proposes a dual-arm collaborative dismantling planning and control method for retired power battery packs. It innovatively combines interactive imitation learning, hierarchical Gaussian process modeling based on adaptive weights, and adaptive impedance control to achieve interactive hierarchical skill learning and flexible collaborative control for the planning and control of dual-arm collaborative dismantling of retired power battery packs. This method, through multi-stage task decomposition and dynamic focusing mechanisms, achieves flexible planning and safe execution of complex operations, solving the problems of low operational adaptability and insufficient accuracy of robots performing battery dismantling tasks in existing technologies.

[0081] It should be noted that the main tasks of the dual-arm collaborative battery disassembly task include, but are not limited to: manually removing screws, using flexible clamps on both arms to independently remove insulating caps and copper sheet connectors, and collaboratively removing the top cover, battery management system, and structural components.

[0082] Reference Figure 2 This illustration shows a flowchart of a dual-arm collaborative dismantling planning and control method for a retired power battery pack, according to an optional embodiment of this application. The method specifically includes the following steps:

[0083] Step 210: During the teaching process of the robot performing the task of disassembling the battery, the robot acquires time-driven demonstration motion trajectory data in the task space, and constructs a dual-arm collaborative execution strategy based on the temporal relationship of the demonstration motion trajectory data.

[0084] Optionally, during the guided teaching process of disassembling the battery, the robot acquires time-driven demonstration motion trajectory data in the task space, and constructs a dual-arm collaborative execution strategy based on the temporal relationship of the demonstration motion trajectory data. Specifically, this may include: collecting motion data for each task teaching session using sensors mounted on the robot during guided battery disassembly; preprocessing the motion data according to... The robot acquires time-driven demonstration motion trajectory data within its task space; based on this data, it analyzes temporal relationships and constructs a dual-arm collaborative execution strategy. ;in, This represents the robot's left arm. This represents the robot's left arm. Indicates the first Second demonstration trajectory Location at any given moment Indicates the first Second demonstration trajectory The posture of the moment Indicates the first Second demonstration trajectory The speed of time, Indicates the first Second demonstration trajectory Force / torque at any moment This indicates the amount of data for each demonstrated trajectory after the robot's left and right arms are aligned. This represents the number of demonstrations.

[0085] Reference Figure 3 As shown in this embodiment, Indicates the number of demonstrations for task teaching, such as That is, it was carried out In each demonstration operation, motion trajectory data of the left and right arms is collected to form demonstration motion trajectory data. The following is an explanation of the motion trajectory sets of the robot's left and right arms:

[0086] The set of motion trajectories representing the robot's left arm, i.e. , indicating the first In this demonstration, the left arm is in the time series arrive The motion parameters within. Among them, The amount of trajectory data used to demonstrate the word's trajectory in the left arm; Based on special orthogonal groups Description, used to characterize spatial attitude; It represents six dimensions, including linear velocity and angular velocity.

[0087] The set of motion trajectories representing the robot's right arm, i.e. , indicating the first In this demonstration, the right arm in the time series arrive The motion parameters within. The explanations of the remaining parameters can refer to the above explanation of the set of motion trajectories of the left arm.

[0088] In this embodiment, the key motion parameters of the left and right arms at different times during multiple demonstrations can be integrated using the above formula to construct a demonstration trajectory dataset of the dual-arm robot in the disassembly task. This provides the original data foundation for subsequent hierarchical Gaussian process modeling based on these trajectories and generation of the robot's autonomous motion trajectory.

[0089] In the specific implementation, a dual-arm collaborative execution strategy is constructed based on the aforementioned temporal relationship. The core of this approach is to analyze the temporal sequence and interaction logic of the two arms' movements during the teaching process to form a set of reusable collaborative operation rules. The temporal relationship refers to the correlation of time dimensions such as the sequence of actions, duration, and interaction nodes between the two arms when performing the disassembly task. For example: the temporal sequence of independent single-arm operation: such as the left arm removing the insulating cap first, and after a certain interval, the right arm removing the copper connector. The temporal sequence of collaborative two-arm operation: such as during the removal of the top cover, the left arm first grabs the left side of the top cover, and after a certain interval, the right arm grabs the right side, both lifting upwards simultaneously at the same speed. This embodiment connects the temporal sequence of the task stages to ensure the continuity of the disassembly process. Furthermore, based on the above temporal relationships, the constructed collaborative two-arm execution strategy is a dynamic rule base. It first constrains the temporal sequence of the task stages: clearly defining the order of each sub-task (such as removing the top cover, disassembling the battery management system, etc.); secondly, it defines the collaborative logic of the two-arm movements, such as the switching timing between single-arm operation and collaborative two-arm operation, the speed during interaction, and force / torque matching rules.

[0090] The execution strategy connects teaching data with autonomous operations, providing temporal constraints for hierarchical Gaussian process modeling, ensuring that high-level task planning conforms to the time logic of the disassembly process, and providing a collaborative benchmark for dual-arm coupling impedance control. This enables the robot to understand the time logic of actions and autonomously coordinate, adapting to the complex and ever-changing operational requirements of battery pack disassembly.

[0091] Step 220: Divide the battery disassembly task into high-level task planning and low-level trajectory generation, and serialize the battery disassembly task into a task group of sub-stages of dual-arm collaboration based on the high-level task planning.

[0092] Step 230: Using the execution strategy as input data and logical basis, in high-level task planning, the spatiotemporal dependencies between different tasks in the task group are analyzed based on Gaussian process modeling to obtain the task trajectory parameters of the two-arm collaboration. In low-level trajectory generation, the local trajectory segments of the left and right arms corresponding to each task group are determined through Gaussian process modeling.

[0093] Steps 220-230 are described uniformly as follows:

[0094] In related technologies, with the continuous upgrading of industrial automation needs, dual-arm robots are increasingly widely used in complex tasks (such as battery disassembly and precision assembly). However, existing technologies mostly rely on fixed strategies or single models, making it difficult to adapt to dynamic environments and multi-stage task requirements. Innovative solutions are urgently needed in the following three aspects: 1) Trajectory planning: Traditional dual-arm cooperative control uses pre-programmed trajectories or single Gaussian process modeling. When the paths of the two robotic arms overlap in a narrow space, traditional methods, lacking a dynamic priority allocation mechanism, may lead to trajectory conflicts or incorrect switching of target points, causing collisions or task interruptions. 2) Cooperative control: Traditional methods rely on fixed stiffness parameters or rigid trajectory tracking. When facing dynamic disturbances, they cannot adaptively adjust, easily leading to tool overload or component damage. 3) Interactive control: Traditional safety mechanisms use fixed thresholds, making it difficult to balance the flexibility and operational accuracy of human intervention.

