Safety-constrained instruction revision and rollback control method for live-line work multi-agent collaboration

Abstract

The application relates to a safety constraint instruction correction and rollback control method for live-line work multi-agent cooperation, which can be used in the technical field of electric power. The method comprises the following steps: collecting state data of multiple agents, and performing timestamp alignment processing to obtain unified states; determining a minimum safety gap and combining personnel proximity risks corresponding to the unified states, mechanical limiting and cooperative interlocking relationships to generate a safety constraint set; obtaining candidate control instructions from at least two agents, and converting the candidate control instructions into unified candidate instructions; performing solving processing on the unified candidate instructions to obtain a correction instruction which meets the safety constraint set and has the minimum deviation from the unified candidate instructions; performing feasibility verification on the correction instruction, and performing consistency verification on cooperative interlocking or task stage compatibility among the multiple agents; and triggering hierarchical rollback control corresponding to the correction instruction in the case that the verification result indicates that the verification fails. The method can improve the efficiency of cooperative control.

Description

Technical Field

[0001] This application relates to the field of power technology, and in particular to a method for modifying and backing down safety constraint commands for multi-agent collaboration in live-line work. Background Technology

[0002] With the continuous advancement of smart grid construction, live-line working, as a key means to ensure the continuous operation of the power system, is gradually developing towards a collaborative working model.

[0003] Currently, multi-party coordination in live-line work is mainly based on the independent control systems of each piece of equipment and on-site manual command. However, when faced with complex and ever-changing working conditions, it is necessary to manually judge and coordinate each safety risk and equipment status, resulting in low efficiency of collaborative control. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for safety constraint instruction correction and rollback control for multi-agent collaboration in live-line work, which can improve the efficiency of collaborative control and address the aforementioned technical problems.

[0005] Firstly, this application provides a method for modifying and backing down safety constraint commands for multi-agent collaborative work on live lines. The method includes:

[0006] According to a preset sampling period, the status data of multiple intelligent agents participating in live-line work are collected, and the status data is timestamped to obtain the unified status of the multiple intelligent agents under a unified time reference; the status data includes at least personnel status data, robot status data, and environmental and electrical parameters.

[0007] The minimum safe clearance for live-line work is determined based on the voltage level and operating conditions. Combined with the personnel approach risk, mechanical limit, and collaborative interlocking relationships corresponding to the unified state, a set of safety constraints for live-line work is generated. This set of safety constraints is shared by the multiple intelligent agents within the same control cycle.

[0008] Acquire candidate control instructions from at least two of the agents, and convert the candidate control instructions into unified candidate instructions represented by a unified control vector;

[0009] Based on the set of security constraints, the unified candidate instruction is solved to obtain the corrected instruction that satisfies the set of security constraints and has the smallest deviation from the unified candidate instruction.

[0010] The feasibility of the correction instructions is verified, and the consistency of the cooperative interlocking or task phase compatibility among the multiple intelligent agents is verified to obtain the verification results.

[0011] If the verification result indicates that the verification failed, the hierarchical rollback control corresponding to the correction instruction is triggered.

[0012] In one embodiment, the step of generating the safety constraint set for live-line work by combining the personnel approach risk, mechanical limit, and collaborative interlocking relationships corresponding to the unified state includes:

[0013] Based on the unified state, the distance function information between the robot's joint configuration and the charged body is determined. According to the minimum safety clearance, the distance function information is linearized to the first order to obtain the electrical clearance constraint.

[0014] Based on the unified state, the shortest distance between the key parts of the personnel and the dangerous parts of the robot is determined, and the shortest distance is linearized to obtain the personnel approach constraint corresponding to the personnel approach risk.

[0015] Based on the aforementioned mechanical limits, determine the mechanical and dynamic constraints;

[0016] Based on the aforementioned collaborative interlocking relationship, determine the collaborative consistency and conflict resolution constraints;

[0017] The set of safety constraints is generated based on the electrical clearance constraints, the personnel access constraints, the mechanical and dynamic constraints, and the coordination and conflict resolution constraints.

[0018] In one embodiment, the step of solving the unified candidate instruction based on the set of security constraints to obtain the corrected instruction that satisfies the set of security constraints and has the smallest deviation from the unified candidate instruction includes:

[0019] The optimization objective is to minimize the deviation between the unified candidate instruction and the instruction to be solved, and the constraint condition is that the instruction to be solved satisfies the set of safety constraints. A quadratic optimization problem is constructed. The deviation is determined based on a preset weight matrix, which is used to characterize the correction cost and cooperative priority of different joints.

[0020] The correction instruction is obtained by solving the quadratic optimization problem.

[0021] In one embodiment, the feasibility verification of the correction instruction and the consistency verification of the cooperative interlocking or task phase compatibility among the multiple intelligent agents, to obtain the verification result, include:

[0022] The feasibility of the correction instruction is verified to obtain the feasibility verification result; the feasibility verification includes a review of the electrical clearance between live parts and the distance between personnel.

[0023] A consistency check is performed on the cooperative interlocks or task phase compatibility among the multiple intelligent agents to obtain a consistency check result; the consistency check includes compatibility checks and processing of task phase interlocks, tool occupancy interlocks, or synchronous action constraints.

[0024] The verification result is determined based on the feasibility verification result and the consistency verification result.

[0025] In one embodiment, triggering the hierarchical rollback control corresponding to the correction instruction when the verification result indicates that the verification failed includes:

[0026] If the verification result indicates that the verification fails, the risk level and degree of infeasibility are determined.

[0027] The retreat level is determined based on the risk level and the degree of infeasibility; the retreat level includes deceleration, maintaining, retreating to a safe position, and emergency braking;

[0028] The graded rollback control is triggered based on the rollback level; the recovery conditions corresponding to the graded rollback control include distance margin and confidence level continuously meeting thresholds, as well as personnel confirmation.

[0029] In one embodiment, the method further includes:

[0030] Within each control cycle, a structured information packet is generated and output; the structured information packet includes at least the unified candidate instruction, the correction instruction, the correction amount, the triggered constraint clause number, the constraint margin, the confidence level, the rollback level, and the rollback reason code.