[0095] To address the shortcomings of existing technologies in trajectory planning, this embodiment uses a hierarchical Gaussian process model based on adaptive weights for task stages to break down complex tasks into logical sub-goals, and sets priorities for each task according to the battery pack structure to avoid overlapping tasks between the two arms.

[0096] To address the complexity of dismantling retired power batteries, this embodiment introduces a hierarchical Gaussian model. This model breaks down the overall task into high-level tasks and low-level actions through hierarchical modeling. Combined with the uncertainty modeling capabilities of Gaussian processes, it achieves a precise mapping from demonstration and teaching to autonomous trajectory planning, adapting to dynamic scenarios such as different battery pack models and changes in component positions.

[0097] Specifically, this embodiment proposes a hierarchical structure of high-level task planning and low-level trajectory generation. The entire complex battery disassembly task is divided into high-level task planning (high-level task layer) and low-level trajectory generation. That is, the disassembly task is transformed into a sequence of sub-tasks that can be efficiently learned and dynamically adjusted, resulting in a quantifiable and adjustable trajectory model. By combining the uncertainty modeling capability of Gaussian processes, the taught trajectory is accurately mapped to the value and the model autonomously plans the trajectory to adapt to different battery pack models.

[0098] During task decomposition, the overall task is serialized into multiple sequentially executed sub-stages. Each sub-stage corresponds to a specific sub-task, forming a task group. The sub-tasks are executed in a fixed order.

[0099] In the trajectory modeling, the high-level task layer is first modeled, and the spatiotemporal dependencies of each sub-stage task are described by Gaussian processes. Based on the dual-arm task trajectory of the battery pack disassembly order contained in each sub-task in the task group, the spatiotemporal dependencies between different tasks are modeled and analyzed, such as the correlation between battery pack and time, and the task trajectory parameters of dual-arm collaboration are output.

[0100] Then, low-level trajectory modeling is performed. Low-level trajectory generation focuses on the specific action implementation of each sub-stage, generating precise local motion trajectories for the left and right arms (such as the motion path of the left arm when removing the insulating cap, the synchronous trajectory of both arms when collaboratively removing the top cover, etc.), and outputting local trajectory segments.

[0101] In an optional embodiment, using the execution strategy as input data and logical basis, in high-level task planning, the spatiotemporal dependencies between different tasks in the task group are analyzed based on Gaussian process modeling to obtain the task trajectory parameters for dual-arm collaboration. Furthermore, in low-level trajectory generation, the local trajectory segments of the left and right arms corresponding to each task group are determined through Gaussian process modeling. This may include: in high-level task planning, targeting the task groups of sub-stages... ,according to By modeling the spatiotemporal dependencies between different tasks using Gaussian processes, the trajectory parameters for dual-arm cooperative tasks are obtained. ;

[0102] In low-level trajectory generation, according to Gaussian process modeling is used to determine each sub-objective. Local trajectory segments corresponding to the left and right arms ;in, Indicates the current battery pack. Indicates the distribution trend of battery packs. This represents the covariance matrix defined by the kernel function. It is based on the mean function obtained by modeling the trajectory statistics using a Gaussian process. It is a kernel function in the form of spatiotemporal decomposition that generates covariance by measuring the smoothness and correlation of trajectories in spatial location and time dimension respectively.

[0103] The parameters in this embodiment are described in detail below:

[0104] In this embodiment, To indicate the disassembly order of battery packs of the same type, each Both include the dual-arm task trajectory of the battery pack disassembly sequence, at which point there are Where S represents the type of battery pack.

[0105] In high-level task modeling, execution strategies are used as sample inputs to build corresponding sub-models. Execution strategies are data with temporal relationships, and in high-level task planning, the serialization of the overall task, including the division of sub-stage task groups, can all be based on the temporal relationships in the execution strategies. The sub-models are used to analyze the battery pack distribution trend and the covariance matrix of the Gaussian process. . Indicates the first The individual stage is in the battery pack ,time The task status (such as the target location of subtasks, execution progress, etc.) is used. Mean function. Reflecting the The individual stage is in the battery pack ,time The overall distribution trend. Covariance matrix. Gaussian kernel function Defined as a measure of the correlation between subtask states under different battery packs and at different time points, the calculation formula is as follows: .in, Used to indicate two different battery packs and In terms of spatial similarity, such as the closer the battery pack structure, the better. The smaller the value, the larger the kernel function value, and the more similar the subtask states. Used to reflect two different points in time and The time correlation is such that the shorter the time interval, the larger the kernel function value, indicating that the subtask state changes smoothly over time. For spatial scale hyperparameters, This is a hyperparameter for the time scale, used to control spatial / temporal correlation.

[0106] In the lower-level trajectory layer, local trajectory segments This represents the specific movement trajectories of the left and right arms, with raw data derived from the demonstration trajectory data and timing information of the sub-task phases in the execution strategy. Local trajectory segments. Parameters including position, attitude, and velocity are mainly determined by the trajectory segment of the left arm. and the trajectory segment of the right arm Composition. Mean function The low-level covariance matrix reflects the typical trajectory model obtained statistically from the demonstration and teaching trajectory. Generated by the spacetime decomposition kernel function, the calculation formula is: This kernel function models spatial and temporal correlations separately to more accurately control the smoothness of the trajectory. Specifically, the spatial kernel function... The main metric measures the smoothness and correlation of trajectories in the spatial dimension under different battery packs. The calculation formula is as follows: , Hyperparameters for controlling the space are used to improve the trajectory's adaptability to different battery packs and location variations; time kernel function. The main metric measures the smoothness and correlation of trajectories over time under different battery packs. The calculation formula is as follows: , Hyperparameters for controlling timing to avoid motion jitter.

[0107] Step 240: Determine the priority weight assigned to each sub-stage in the task group, and dynamically adjust the priority weight based on the task trajectory parameters, the local trajectory segment, and the priority weight according to the currently executing task stage detected in real time, to obtain the trajectory parameters used for dual-arm trajectory planning.

[0108] To achieve a multi-stage, multi-strategy fusion method for dual-arm trajectory planning in complex task structures, this embodiment introduces dynamic weights based on Gaussian modeling to fuse the prediction results of various sub-models. First, according to the decomposed process flow, the task is divided into multiple sequentially executed sub-stages, and a fixed priority weight is assigned to each sub-stage. The allocation of fixed priority weights can decrease in task order, ensuring that high-priority tasks dominate trajectory planning and avoiding task overlap conflicts. Then, the currently executing task stage is detected in real time, based on the matching degree between the robot arm's end-effector pose and the preset sub-task target position. Based on the current stage, the weights of each sub-task are dynamically adjusted to obtain dynamic weights (or attention weights), which are used to fuse the outputs of each sub-model to obtain the final trajectory output mean and variance, serving as the trajectory parameters for dual-arm trajectory planning.

[0109] Therefore, this embodiment proposes a hierarchical Gaussian process model based on adaptive weights, which dynamically adjusts the task execution priority according to the current battery structure and task status, realizing the decoupling of task hierarchy and dynamic optimization of trajectory planning, and solving the trajectory conflict and path interference problems that occur when traditional methods face complex structures and task intersections.

[0110] In one optional embodiment, this embodiment determines the priority weight assigned to each sub-stage in the task group, and dynamically adjusts it based on the task trajectory parameters, the local trajectory segment, and the priority weight according to the currently executing task stage detected in real time, to obtain trajectory parameters for dual-arm trajectory planning. This may include: according to the task order corresponding to the sub-stage, based on... Assign fixed priority weights to each sub-stage. Based on the real-time detection of the matching degree between the end-effector poses of each robotic arm and the preset sub-task target positions, the current task stage is determined as the current stage; based on the current stage, according to By analyzing the correlation between the current stage and each sub-stage, dynamic attention weights are obtained. Attention weight By integrating the prediction results of each sub-model in Gaussian process modeling, based on Determine the final trajectory output mean , and, according to Determine the final trajectory output variance ; Output mean based on trajectory and trajectory output variance The trajectory parameters used for dual-arm trajectory planning are determined; among them, The total number of tasks. Number the current task phase. To control the weight decay rate of tasks that are not in the current stage.

[0111] The relevant formulas and parameters in this embodiment are explained in detail:

[0112] In this embodiment, the tasks are divided according to their execution order. Each sub-stage is pre-assigned a fixed priority weight. The allocation of fixed weights mainly considers the total number of tasks. and the current sub-stage , .

[0113] In this embodiment, during dynamic weight allocation, the pose of each robotic arm's end effector is first detected in real time. This is then combined with the preset sub-task target position (i.e., the target pose predefined for each sub-task stage; typically, the robotic arm's end effector should reach this target pose upon completion of the sub-task). The matching degree is determined by quantifying the pose error, thereby determining the current stage. Focusing on the current stage, adjust the weights of each subtask. Current Stage With adjacent stages Preserve the weights to obtain dynamic weights The weights for other stages can be set to 0. The final trajectory mean is generated by dynamically weighting and fusing the prediction results from the Gaussian process model of each sub-stage. and variance The final trajectory mean Through dynamic weights and the first Calculation of the predicted mean for each sub-stage; final variance. Through dynamic weights and the first Calculation of prediction variance for each sub-stage.

[0114] Therefore, this embodiment adopts a hierarchical model combined with dynamic weight fusion to achieve the following: ① Reduce task complexity: The hierarchical structure decomposes the complex battery disassembly task into independently modelable sub-problems, simplifying the learning difficulty; ② Utilize the uncertainty of Gaussian processes: The covariance matrix quantifies unknown factors in task execution (such as battery pack differences and positional deviations), providing a basis for subsequent dynamic adjustments; ③ Enhance adaptability: The spatiotemporal kernel function controls the sensitivity to different environments through hyperparameters, enabling the model to adapt to different battery models and dynamic environments; ④ Provide a collaborative foundation: The high-level model ensures that sub-stages are executed in the process sequence (avoiding task conflicts), while the low-level model generates specific trajectories for dual-arm coordination (ensuring synchronized actions), providing accurate references for subsequent dual-arm coupling control; ⑤ Achieve multi-stage, multi-strategy fusion trajectory planning for complex disassembly tasks, dynamically focusing on the current task, improving the real-time performance and accuracy of trajectory planning, and meeting the needs of different batteries and dynamic environments.

[0115] Step 250: For each robotic arm of the robot, a Cartesian impedance control strategy is adopted. Based on trajectory parameter analysis, single-arm compliant control and dual-arm cooperative coupling control are used to obtain compliant control parameters and relative attitude stability parameters.

[0116] Optionally, the above-mentioned Cartesian impedance control strategy for each robotic arm, based on trajectory parameter analysis of single-arm compliant control and dual-arm cooperative coupling control, to obtain compliant control parameters and relative attitude stability parameters, may include the following sub-steps:

[0117] Sub-step 2501: Based on the trajectory parameters, determine the robot's body characteristic parameters, dynamic interaction parameters, motion state parameters, and obtain preset cooperative constraint parameters.

[0118] The body characteristic parameters include the joint space inertia matrix and Jacobian matrix in the robot's inherent physical characteristic parameters. The dynamic interaction parameters are real-time parameters related to motion control and environmental interaction. The real-time parameters include the current target pose, actual pose, and external forces. The body characteristic parameters include the actual pose and actual velocity of each robotic arm. The cooperative constraint parameters include the desired relative pose, coupling stiffness, and coupling damping.

[0119] Sub-step 2502: Based on the body characteristic parameters and the dynamic interaction parameters, the stiffness and damping parameters are analyzed using a Cartesian impedance control strategy to determine the acceleration command of the robot end effector as the compliance control parameter.