[0031] In one embodiment, determining the minimum safe clearance for live-line work based on the voltage level and operating conditions includes:

[0032] Conservative factors are determined based on perceived confidence and communication confidence.

[0033] The minimum safety clearance is determined based on the voltage level, the operating conditions, and the conservative factor.

[0034] Secondly, this application also provides a safety constraint command correction and rollback control device for multi-agent collaborative live-line work. The device includes:

[0035] The data acquisition module is used to collect the status data of multiple intelligent agents participating in live-line work according to a preset sampling period, and to perform timestamp alignment processing on the status data to obtain the unified status of the multiple intelligent agents under a unified time reference; the status data includes at least personnel status data, robot status data, and environmental and electrical parameters.

[0036] The set generation module is used to determine the minimum safe clearance for the live-line operation based on the voltage level and working conditions, and to generate a set of safety constraints for the live-line operation by combining the personnel approach risk, mechanical limit and cooperative interlock relationship corresponding to the unified state; the set of safety constraints is shared by the multiple intelligent agents within the same control cycle;

[0037] The instruction acquisition module is used to acquire candidate control instructions from at least two of the intelligent agents and convert the candidate control instructions into unified candidate instructions represented by a unified control vector.

[0038] The instruction processing module is used to solve the unified candidate instruction based on the set of security constraints to obtain the corrected instruction that satisfies the set of security constraints and has the smallest deviation from the unified candidate instruction.

[0039] The instruction verification module is used to verify the feasibility of the correction instruction and to verify the consistency of the cooperative interlocking or task phase compatibility among the multiple intelligent agents, and to obtain the verification result.

[0040] The rollback control module is used to trigger the hierarchical rollback control corresponding to the correction instruction when the verification result indicates that the verification has failed.

[0041] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0042] According to a preset sampling period, the status data of multiple intelligent agents participating in live-line work are collected, and the status data is timestamped to obtain the unified status of the multiple intelligent agents under a unified time reference; the status data includes at least personnel status data, robot status data, and environmental and electrical parameters.

[0043] The minimum safe clearance for live-line work is determined based on the voltage level and operating conditions. Combined with the personnel approach risk, mechanical limit, and collaborative interlocking relationships corresponding to the unified state, a set of safety constraints for live-line work is generated. This set of safety constraints is shared by the multiple intelligent agents within the same control cycle.

[0044] Acquire candidate control instructions from at least two of the agents, and convert the candidate control instructions into unified candidate instructions represented by a unified control vector;

[0045] Based on the set of security constraints, the unified candidate instruction is solved to obtain the corrected instruction that satisfies the set of security constraints and has the smallest deviation from the unified candidate instruction.

[0046] The feasibility of the correction instructions is verified, and the consistency of the cooperative interlocking or task phase compatibility among the multiple intelligent agents is verified to obtain the verification results.

[0047] If the verification result indicates that the verification failed, the hierarchical rollback control corresponding to the correction instruction is triggered.

[0048] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:

[0049] According to a preset sampling period, the status data of multiple intelligent agents participating in live-line work are collected, and the status data is timestamped to obtain the unified status of the multiple intelligent agents under a unified time reference; the status data includes at least personnel status data, robot status data, and environmental and electrical parameters.

[0050] The minimum safe clearance for live-line work is determined based on the voltage level and operating conditions. Combined with the personnel approach risk, mechanical limit, and collaborative interlocking relationships corresponding to the unified state, a set of safety constraints for live-line work is generated. This set of safety constraints is shared by the multiple intelligent agents within the same control cycle.

[0051] Acquire candidate control instructions from at least two of the agents, and convert the candidate control instructions into unified candidate instructions represented by a unified control vector;

[0052] Based on the set of security constraints, the unified candidate instruction is solved to obtain the corrected instruction that satisfies the set of security constraints and has the smallest deviation from the unified candidate instruction.

[0053] The feasibility of the correction instructions is verified, and the consistency of the cooperative interlocking or task phase compatibility among the multiple intelligent agents is verified to obtain the verification results.

[0054] If the verification result indicates that the verification failed, the hierarchical rollback control corresponding to the correction instruction is triggered.

[0055] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0056] According to a preset sampling period, the status data of multiple intelligent agents participating in live-line work are collected, and the status data is timestamped to obtain the unified status of the multiple intelligent agents under a unified time reference; the status data includes at least personnel status data, robot status data, and environmental and electrical parameters.

[0057] The minimum safe clearance for live-line work is determined based on the voltage level and operating conditions. Combined with the personnel approach risk, mechanical limit, and collaborative interlocking relationships corresponding to the unified state, a set of safety constraints for live-line work is generated. This set of safety constraints is shared by the multiple intelligent agents within the same control cycle.

[0058] Acquire candidate control instructions from at least two of the agents, and convert the candidate control instructions into unified candidate instructions represented by a unified control vector;

[0059] Based on the set of security constraints, the unified candidate instruction is solved to obtain the corrected instruction that satisfies the set of security constraints and has the smallest deviation from the unified candidate instruction.

[0060] The feasibility of the correction instructions is verified, and the consistency of the cooperative interlocking or task phase compatibility among the multiple intelligent agents is verified to obtain the verification results.

[0061] If the verification result indicates that the verification failed, the hierarchical rollback control corresponding to the correction instruction is triggered.

[0062] The aforementioned method, device, computer equipment, computer-readable storage medium, and computer program product for safety constraint instruction correction and rollback control in live-line work multi-agent collaboration collect state data from multiple agents participating in live-line work according to a preset sampling period and perform timestamp alignment processing to obtain a unified state under a unified time reference, which helps to eliminate time asynchrony problems in multi-agent information interaction; by determining the minimum safety gap based on voltage level and working conditions, and combining personnel approach risk, mechanical limit, and collaborative interlocking relationships to generate a set of safety constraints that can be shared by multiple agents, it helps to unify the safety criteria for multi-agent collaboration; by converting candidate control instructions from at least two agents into unified candidate instructions, and solving for the correction instruction that satisfies the constraints and has the smallest deviation based on the set of safety constraints, it helps to maintain the original collaborative intent to the greatest extent while meeting safety requirements and reducing unnecessary downtime; through feasibility verification and consistency verification, and triggering hierarchical rollback control when verification fails, it helps to promptly intercept unexecutable or conflicting instructions and provide an anomaly handling mechanism, thereby improving the efficiency and continuity of collaborative control while ensuring the safety of live-line work. Attached Figure Description

[0063] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0064] Figure 1 This is a flowchart illustrating a method for modifying and rolling back safety constraint commands for multi-agent collaboration in live-line work, as shown in one embodiment.

[0065] Figure 2 This is a flowchart illustrating the steps for generating a set of security constraints in one embodiment;

[0066] Figure 3 This is a flowchart illustrating a safety constraint command correction and rollback control method for multi-agent collaboration in live-line work, as described in another embodiment.

[0067] Figure 4 This is a structural block diagram of a safety constraint command correction and rollback control device for multi-agent collaboration in live-line work, as shown in one embodiment.

[0068] Figure 5 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0069] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0070] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0071] In one exemplary embodiment, such as Figure 1As shown, a method for correcting and rolling back safety constraint commands for multi-agent collaboration in live-line work is provided. This embodiment illustrates the method by applying it to a terminal. It is understood that this method can also be applied to a server, or to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. The terminal can be, but is not limited to, various personal computers, laptops, smartphones, tablets, etc.; the server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. In this embodiment, the method includes the following steps:

[0072] Step S101: Collect the status data of multiple intelligent agents participating in the live-line operation according to the preset sampling period, and perform timestamp alignment processing on the status data to obtain the unified status of multiple intelligent agents under a unified time reference; the status data includes at least personnel status data, robot status data, and environmental and electrical parameters.

[0073] Step S102: Determine the minimum safe clearance for live-line work based on the voltage level and working conditions, and generate a set of safety constraints for live-line work by combining the personnel approach risk, mechanical limit and collaborative interlocking relationships corresponding to the unified state; the set of safety constraints can be shared by multiple intelligent agents within the same control cycle;

[0074] Step S103: Obtain candidate control instructions from at least two agents, and convert the candidate control instructions into unified candidate instructions represented by a unified control vector;

[0075] Step S104: Based on the set of security constraints, solve the unified candidate instructions to obtain the corrected instructions that satisfy the set of security constraints and have the smallest deviation from the unified candidate instructions.

[0076] Step S105: Perform feasibility verification on the correction instructions and consistency verification on the cooperative interlocking or task phase compatibility among multiple intelligent agents to obtain the verification results.

[0077] Step S106: If the verification result indicates that the verification failed, trigger the hierarchical rollback control corresponding to the correction instruction.

[0078] Live-line work can be the work performed during the maintenance and operation of a power system to repair equipment, handle defects, and replace components without interrupting power.

[0079] Among them, multiple agents can be a set of entities that participate in the collaborative control of live-line work and can generate control intentions or control commands.

[0080] Among them, state data can be data describing the physical state, motion state, or task state of multiple agents.

[0081] The timestamp can be an identifier used to mark the time when status data was collected.

[0082] In this context, a unified state can be a set of state data of multiple agents belonging to the same time base after being timestamped and aligned.

[0083] Personnel status data can be information describing the status of workers or supervisors, such as the three-dimensional position and velocity information of key parts of the human body.

[0084] Among them, robot state data can be information describing the state of the robot, such as joint position, joint velocity, end effector pose and end effector velocity.

[0085] Among them, environmental and electrical parameters can be parameters that describe the work site environment and the electrical properties of live conductors, such as voltage level, work method and tool status.

[0086] The voltage level can be the rated voltage level of the live conductor.

[0087] Among them, the working conditions can be the specific working method and the insulation status of the tools at the live working site.

[0088] The minimum safety clearance can be the minimum distance threshold that must be maintained between a live conductor and an accessible conductive part, person, or robot under a given voltage level and operating conditions.

[0089] Among them, the risk of personnel proximity can be the safety hazard caused by the close proximity of personnel to critical parts of the robot and dangerous parts of the robot.

[0090] Among them, mechanical limits can be mechanical and dynamic restrictions such as upper limits of robot joint speed, soft limits of joint position, or upper limits of end-effector speed.

[0091] Among them, the cooperative interlock relationship can be the occupancy relationship and interlock logic in multi-agent cooperative tasks, such as prohibiting another agent from initiating conflicting actions when a certain tool is occupied.

[0092] The safety constraint set can be a set of constraints that must be met within the current control cycle, including electrical clearance constraints, personnel access constraints, mechanical limit constraints, and cooperative interlocking constraints.

[0093] The control cycle can be a fixed time interval for the system to collect status data, calculate instructions, and issue commands.

[0094] Among them, candidate control instructions can be control quantities or action suggestions proposed by each agent within the current control cycle.

[0095] The unified control vector can be a mathematical vector form used to uniformly represent candidate control commands from different sources, such as a joint velocity vector.

[0096] Among them, the unified candidate instruction can be a candidate control instruction that has been converted and aggregated and is represented by a unified control vector.

[0097] Among them, the correction instruction can be a control instruction obtained by solving and calculating the unified candidate instruction based on the principle of minimum change within the set of safety constraints.

[0098] Task phase compatibility refers to whether the action intentions of multiple agents in the current task phase are compatible and do not conflict with each other.

[0099] The verification result can be a judgment of pass or fail obtained after feasibility verification and consistency verification.

[0100] Among them, graded backoff control can be a backoff control action of different intensity output by the system according to different risk levels when the verification fails or the risk increases suddenly, such as deceleration, holding, evacuation or emergency braking.