[0120] Sub-step 2503: Based on the motion state parameters and the cooperative constraint parameters, a coupling force control mechanism is introduced to analyze the relative attitude stability parameters that maintain the relative attitude of each robotic arm from deviating from the desired target.

[0121] A unified description is provided for sub-steps 2501-2503:

[0122] To address the shortcomings of existing technologies in collaborative control and to achieve compliance and synchronization of dual-arm robots in complex tasks, this embodiment introduces adaptive impedance control technology. It adopts a Cartesian impedance control framework with dual-arm coupling and dynamically adjusts stiffness and damping parameters based on the cognitive uncertainty output by the hierarchical model.

[0123] In its implementation, the core of the Cartesian impedance control framework is to address two types of operational requirements in battery disassembly: independent single-arm operation and coordinated dual-arm operation. These two aspects serve as the targets of impedance control. Impedance control for independent single-arm operation primarily ensures operational compliance; impedance control for coordinated dual-arm operation primarily ensures motion synchronization and avoids relative pose deviations. Impedance control is implemented using relevant matrices / parameters as baseline inputs. By analyzing trajectory parameters fused based on dynamic weights, the relevant matrices and parameters required for impedance control are determined.

[0124] In single-arm compliant control, for each robotic arm, a classic Cartesian impedance control strategy is adopted. The spatial relationship matrix, stiffness matrix, and damping matrix, combined with pose error and external force, are used as the reference for impedance control. By simulating the system characteristics of spring-damped systems, the end effector of the robotic arm has a compliant response, thereby enabling flexible position adjustment.

[0125] In dual-arm coupling and cooperative control, a coupling force control mechanism is introduced for scenarios where dual arms perform tasks in coordination. The mechanism mainly uses the coupling force of the left and right arms, the desired relative pose, and the actual relative pose as inputs. Combined with coupling stiffness and coupling damping, a cooperative constraint is applied between the left and right arms to maintain their relative poses from deviating from the preset desired target and ensure synchronous movement.

[0126] In an optional embodiment, the above-mentioned analysis of stiffness and damping parameters based on the body characteristic parameters and the dynamic interaction parameters using a Cartesian impedance control strategy to determine the acceleration command of the robot end effector as a compliance control parameter includes: using the body characteristic parameters and the dynamic interaction parameters as inputs, employing a Cartesian impedance control strategy, according to... Calculate the acceleration commands for the robot's end effector; among which, For robot joint space inertial matrix And Jacobi matrix The calculated mission space inertia matrix, Here is the stiffness matrix. Here is the damping matrix. This represents the pose error between the current target position generated by the trajectory planning layer and the actual position of the current end effector. It represents an external force.

[0127] Reference Figure 3 As shown, in this embodiment, As the mission space inertia matrix, it mainly reflects the mass characteristics of the end effector in different directions (position + attitude) in mission space, and can be obtained through: The calculation shows that this formula converts the inertial characteristics of the joint space into a relationship in the end-effector task space, ensuring that control commands match the end-effector motion characteristics.

[0128] This represents the acceleration of the robotic arm's end effector in the task space (which can be calculated) and is used as the output of the control equation.

[0129] diagonal matrix It belongs to the stiffness matrix and mainly defines the response intensity of the end effector to pose errors in different directions, such as The larger the value, the more sensitive the end effector is to errors deviating from the target pose, allowing for faster adjustment back to the target position. In practical implementation, the calculation formula is: .in, It is a diagonal matrix used to define the stiffness of each degree of freedom; This is the rotation matrix corresponding to the target pose.

[0130] This is the damping matrix, simulating the characteristics of a damper used to suppress system oscillations. It is typically designed as critical damping to improve system stability. In practical implementation, its calculation formula is: .in, It is a basic diagonal matrix that is matched with the stiffness matrix.

[0131] The speed of the robotic arm's end effector within the task space. Damping force is used to suppress excessively fast motion and prevent the end effector from deviating from the target due to excessive inertia.

[0132] The pose error is the deviation between the actual pose of the current robotic arm end effector and the target pose. For example, the larger the error, the more... The stronger the recovery adjustment, the closer the actual pose will be to the target pose. In actual implementation, The calculation formula is: .in, The current target pose generated by the trajectory planning layer. This represents the current actual pose of the robotic arm's end effector.

[0133] In calculation Introducing external forces into the governing equations This enables the robotic arm to sense external forces and adjust its movement.

[0134] In an optional implementation, the above-mentioned analysis of the relative attitude stability parameters, based on the motion state parameters and the cooperative constraint parameters, using a coupling force control mechanism to maintain the relative attitude of each robotic arm from deviating from the desired target, may specifically include: using the motion state parameters and the cooperative constraint parameters as inputs, introducing a coupling force control mechanism, according to... Calculate the relative attitude stability parameters; For the preset desired relative pose, Indicates coupling stiffness, It is a coupling damping.

[0135] Reference Figure 3 In this embodiment, the coupling force of the two arms Left arm coupling force Coupling force with the right arm This creates a cross constraint, forcing the coordinated movement of both arms.

[0136] As the desired relative pose, it defines the relative position and orientation that the two arms should maintain during collaborative operation, such as the structural constraints in the collaborative operation of the two arms during the collaborative disassembly of the top cover. In actual implementation, Through the formula: , calculated to be. Among them, The target pose for the right arm. The target pose for the left arm.

[0137] Left arm actual position Compared with the actual position of the right arm ,pass Calculate the actual relative pose to reflect the deviation of the current relative position of the two arms.

[0138] Coupling Stiffness Used to determine the correction strength for pose deviations of the two arms, and to determine the degree to which the relative position between the two arms is maintained; high coupling stiffness can improve cooperative stability, such as... The larger the arm, the closer the arms are to the desired relative position.

[0139] Coupled Damping It is mainly used to reduce the vibration or impact caused by the difference in speed between the two arms.

[0140] Therefore, this embodiment proposes a dual-arm coupled Cartesian impedance control strategy. During the single-arm execution phase, it provides target-based variable impedance compliant control, while during the dual-arm coordination phase, a coupling force control mechanism maintains a stable relative attitude between the two arms, improving the accuracy and consistency of the dual-arm coordinated operation. This combination solves the problem of traditional rigid control being unable to adapt to dynamic environments and meets the dual requirements of precise operation and coordinated precision in power battery disassembly.

[0141] Step 260 introduces a layered uncertainty-driven stiffness adaptive adjustment strategy. Based on compliance control parameters and relative attitude stability parameters, the uncertainties of the upper and lower layers are analyzed by fusion, and adjustment control and constraint control are performed according to the total fusion uncertainty to obtain an adaptive control constraint strategy.

[0142] Step 270: Based on the control constraint strategy, adaptive control is performed in space and task state when the robot performs the battery disassembly task, and the robot's pose and force are constrained.

[0143] Steps 260-270 are described uniformly as follows:

[0144] To address the shortcomings of existing technologies in interactive control, this application designs a stiffness adjustment mechanism driven by cognitive uncertainty. This mechanism combines pose error and force constraints for dual protection, limiting the range of motion and the applied force through both position and force constraints. To ensure operational safety and flexibility during human-robot collaboration, a hierarchical uncertainty-driven adaptive stiffness adjustment strategy is introduced. When the robot encounters high-risk or uncertain areas during task execution, the system proactively reduces stiffness, allowing for manual intervention or automatic compliant obstacle avoidance.

[0145] In its implementation, this embodiment first addresses stiffness adjustment by using uncertainties at both high and low levels. The total uncertainty is calculated through dynamic weighted fusion, and this calculation serves as the benchmark for stiffness adjustment. Then, to ensure the system does not experience abrupt changes in motion or excessive force due to high stiffness during execution, this embodiment introduces displacement saturation constraints and stiffness constraints to limit the stiffness and velocity of the robot's dual arms. Through the synergistic fusion of these multi-level methods, accurate trajectory generation, stable collaborative control, and safe human-machine interaction for dual-arm robots in complex tasks are achieved, providing a reliable solution for applications such as industrial battery disassembly.

[0146] Therefore, this embodiment proposes a layered uncertainty-driven stiffness adjustment mechanism, which realizes the adaptive control of robot stiffness and damping parameters under spatial and task conditions, enables real-time intervention and trajectory correction by external forces, and combines a dual constraint mechanism of pose and force to limit the robot's interactive force and range of motion.

[0147] In an optional embodiment, the aforementioned hierarchical uncertainty-driven stiffness adaptive adjustment strategy, based on compliant control parameters and relative attitude stability parameters, fuses and analyzes uncertainties at both high and low levels. Specifically, it may include: analyzing high-level uncertainties when the robot encounters risk areas during task execution based on compliant control parameters and relative attitude stability parameters. and low-level uncertainty Based on high-level uncertainty and low-level uncertainty The total uncertainty is obtained through dynamic weighted fusion. .

[0148] Reference Figure 3 As shown in this embodiment, hierarchical uncertainty refers to the uncertainty generated at different levels from task planning to trajectory execution, which is dynamically weighted and fused into total uncertainty, serving as the basis for stiffness adjustment.

[0149] Among them, high-level uncertainty It primarily reflects fluctuations in task structure and sub-objective dependencies, such as structural differences between different batteries and temporal differences in sub-task switching. It can be quantified based on the covariance matrix of a high-level Gaussian process model. This reflects the overall dispersion of the task distribution.

[0150] Low-level uncertainty It primarily reflects fluctuations in specific operational trajectories, such as deviations between the real-time position and velocity of the robotic arm's end effector and the theoretical trajectory; differences in the smoothness of local trajectory segments during independent single-arm operation or dual-arm collaborative operation. It can be quantified based on the covariance matrix of a low-level Gaussian process model. It reflects the subtle fluctuations of local movements.

[0151] The weights in the foregoing embodiments (That is, the task-related weights generated by adaptive weighting) dynamically weight and fuse uncertainties from different layers, highlighting uncertainties relevant to the current operation. The weighted fusion formula can be: This is used to ensure that the total uncertainty is more in line with actual operational needs.

[0152] In an optional embodiment, the above-mentioned adjustment control and constraint control based on the fused total uncertainty to obtain an adaptive control constraint strategy may specifically include: using the total uncertainty... Based on, according to Calculate the adjustment and control stiffness Based on adjusting control stiffness Introducing displacement saturation constraints The analysis focuses on the robot's end-effector displacement within a safe range, and introduces a maximum stiffness constraint. Analyze the stiffness matrix Stiffness values ​​in each principal direction; based on adjusting and controlling stiffness Based on the end displacement and stiffness values, an adaptive control constraint strategy is determined; among which, For position The diagonal stiffness matrix at the position, This is the matrix representing the maximum stiffness value. For adjustment coefficients, For end displacement, The stiffness eigenvector matrix, The maximum safe speed allowed for the mission. This represents the stiffness values ​​of the stiffness matrix in each principal direction. This is the maximum safe contact force allowed by the robotic arm.

[0153] Reference Figure 3 In this embodiment, the total uncertainty is correlated with the stiffness using an exponential function to calculate the adjustment control stiffness. (usually indicating location) The real-time stiffness matrix at a given location is in diagonal form, with each degree of freedom adjusted independently, thus achieving adaptive adjustment where high uncertainty corresponds to low stiffness. The matrix representing the maximum stiffness value is used as an example. It can refer to the maximum rigidity of the robotic arm under environmental conditions; adjustment coefficient This can be used as a sensitivity to control the change of stiffness with uncertainty, such as The larger the value, the more sensitive the stiffness is to uncertainty; That is, location The total uncertainty at the location. For example, its adjustment logic could be: when the robotic arm enters a high uncertainty area (such as disassembling copper sheets in a confined space, or when the position of a component is unknown). Increase, exponential term Reduced, leading to Lowering the margin of the robotic arm—allowing for more flexible trajectory adjustments during manual intervention (such as a gentle push from the operator to correct the position), or automatically and smoothly avoiding obstacles before a collision to prevent hard contact damage to components. Conversely, in areas of low uncertainty (such as fixed trajectory segments in standardized operations). Small, stiffness close to Ensure operational precision (such as stable gripping of insulating caps).