[0101] Optionally, the terminal collects state data from multiple agents participating in live-line work according to a preset sampling period, and performs timestamp alignment processing on the state data to obtain a unified state of multiple agents under a unified time reference. The state data includes at least personnel state data, robot state data, and environmental and electrical parameters. The terminal determines the minimum safe clearance for live-line work based on the voltage level and working conditions, and generates a set of safety constraints for live-line work by combining the personnel approach risk, mechanical limit, and cooperative interlocking relationships corresponding to the unified state. The set of safety constraints is shared by multiple agents within the same control cycle. The terminal obtains candidate control instructions from at least two agents and converts the candidate control instructions into unified candidate instructions represented by a unified control vector. Based on the set of safety constraints, the terminal solves the unified candidate instructions to obtain a correction instruction that satisfies the set of safety constraints and has the smallest deviation from the unified candidate instructions. The terminal performs feasibility verification on the correction instruction and consistency verification on the cooperative interlocking or task phase compatibility among multiple agents to obtain a verification result. If the verification result indicates that the verification fails, the terminal triggers the hierarchical backoff control corresponding to the correction instruction.

[0102] For example, the terminal collects state data from personnel, robots, and platform agents participating in live-line work via sensors and communication interfaces, using a preset fixed time interval as the sampling period. It then uses interpolation algorithms to timestamp-align the state data, eliminating time differences caused by data transmission delays and obtaining a unified state for multiple agents under a unified time reference. The state data includes at least personnel state data (positions of key body parts), robot state data (joint positions and speeds), and environmental and electrical parameters including voltage levels. Based on the voltage level and specific working conditions, the terminal determines the minimum safe clearance for live-line work by querying a preset mapping table. Combining this with the personnel approach risk, mechanical limits, and interlocking relationships corresponding to the unified state, it constructs a set of safety constraints for live-line work. This set of safety constraints is shared by multiple agents within the same control cycle to ensure global safety criteria. Consistency: The terminal acquires candidate control commands from at least two agents and converts the different forms of candidate control commands into unified candidate commands represented by a unified control vector for unified computation. Based on a set of safety constraints, the terminal uses an optimization algorithm to solve the unified candidate commands, obtaining a corrected command that satisfies the set of safety constraints and has the smallest deviation from the unified candidate commands, thereby maintaining the original collaborative operation intent as much as possible while ensuring safety. The terminal performs feasibility verification on the corrected commands through geometric verification and other methods, and performs consistency verification on the collaborative interlocking or task phase compatibility between multiple agents, obtaining a verification result indicating whether the command is safe and executable. If the verification result indicates that the verification fails, the terminal triggers graded backoff control corresponding to the corrected command, including different levels such as deceleration, holding, withdrawal, or braking, according to the current risk level, to ensure the overall safety of the live-line working system.

[0103] In the aforementioned method for correcting and rolling back safety constraint instructions for multi-agent collaboration in live-line work, the state data of multiple agents participating in the live-line work are collected according to a preset sampling period and timestamped to obtain a unified state under a unified time reference, which helps to eliminate the time asynchrony problem in multi-agent information interaction. By determining the minimum safety gap based on the voltage level and working conditions, and combining personnel approach risk, mechanical limit, and collaborative interlocking relationships to generate a set of safety constraints that can be shared by multiple agents, it helps to unify the safety criteria for multi-agent collaboration. By converting candidate control instructions from at least two agents into unified candidate instructions, and solving for the corrected instructions that satisfy the constraints and have the smallest deviation based on the set of safety constraints, it helps to maintain the original collaborative intent to the greatest extent while meeting safety requirements and reducing unnecessary downtime. Through feasibility verification and consistency verification, and triggering hierarchical rollback control when verification fails, it helps to intercept unexecutable or conflicting instructions in a timely manner and provides an anomaly handling mechanism, thereby improving the efficiency and continuity of collaborative control while ensuring the safety of live-line work.

[0104] In one exemplary embodiment, reference is made to Figure 2 By combining the personnel approach risk, mechanical limit, and collaborative interlocking relationships corresponding to a unified state, a set of safety constraints for live-line work is generated, including:

[0105] Step S201: Based on the unified state, determine the distance function information between the robot's joint configuration and the charged body. According to the minimum safety clearance, perform first-order linearization on the distance function information to obtain the electrical clearance constraint.

[0106] Step S202: Based on the unified state, determine the shortest distance between the key parts of the personnel and the dangerous parts of the robot, and perform first-order linearization on the shortest distance to obtain the personnel approach constraints corresponding to the personnel approach risk.

[0107] Step S203: Determine the mechanical and dynamic constraints based on the mechanical limits;

[0108] Step S204: Determine the coordination consistency and conflict resolution constraints based on the coordination and interlocking relationship;

[0109] Step S205: Generate a set of safety constraints based on electrical clearance constraints, personnel access constraints, mechanical and dynamic constraints, and coordination and conflict resolution constraints.

[0110] The distance function information can be a function that describes the closest distance between the robot and a charged body or equivalent risk source under a preset joint configuration.

[0111] The first-order linearization process can be a linearization approximation operation, such as performing a first-order expansion on the distance function at the current time, to form a computable linear constraint.

[0112] Among them, the electrical clearance constraint can be a linear inequality constraint used to ensure that the distance between the robot and the charged body in the next control cycle is not less than the conservative minimum safety clearance.

[0113] Dangerous parts can be parts of the robot that can come into contact with people, such as the end effector and the surface of the linkage.

[0114] Among them, the closest distance can be the shortest straight-line distance between the critical parts of the personnel and the dangerous parts of the robot.

[0115] Among them, the personnel approach constraint can be a linear inequality constraint used to ensure that the robot's movements converge or decelerate in a specified direction when personnel approach.

[0116] Among them, mechanical and dynamic constraints can be physical and kinematic limitations on the robot, including upper limits on joint velocities, soft limits on joint positions, and upper limits on end-effector velocities.

[0117] Among them, the constraints on coordination consistency and conflict resolution can be equality or inequality constraints that constrain the occupancy and interlocking relationships in multi-agent collaborative tasks, such as restrictions on synchronized actions or interlocking actions.