[0154] To avoid the risks associated with stiffness adjustment, ensure that the displacement response remains within a safe range under low stiffness, prevent sudden changes in system motion during execution, and prevent excessive forces due to high stiffness, this embodiment introduces two types of constraints: displacement saturation constraints and maximum stiffness constraints. The following explains these two types of constraints:

[0155] At low stiffness, the end effector of a robotic arm is prone to large displacements due to external forces (such as environmental reaction forces), potentially exceeding safe limits. Displacement saturation constraints are used to limit the maximum displacement under low stiffness. Specifically, in the displacement saturation constraint equation... The stiffness eigenvector matrix represents the actual displacement of the robotic arm's end effector. It reflects the distribution of stiffness in different directions; The term represents the inverse square root of the stiffness matrix; the lower the stiffness, the larger this value, but it is limited by... Maximum safe speed This is used to prevent excessively rapid displacement changes. By introducing displacement saturation constraints, it is ensured that the displacement will not change abruptly or become excessively large under low stiffness, thus preventing operational instability.

[0156] At high stiffness, the end effector of the robotic arm may generate excessive forces due to component contact, potentially leading to component damage. Therefore, a maximum stiffness constraint is introduced to limit the maximum stiffness in each direction. The maximum stiffness constraint equation is as follows: This represents the stiffness values ​​of the stiffness matrix in each principal direction; This is the maximum safe contact force allowed by the robotic arm.

[0157] Step 280: Based on the control constraint strategy, adaptive control is performed in space and task state when the robot performs the battery disassembly task, and the robot's pose and force are constrained.

[0158] In summary, the embodiments of this application firstly involve the operator guiding both arms to complete the battery disassembly task teaching through physical interaction, performing data acquisition and preprocessing to obtain multiple time-driven single-arm task trajectories and collaborative dual-arm task trajectories in the robot's task space, and constructing a dual-arm collaborative execution strategy based on temporal relationships. Then, using the demonstrated motion trajectory as a reference, an adaptive weight allocation strategy based on task stages is introduced to construct a hierarchical Gaussian process model. According to the standard battery pack disassembly process, the task is divided into sequentially executed sub-stages. Simultaneously, combining the spatial relationship between the real-time end-effector position and the reference positions of each sub-task, the entire complex battery disassembly task is divided into high-level and low-level tasks. The high-level task layer models the temporal and spatial dependencies between sub-targets, while the low-level task layer designs an independent Gaussian process for each sub-target, generating detailed trajectory segments. Secondly, variable impedance control is adopted to establish a dual-arm coupled Cartesian impedance control strategy to achieve motion synchronization and compliant operation. When the robotic arm performs tasks independently, such as removing the insulating cap and copper sheet connectors, target-based variable impedance control is executed, and compliant control of the end-effector pose error is achieved through stiffness and damping. When performing tasks involving dual-arm collaboration, such as collaboratively removing the top cover, battery management system, and structural components, a coupling force control mechanism maintains relative pose constraints, ensuring the stability of collaborative operations during disassembly. Finally, during the actual collaborative control of the dual arms, a hierarchical uncertainty-driven stiffness adjustment mechanism is proposed, integrating high-level task uncertainties and low-level motion uncertainties. This mechanism dynamically adjusts stiffness, allowing the operator to correct the trajectory in real time using external forces. Combined with pose constraints and maximum force constraints, it limits the robot's movement speed and interaction forces, preventing tool overload or accidental collisions.

[0159] As can be seen, the embodiments of this application can achieve task-level and motion-level decoupling through a hierarchical Gaussian process model based on task-stage adaptive weights, ensure the accuracy of dual-arm collaboration through coupling impedance control, and balance the flexibility and accuracy of human-machine interaction through dynamic stiffness adjustment. In complex industrial scenarios such as battery disassembly, this significantly improves task adaptability, operational safety and human-machine collaboration efficiency, and solves the problems existing in the prior art when performing battery disassembly tasks with robots.

[0160] It should be noted that, for the sake of simplicity, the method embodiments are described as a series of actions. However, those skilled in the art should know that the embodiments of this application are not limited to the described order of actions, because according to the embodiments of this application, some steps may be performed in other orders or simultaneously.

[0161] like Figure 4 As shown in the figure, this application embodiment also provides a dual-arm collaborative dismantling planning and control system 400 for retired power battery packs, including:

[0162] The demonstration motion trajectory analysis module 410 is used to acquire time-driven demonstration motion trajectory data of the robot in the task space during the teaching process of the robot being guided to perform the task of disassembling the battery, and to construct a dual-arm collaborative execution strategy based on the temporal relationship of the demonstration motion trajectory data.

[0163] The Gaussian process modeling module 420 is used to perform hierarchical Gaussian process modeling based on the execution strategy, and combined with the introduced adaptive weights, analyze the high-level task planning and low-level trajectory generation in complex battery disassembly to obtain trajectory parameters for dual-arm trajectory planning in complex task structures.

[0164] The control and adjustment module 430 is used to control each robotic arm of the robot by adopting a Cartesian impedance control strategy and analyzing the single-arm compliant control and dual-arm cooperative coupling control based on trajectory parameters to obtain compliant control parameters and relative attitude stability parameters.

[0165] The control constraint strategy generation module 440 is used to introduce a stiffness adaptive adjustment strategy driven by hierarchical uncertainty. Based on compliance control parameters and relative attitude stability parameters, it integrates and analyzes the uncertainties of the high and low layers, and performs adjustment control and constraint control according to the total fused uncertainty to obtain an adaptive control constraint strategy.

[0166] The control constraint module 450 is used to perform adaptive control in space and task state when the robot performs the task of battery disassembly based on the control constraint strategy, and to constrain the robot's pose and force.

[0167] Optional Gaussian process modeling module, including:

[0168] The hierarchical planning submodule is used to divide the battery disassembly task into high-level task planning and low-level trajectory generation, and serialize the battery disassembly task into a task group of sub-stages for dual-arm collaboration based on the high-level task planning.

[0169] The hierarchical modeling submodule is used to take the execution strategy as input data and logical basis, and in high-level task planning, analyze the spatiotemporal dependencies between different tasks in the task group based on Gaussian process modeling to obtain the task trajectory parameters of the two-arm collaboration, and in low-level trajectory generation, determine the local trajectory segments of the left and right arms corresponding to each task group through Gaussian process modeling.

[0170] The dynamic weighted fusion submodule is used to determine the priority weight assigned to each sub-stage in the task group. Based on the currently executing task stage detected in real time, the module dynamically adjusts the priority weight to obtain trajectory parameters for dual-arm trajectory planning.

[0171] Optional, the control and adjustment module includes:

[0172] The multi-dimensional parameter acquisition submodule is used to determine the robot's body characteristic parameters, dynamic interaction parameters, and motion state parameters based on the trajectory parameters, and to acquire preset cooperative constraint parameters.

[0173] The first impedance control submodule is used to analyze stiffness and damping parameters based on the body characteristic parameters and the dynamic interaction parameters, and to determine the acceleration command of the robot end effector as a compliance control parameter by adopting a Cartesian impedance control strategy.

[0174] The second impedance control submodule is used to analyze the relative attitude stability parameters that maintain the relative attitude of each robotic arm from deviating from the desired target by introducing a coupling force control mechanism based on the motion state parameters and the cooperative constraint parameters.

[0175] The body characteristic parameters include the joint space inertia matrix and Jacobian matrix in the robot's inherent physical characteristic parameters. The dynamic interaction parameters are real-time parameters related to motion control and environmental interaction. The real-time parameters include the current target pose, actual pose, and external forces. The body characteristic parameters include the actual pose and actual velocity of each robotic arm. The cooperative constraint parameters include the desired relative pose, coupling stiffness, and coupling damping.

[0176] It should be noted that the dual-arm collaborative dismantling planning and control system for retired power battery packs provided in the embodiments of this application can execute the dual-arm collaborative dismantling planning and control method for retired power battery packs provided in any embodiment of this application, and has the corresponding functions and beneficial effects of the execution method.

[0177] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0178] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for planning and controlling the collaborative dismantling of retired power battery packs using a dual-arm system, characterized in that, include: During the teaching process of the robot to perform the task of disassembling batteries, the robot's time-driven demonstration motion trajectory data in the task space is acquired, and a dual-arm collaborative execution strategy is constructed based on the temporal relationship of the demonstration motion trajectory data. Based on the execution strategy, a hierarchical Gaussian process model is performed. Combined with the introduced adaptive weights, the high-level task planning and low-level trajectory generation in complex battery disassembly are analyzed to obtain trajectory parameters for dual-arm trajectory planning in complex task structures. For each robotic arm, a Cartesian impedance control strategy is adopted. Based on trajectory parameter analysis, single-arm compliant control and dual-arm cooperative coupling control are obtained to obtain compliant control parameters and relative attitude stability parameters. A layered uncertainty-driven stiffness adaptive adjustment strategy is introduced. Based on compliance control parameters and relative attitude stability parameters, the uncertainties of the upper and lower layers are analyzed by fusion. Adjustment control and constraint control are performed according to the total fusion uncertainty to obtain an adaptive control constraint strategy. Based on the aforementioned control constraint strategy, adaptive control is performed in space and task state when the robot performs the task of battery disassembly, and the robot's pose and force are constrained.

2. The method according to claim 1, characterized in that, During the guided teaching process of the robot to disassemble batteries, time-driven demonstration motion trajectory data of the robot in the task space is acquired. Based on the temporal relationship of the demonstration motion trajectory data, a dual-arm collaborative execution strategy is constructed, including: When the robot is guided to disassemble the battery, motion data is collected during each teaching task through sensors installed on the robot. Preprocessing is performed based on the motion data, according to... Acquire time-driven demonstration motion trajectory data in the robot's task space; Based on the demonstrated motion trajectory data, the temporal relationships are analyzed, and a dual-arm coordinated execution strategy is constructed based on these temporal relationships. ; in, This represents the robot's left arm. This represents the robot's left arm. Indicates the first Second demonstration trajectory Location at any given moment Indicates the first Second demonstration trajectory The posture of the moment Indicates the first Second demonstration trajectory The speed of time, Indicates the first Second demonstration trajectory Force / torque at any moment This indicates the amount of data for each demonstrated trajectory after the robot's left and right arms are aligned. This represents the number of demonstrations.

3. The method according to claim 1, characterized in that, Based on the aforementioned execution strategy, a hierarchical Gaussian process model is performed. Combined with introduced adaptive weights, the high-level task planning and low-level trajectory generation in complex battery disassembly are analyzed to obtain trajectory parameters for dual-arm trajectory planning in complex task structures, including: The battery disassembly task is divided into high-level task planning and low-level trajectory generation, and the battery disassembly task is serialized into a task group of sub-stages with dual-arm collaboration based on the high-level task planning. Using the execution strategy as input data and logical basis, in high-level task planning, the spatiotemporal dependencies between different tasks in the task group are analyzed based on Gaussian process modeling to obtain the task trajectory parameters of the two-arm collaboration. In low-level trajectory generation, the local trajectory segments of the left and right arms corresponding to each task group are determined through Gaussian process modeling. The priority weights assigned to each sub-stage in the task group are determined, and the current execution task stage is dynamically adjusted based on the task trajectory parameters, the local trajectory segments, and the priority weights to obtain the trajectory parameters used for dual-arm trajectory planning.

4. The method according to claim 3, characterized in that, Using the aforementioned execution strategy as input data and logical foundation, in high-level task planning, Gaussian process modeling is used to analyze the spatiotemporal dependencies between different tasks in the task group, obtaining the task trajectory parameters for dual-arm collaboration. Furthermore, in low-level trajectory generation, Gaussian process modeling is used to determine the local trajectory segments of the left and right arms corresponding to each task group, including: In high-level mission planning, task groups are defined for sub-phases. ,according to By modeling the spatiotemporal dependencies between different tasks using Gaussian processes, the trajectory parameters for dual-arm cooperative tasks are obtained. ; In low-level trajectory generation, according to Gaussian process modeling is used to determine each sub-objective. Local trajectory segments corresponding to the left and right arms ; in, Indicates the current battery pack. Indicates the distribution trend of battery packs. This represents the covariance matrix defined by the kernel function. It is based on the mean function obtained by modeling the trajectory statistics using a Gaussian process. It is a kernel function in the form of spatiotemporal decomposition that generates covariance by measuring the smoothness and correlation of trajectories in spatial location and time dimension respectively.