[0118] Optionally, based on a unified state, the terminal determines the distance function information between the robot's joint configuration and the charged body. Based on the minimum safety clearance, it performs first-order linearization on the distance function information to obtain electrical clearance constraints. Based on the unified state, it determines the shortest distance between key personnel parts and dangerous parts of the robot, and performs first-order linearization on the shortest distance to obtain personnel access constraints corresponding to personnel access risks. Based on mechanical limits, it determines mechanical and dynamic constraints including upper limits for joint speeds and end-effector speeds. Based on cooperative interlocking relationships, it determines cooperative consistency and conflict resolution constraints to restrict synchronous or interlocked actions. Based on electrical clearance constraints, personnel access constraints, mechanical and dynamic constraints, and cooperative consistency and conflict resolution constraints, it generates a set of safety constraints.

[0119] The technical solution provided in this embodiment generates a set of safety constraints based on electrical clearance constraints, personnel access constraints, mechanical and dynamic constraints, and coordination consistency and conflict resolution constraints. This is beneficial for comprehensively covering various safety risks and coordination conflicts in live-line work, thereby improving the safety of multi-agent collaborative control.

[0120] In an exemplary embodiment, based on a set of safety constraints, a unified candidate instruction is solved to obtain a corrected instruction that satisfies the set of safety constraints and has the smallest deviation from the unified candidate instruction. This includes: constructing a quadratic optimization problem with the goal of minimizing the deviation between the unified candidate instruction and the instruction to be solved, and with the instruction to be solved satisfying the set of safety constraints as the constraint condition; wherein the deviation is determined based on a preset weight matrix, which is used to characterize the correction cost and collaborative priority of different joints; and solving the quadratic optimization problem to obtain the corrected instruction.

[0121] The instruction to be solved can be a control vector that satisfies the set of safety constraints and needs to be solved in the optimization problem.

[0122] The deviation can be the degree of difference between the unified candidate instruction and the instruction to be solved in the vector space.

[0123] Among them, the quadratic optimization problem can be a mathematical optimization problem in which the objective function is a quadratic function and the constraints are linear constraints.

[0124] The weight matrix can be a diagonal matrix used to adjust the weights of different joints when calculating deviations, reflecting the correction costs and coordination priorities of different joints.

[0125] Optionally, the terminal constructs a quadratic optimization problem with the goal of minimizing the deviation between the unified candidate instruction and the instruction to be solved, and with the instruction to be solved satisfying the set of safety constraints as the constraint condition. The deviation is determined based on a preset weight matrix, which is used to characterize the correction cost and coordination priority of different joints. A standard quadratic programming solver is used to solve the quadratic optimization problem, and the optimal solution is calculated in real time in each control cycle to obtain the correction instruction.

[0126] The technical solution provided in this embodiment obtains the correction instructions by solving the quadratic optimization problem, which is beneficial to the smooth correction of multi-agent instructions, thereby improving the continuity and execution efficiency of collaborative operations.

[0127] In an exemplary embodiment, the feasibility of the correction instruction and the consistency of the collaborative interlocking or task phase compatibility among multiple intelligent agents are verified to obtain a verification result. This includes: performing a feasibility verification on the correction instruction to obtain a feasibility verification result; the feasibility verification includes reviewing the electrical clearance between energized bodies and the distance between personnel; performing a consistency verification on the collaborative interlocking or task phase compatibility among multiple intelligent agents to obtain a consistency verification result; the consistency verification includes compatibility checks on task phase interlocking, tool occupancy interlocking, or synchronous action constraints; and determining the verification result based on the feasibility verification result and the consistency verification result.

[0128] Among them, the feasibility verification result can be a judgment on whether the instruction obtained after performing feasibility checks such as geometric verification on the modified instruction meets the physical security boundary.

[0129] The verification process can involve substituting the corrected state back into the distance function for geometric calculations to confirm whether the electrical clearance and personnel spacing truly meet the requirements.

[0130] Among them, the consistency verification result can be a judgment on whether there is a logical conflict in the control instructions obtained after performing a compatibility check on the control instructions of multiple agents.

[0131] Among them, the compatibility check process can be a process of checking whether the instructions issued by different intelligent agents are compatible with each other in terms of task phase interlocking, tool occupancy interlocking, or synchronous action constraints.

[0132] Optionally, the terminal performs a feasibility check on the correction command by substituting the predicted state corresponding to the correction command back into the distance function for verification, confirming whether the electrical clearance of the energized body and the distance between personnel meet the safety requirements, and obtaining a feasibility check result; it also performs a consistency check on the cooperative interlocking or task phase compatibility among multiple agents by performing compatibility checks on task phase interlocking, tool occupancy interlocking, or synchronous action constraints, determining whether there are conflicts in the commands of multiple agents, and obtaining a consistency check result; based on the feasibility check result and the consistency check result, it comprehensively determines whether the correction command can be safely issued, and determines the check result.

[0133] The technical solution provided in this embodiment obtains a feasibility verification result by reviewing the electrical clearance between charged bodies and the distance between personnel, which helps to avoid violating safety constraints. It also obtains a consistency verification result by performing a compatibility check, which helps to prevent locally feasible instructions from causing conflicts in global coordination, thereby improving the security and executability of issued instructions.

[0134] In an exemplary embodiment, if the verification result indicates that the verification failed, a graded rollback control corresponding to the correction instruction is triggered, including: determining the risk level and infeasibility degree if the verification result indicates that the verification failed; determining the rollback level based on the risk level and infeasibility degree; the rollback level includes deceleration, maintaining, evacuating to a safe posture, and emergency braking; triggering the graded rollback control based on the rollback level; the recovery conditions corresponding to the graded rollback control include distance margin and confidence level continuously meeting thresholds, as well as personnel confirmation.

[0135] The risk level can be a level that describes the severity of security threats in the current work scenario.

[0136] The degree of infeasibility can describe the severity of the violation of security constraints by the modification instruction or the state where the secondary optimization problem has no solution.

[0137] The retreat level can be an emergency response level classified according to the risk level and the degree of infeasibility, including deceleration, maintaining, evacuation to a safe position and emergency braking.

[0138] Among them, the recovery conditions can be the prerequisites that the system needs to meet to smoothly transition from the hierarchical back-off control state back to the normal cooperative control state.

[0139] The distance margin can be the difference between the actual distance and the minimum safety gap or safety distance threshold.

[0140] Optionally, if the verification result indicates that the verification failed, the terminal evaluates the current system status data to determine the risk level and infeasibility. Based on the risk level and infeasibility, it determines the corresponding rollback level. Rollback levels include deceleration, maintaining, retreating to a safe posture, and emergency braking. Based on the rollback level, it triggers graded rollback control and outputs rollback control actions of corresponding strength. After triggering graded rollback control, it continuously monitors the system status. When the distance margin and confidence level continuously meet the thresholds and the recovery conditions are confirmed by personnel, the system (power system / live-line working system) smoothly recovers from the rollback state to the normal collaborative control state.

[0141] The technical solution provided in this embodiment, by determining the rollback level, facilitates the adoption of matching emergency response measures for different levels of security risks. By triggering graded rollback control according to the rollback level and setting recovery conditions, it is beneficial to form a closed-loop control mechanism for abnormal handling and recovery, thereby improving emergency response efficiency and enhancing risk controllability.

[0142] In an exemplary embodiment, the method further includes: generating and outputting a structured information package in each control cycle; the structured information package includes at least a unified candidate instruction, a correction instruction, a correction amount, the triggered constraint clause number, the constraint margin, the confidence level, the rollback level, and the rollback reason code.

[0143] The structured information package can be a data set containing key states and decision information within the current control cycle of the system, encapsulated according to a preset data structure.

[0144] The correction amount can be the difference between the correction instruction and the unified candidate instruction.

[0145] The triggered constraint clause number can be a unique identifier for a specific security constraint rule that is in effect or violated within the current control cycle.

[0146] Among them, the constraint margin can be the remaining margin between the current state and the boundary of the safety constraint that is triggered.

[0147] The rollback reason code can be a code used to identify the specific reason for triggering the hierarchical rollback control.

[0148] Optionally, the terminal collects various data during system operation in each control cycle, generates and outputs a structured information package; the structured information package includes at least a unified candidate instruction, a correction instruction, a correction amount, the triggered constraint clause number, constraint margin, confidence level, rollback level, and rollback reason code; the structured information package is published to a visualization interface or log system through a communication network for real-time display of security status and recording of audit logs.

[0149] The technical solution provided in this embodiment generates and outputs a structured information package containing data such as correction amounts and triggered constraint clause numbers within each control cycle. This helps to provide on-site operators with a clear basis for safety boundary changes and an explanation of handling strategies, thereby enhancing the interpretability of the system and the traceability of project acceptance.

[0150] In one exemplary embodiment, determining the minimum safe clearance for live-line work based on voltage level and working conditions includes: determining a conservatism factor based on perception confidence and communication confidence; and determining the minimum safe clearance based on voltage level, working conditions, and the conservatism factor.

[0151] Among them, perception confidence can be a scalar indicator that describes the stability and reliability of the detection results of human body perception devices.

[0152] Among them, communication confidence can be a scalar indicator that describes the quality of communication links and the reliability of data transmission between multiple agents.

[0153] The conservative factor can be a coefficient calculated based on perceived confidence and communication confidence, used to adaptively adjust the size of the safety margin.

[0154] Optionally, the terminal acquires detection data from the human body sensing device and quality data from the communication link, determines a conservative factor based on the sensing confidence and communication confidence, queries the minimum distance threshold of the foundation based on the voltage level and working conditions, and modulates the minimum distance threshold of the foundation using the conservative factor to determine the minimum safe clearance for live-line work, and automatically expands the safety margin when the confidence decreases.

[0155] The technical solution provided in this embodiment determines a conservative factor based on perception confidence and communication confidence, which helps to quantify the uncertainty in the perception and communication process. By combining the conservative factor to determine the minimum safety gap, it is beneficial to adaptively expand the safety distance under adverse conditions such as obstruction, reflection, or weak network, thereby improving the robustness of the system under uncertain conditions and reducing the risk of misjudgment and missed judgment.

[0156] The following example illustrates the safety constraint command correction and rollback control method for multi-agent cooperative live-line working provided in this application. (Refer to...) Figure 3 This embodiment uses the application of this method to a terminal as an example for illustration.

[0157] Live-line working is a crucial method for power system maintenance and operation. It allows for equipment repair, defect handling, and component replacement without power outages, significantly reducing power loss and improving power supply reliability. With the increasing complexity of distribution network structures, the rising frequency of operations, and stringent safety management requirements, live-line working is gradually evolving from the traditional "manual labor + insulated tools" model to a more "human-machine collaborative, equipment-based, and intelligent" approach. In recent years, equipment such as live-line working robots, insulated working platforms, intelligent tools, and on-site video monitoring systems have been applied. Some operational steps are now completed by robots or with robot assistance, reducing the time and frequency of personnel directly approaching live conductors and improving operational consistency and controllability.

[0158] Live-line working robot systems typically consist of a mobile chassis or work platform, a robotic arm actuator, an end effector, an insulating protective structure, a sensing system, and a control system. To meet on-site operational needs, the system must simultaneously handle tasks such as environmental modeling, object identification, trajectory planning, force-position hybrid control, safety protection, and human-machine interaction. In typical applications, operators are responsible for process confirmation, tool replacement, on-site command, and emergency response, while robots perform repetitive or high-risk actions such as bolt tightening, wire stripping and crimping, lead wire installation, and foreign object removal. The two collaborate through voice, gestures, control terminals, or platform systems to complete the work process. Meanwhile, with the implementation of "platform-based management and control" and "edge computing" in power operation and maintenance scenarios, multiple entities with decision-making and control capabilities, such as dispatching platforms, edge nodes, video gateways, and sensor terminals, often exist on-site, forming a multi-agent collaborative operation model involving "humans, robots, platforms / edge nodes, and sensor terminals."