5. The method according to claim 3, characterized in that, The priority weights assigned to each sub-stage in the task group are determined, and based on the currently executing task stage detected in real time, the priority weights are dynamically adjusted to obtain trajectory parameters for dual-arm trajectory planning, including: According to the task order corresponding to the sub-stage, Assign fixed priority weights to each sub-stage. ; Based on the matching degree between the end pose of each robotic arm and the preset sub-task target position detected in real time, the current task stage is determined and designated as the current stage. Based on the current stage, according to By analyzing the correlation between the current stage and each sub-stage, dynamic attention weights are obtained. ; Attention weight By integrating the prediction results of each sub-model in Gaussian process modeling, based on Determine the final trajectory output mean , and, according to Determine the final trajectory output variance ; Based on trajectory output mean and trajectory output variance Determine the trajectory parameters used for dual-arm trajectory planning; in, The total number of tasks. Number the current task phase. To control the weight decay rate of tasks that are not in the current stage.

6. The method according to claim 1, characterized in that, For each robotic arm, a Cartesian impedance control strategy is adopted. Based on trajectory parameter analysis, single-arm compliant control and dual-arm cooperative coupling control are obtained, yielding compliant control parameters and relative attitude stability parameters, including: Based on the trajectory parameters, the robot's body characteristic parameters, dynamic interaction parameters, and motion state parameters are determined, and preset cooperative constraint parameters are obtained. Based on the body characteristic parameters and the dynamic interaction parameters, the stiffness and damping parameters are analyzed using a Cartesian impedance control strategy to determine the acceleration command of the robot end effector, which serves as the compliance control parameter. Based on the motion state parameters and the cooperative constraint parameters, a coupling force control mechanism is introduced to analyze the relative attitude stability parameters that maintain the relative attitude of each robotic arm from deviating from the desired target. The body characteristic parameters include the joint space inertia matrix and Jacobian matrix in the robot's inherent physical characteristic parameters. The dynamic interaction parameters are real-time parameters related to motion control and environmental interaction. The real-time parameters include the current target pose, actual pose, and external forces. The body characteristic parameters include the actual pose and actual velocity of each robotic arm. The cooperative constraint parameters include the desired relative pose, coupling stiffness, and coupling damping.

7. The method according to claim 6, characterized in that, Based on the aforementioned body characteristic parameters and dynamic interaction parameters, a Cartesian impedance control strategy is used to analyze stiffness and damping parameters, and determine the acceleration command of the robot's end effector as a compliance control parameter, including: Using the aforementioned body characteristic parameters and the aforementioned dynamic interaction parameters as inputs, a Cartesian impedance control strategy is employed, based on... Calculate the acceleration commands for the robot's end effector; Specifically, based on the motion state parameters and the cooperative constraint parameters, a coupling force control mechanism is introduced to analyze the relative attitude stability parameters that maintain the relative attitude of each robotic arm from deviating from the desired target. This includes: using the motion state parameters and the cooperative constraint parameters as inputs, introducing a coupling force control mechanism, and according to... Calculate the relative attitude stability parameters; For robot joint space inertial matrix And Jacobi matrix The calculated mission space inertia matrix, Here is the stiffness matrix. Here is the damping matrix. This represents the pose error between the current target position generated by the trajectory planning layer and the actual position of the current end effector. Indicates external force. For the preset desired relative pose, Indicates coupling stiffness. It is a coupling damping.

8. The method according to claim 1, characterized in that, A layered uncertainty-driven adaptive stiffness adjustment strategy is introduced, based on compliance control parameters and relative attitude stability parameters, to fuse the uncertainties of the upper and lower layers, including: Based on compliant control parameters and relative attitude stability parameters, we analyze the high-level uncertainties when the robot encounters risk areas during task execution. and low-level uncertainty ; Based on high-level uncertainty and low-level uncertainty The total uncertainty is obtained through dynamic weighted fusion. .

9. The method according to claim 8, characterized in that, Based on the total uncertainty of the fusion, adjustment control and constraint control are performed to obtain an adaptive control constraint strategy, including: With total uncertainty Based on, Calculate the adjustment and control stiffness ; Based on adjusting control stiffness Introducing displacement saturation constraints The analysis focuses on the robot's end-effector displacement within a safe range, and introduces a maximum stiffness constraint. Analyze the stiffness matrix Stiffness values ​​in each principal direction; Based on adjusting control stiffness Based on the end displacement and stiffness values, an adaptive control constraint strategy is determined. in, For position The diagonal stiffness matrix at the position, This is the matrix representing the maximum stiffness value. For adjustment coefficients, The stiffness eigenvector matrix, The maximum safe speed allowed for the mission. This represents the stiffness values ​​of the stiffness matrix in each principal direction. This is the maximum safe contact force allowed by the robotic arm.

10. A dual-arm collaborative dismantling planning and control system for retired power battery packs, characterized in that, include: The demonstration motion trajectory analysis module is used to acquire time-driven demonstration motion trajectory data of the robot in the task space during the teaching process of the robot performing the task of disassembling batteries under guidance, and to construct a dual-arm collaborative execution strategy based on the temporal relationship of the demonstration motion trajectory data. The Gaussian process modeling module is used to perform hierarchical Gaussian process modeling based on the execution strategy. Combined with the introduced adaptive weights, it analyzes the high-level task planning and low-level trajectory generation in complex battery disassembly to obtain trajectory parameters for dual-arm trajectory planning in complex task structures. The control and adjustment module is used to control each robotic arm of the robot by adopting a Cartesian impedance control strategy. Based on trajectory parameter analysis, it analyzes the single-arm compliant control and the dual-arm cooperative coupling control to obtain compliant control parameters and relative attitude stability parameters. The control constraint strategy generation module is used to introduce a stiffness adaptive adjustment strategy driven by hierarchical uncertainty. Based on compliance control parameters and relative attitude stability parameters, it integrates the analysis of uncertainties in the high and low layers, and performs adjustment control and constraint control according to the total fused uncertainty to obtain an adaptive control constraint strategy. The control constraint module is used to perform adaptive control in space and task state when the robot performs the battery disassembly task based on the control constraint strategy, and to constrain the robot's pose and force.

Images