[0159] In the aforementioned multi-agent collaborative live-line work, safety control is the core constraint. Traditional safety measures mainly rely on minimum safe distances corresponding to voltage levels, selection specifications for insulated tools, equipotential / insulation work procedures, and on-site monitoring systems. Related robot systems typically employ basic protection methods such as geometric obstacle avoidance, speed limits, no-entry zones in the workspace, collision detection, and emergency stops. In more complex systems, vision- or point cloud-based personnel detection and area isolation are introduced, establishing independent safety boundaries for personnel and robots, triggering deceleration or shutdown when personnel approach or robots cross these boundaries. For platform-side control systems, risk alerts are often implemented through rules, strategies, or alarm thresholds. For example, alarms are triggered and manual intervention is requested when personnel enter a hazardous area, robots approach live conductors, or communication links malfunction.

[0160] However, from the perspective of related technical implementation methods, the control and safety constraints of multi-agent collaborative live-line work often exhibit "distributed" or "loosely coupled" characteristics: different agents generate control commands or action suggestions separately, which are then converged through simple prioritization, interlocking logic, or manual confirmation. The trajectory planning and obstacle avoidance modules on the robot controller side are usually based on their own perception results and local constraints; the safety strategies on the platform or edge node side are mainly based on alarms, speed limits, or shutdowns; and the operator's intentions and changes in actions are indirectly reflected through manual commands or operating terminals. Because the modeling granularity, input information, and triggering conditions of safety constraints are not uniform among the agents, situations can easily arise during the collaborative process where "different agents give different safety judgments at the same time," leading to command conflicts, execution delays, or excessive conservatism.

[0161] Furthermore, live-line working scenarios are characterized by both strong constraints and significant uncertainties: on the one hand, voltage levels and insulation conditions dictate rigid minimum safety clearances and operational rules; on the other hand, the complex on-site environment, common obstructions and reflections, rapid changes in personnel posture and intent, and potential delays and packet loss in communication links all contribute to uncertainties in the synchronization of perception and information. Related technologies often address these uncertainties by increasing safety distances, reducing speed, or directly shutting down the system. While these strategies improve safety margins, they can easily lead to frequent accidental shutdowns, poor operational continuity, and decreased collaboration efficiency. Moreover, when multiple agents participate in parallel, the shutdown and recovery processes lack a unified handling logic, affecting the work cycle and emergency response effectiveness.

[0162] In summary, under the relevant human-machine collaboration and platform-based operation and maintenance system for live-line work, multi-agent collaborative control has become a trend. However, related technologies generally lack a technical solution that can uniformly formalize the safety constraints unique to live-line work, perform consistent correction and conflict resolution on candidate control commands from different agents, and provide verifiable hierarchical rollback and recovery mechanisms in infeasible or high-risk situations, thereby ensuring both electrical safety and the continuity and efficiency of collaborative work.

[0163] The most significant and critical technical issue is that most multi-agent collaborative live-line working systems lack a mechanism to uniformly and formally express the safety rules specific to live-line work and integrate them throughout the entire collaborative control process. This results in each agent independently generating control commands or action suggestions under different information sources, time bases, and triggering logics, ultimately only able to converge through simple prioritization, interlocking, or manual confirmation. Due to the lack of a unified set of safety constraints and consistency verification, conflicts, mutual rejection, or repeated switching between candidate commands easily occur during the collaborative process, manifesting as frequent shutdowns, work interruptions, command oscillations, and control lags. It is difficult to maintain continuous and stable collaborative execution while meeting electrical safety clearances and operating procedures. This deficiency directly creates a contradiction where "safety relies on conservative strategies, and efficiency relies on manual coordination," becoming a key issue restricting the implementation and large-scale application of multi-agent collaborative live-line work.

[0164] Secondary issues:

[0165] (1) The safety control strategy relies too much on coarse-grained means such as “stopping / speed limiting”, resulting in low efficiency and uncontrollable cycle time.

[0166] When faced with personnel approaching, robots crossing boundaries, or sensor or communication malfunctions, related technologies often employ uniform handling methods such as increasing the safety distance, reducing speed, or emergency shutdown. While these methods improve safety margins, they lack the ability to make "minimum alteration corrections" to candidate instructions. This makes it impossible to achieve continuous operation when risks are controllable, leading to frequent accidental shutdowns, long waiting times, and large fluctuations in work cycle time. In particular, these inefficiencies are amplified when multiple agents collaborate in parallel.

[0167] (2) The feasibility of candidate instructions and the consistency of multi-agent systems lack systematic verification, which can easily lead to conflicts and unexecutable instructions.

[0168] In collaborative operations, robot controllers, platform strategies, edge nodes, and manual operations may simultaneously generate instructions or constraints on the same execution object. Many related systems lack a two-level mechanism of "feasibility verification—consistency verification," which easily leads to situations where instructions are locally feasible but globally infeasible, or individual constraints are satisfied but the group conflicts with each other, resulting in execution failures, action interruptions, or repeated triggering of safety boundaries.

[0169] (3) There is a lack of quantifiable mechanisms for handling uncertainties in perception and communication, which can easily lead to misjudgment or omission.

[0170] Live-line work sites are prone to obstructions, strong reflections, complex backgrounds, and weak network latency, leading to fluctuations in perception results such as personnel position, critical parts, and tool posture. Coordination links may also experience delays, packet loss, and asynchrony. Related technologies often lack constraint strength modulation and triggering strategies based on confidence levels or link quality, relying instead on fixed thresholds or empirical rules. This can result in either overly conservative approaches leading to false shutdowns under uncertain conditions, or insensitive thresholds causing missed detections.

[0171] (4) The lack of a unified handling logic for graded retreat and recovery results in high emergency response costs and poor risk controllability.

[0172] When candidate instructions become infeasible or the risk level suddenly increases, the relevant systems often simplify the response to an emergency stop or manual takeover, lacking a set of retreat actions (deceleration, maintaining, evacuating to a safe position, locking critical degrees of freedom, etc.) categorized by risk level, along with their triggering, sustaining, and recovery conditions. As a result, emergency response relies on personnel experience, recovery procedures are inconsistent, and problems such as repeated start-stop cycles, recovery difficulties, or excessively slow response are prone to occur, increasing on-site management costs and safety uncertainty.

[0173] (5) The configuration of key parameters and strategies for live-line work depends on manual debugging, and the cost of migration across scenarios is high.

[0174] Different voltage levels, tower structures, tool configurations, and work procedures correspond to different safety clearances, restricted areas, and control limits. Related systems often lack traceable parameter mapping and automatic configuration mechanisms, requiring repeated on-site debugging of thresholds and strategies. This results in high configuration costs, susceptibility to errors, and difficulty in reuse, impacting project delivery efficiency and scalability.

[0175] (6) The lack of structured explanation and traceability information for safety results affects visual prompts and acceptance assessments.

[0176] Most systems only provide alarm or shutdown status, lacking structured output of "trigger reasons, constraints, correction amounts, and rollback levels," making it difficult for operators to understand the basis for changes in safety boundaries. The platform struggles to generate auditable records and evaluation indicators, hindering on-site guidance, problem localization, and subsequent strategy iteration, as well as the formation of verifiable engineering deliverables.

[0177] The above technical issues collectively result in the fact that related multi-agent collaborative live-line working systems struggle to balance safety with continuity and efficiency, and lack a unified, verifiable, and traceable mechanism for correcting and reverting safety constraint commands.

[0178] This embodiment proposes a safety constraint command correction and rollback control method for multi-agent collaboration in live-line work. The core idea is to represent the electrical safety rules, personnel approach risks, collaborative conflict relationships, and communication and perception uncertainties of live-line work as a computable set of safety constraints under a unified time reference. Candidate control commands generated by multiple agents are corrected with minimum changes within the constraints, ensuring that the output commands satisfy safety constraints while maintaining the original collaborative intent as much as possible. When constraints cannot be met or uncertainties lead to a sudden increase in risk, graded rollback control is triggered according to the risk level, and the system smoothly returns to normal collaborative control after the recovery conditions are met. This closed-loop mechanism solves problems in the background technology such as inconsistent safety criteria, command conflicts, excessive shutdowns, misjudgments under weak network and obstruction conditions, and the lack of unified logic in emergency response.

[0179] 1. Terminology Definitions and Key Variable Descriptions:

[0180] To facilitate implementation by those skilled in the art, the key terms in this embodiment are defined as follows.

[0181] Multi-agent refers to a set of entities that participate in the collaborative control of live-line work and can generate control intentions or commands, including at least human agents, robot agents, and platform agents. Human agents refer to operators or supervisors, whose state is characterized by the position, posture, approach tendency, and intention confidence of key body parts. Robot agents refer to at least one working robot, whose state is characterized by joint position, joint velocity, end-effector pose, end-effector velocity, torque, or force sensing information. Platform agents refer to edge computing nodes or scheduling platforms, whose state includes task stage, collaboration rules, communication link quality, and safety policy parameters.

[0182] Candidate control commands refer to the control quantities or action suggestions proposed by each agent within the current control cycle. For unified processing, this embodiment abstracts candidate commands into control vectors. When a robot uses joint speed control, This represents the joint velocity vector; when end-effector velocity control is used... This represents the vector of linear velocity and angular velocity at the end point; when using trajectory point control... This represents the trajectory increment vector for the next control cycle. This embodiment preferably uses a joint speed control mode for compatibility with common industrial robot controllers.

[0183] The set of safety constraints refers to the set of constraints that must be satisfied within the current period, denoted as . This embodiment represents it as a combination of linear inequality constraints and equality constraints:

[0184] .

[0185] in , It is generated by electrical clearance constraints, personnel access constraints, mechanical limit constraints, and cooperative conflict constraints. , It can be used to represent synchronous actions or interlocking relationships that must be satisfied.

[0186] Minimum safety clearance refers to the minimum distance threshold that must be maintained between a live conductor and accessible conductive parts, personnel, or robots under a given voltage level and operating conditions. It is denoted as [missing information - likely a typo]. In this embodiment The voltage level, operating mode, and tool insulation status are determined, and a mapping table method is preferred, i.e., the value is obtained by looking up the index of voltage level, operating mode, and tool insulation status. , and serve as the basic parameters for constraint generation.

[0187] Confidence level is a scalar measure describing the reliability of perception and communication, ranging from [0, 1]. The confidence level for human perception is denoted as... The confidence level of the communication link is denoted as This embodiment modulates the conservatism of the constraints and the backoff trigger threshold with confidence level, so that the system maintains safety and robustness under uncertain conditions such as weak network, occlusion, and strong reflection.

[0188] Graded rollback refers to the system outputting rollback control of varying strengths according to levels when command correction is not feasible or the risk suddenly increases. The rollback level is denoted as L, and is preferably set to four levels, corresponding to deceleration, holding, retreating to a safe posture, and emergency braking, respectively.

[0189] 2. System software structure and data flow:

[0190] The software in this embodiment runs on a control platform or edge node and communicates with the robot controller. The software modules include a state acquisition and alignment module, a safety constraint generation module, a candidate instruction aggregation module, an instruction correction module, a feasibility and consistency verification module, a hierarchical rollback and recovery module, and a log and visualization interface module. The state acquisition and alignment module acquires data from human sensing devices, the robot controller, and the platform task management system and performs time alignment; the safety constraint generation module generates a constraint set based on the aligned state. The candidate instruction aggregation module collects candidate instructions from the robot body planner, platform scheduler, and human-machine interaction terminal and unifies them into vectors. The instruction correction module calculates correction instructions within the constraint set. The verification module verifies... Perform feasibility and consistency checks; when the checks fail, the rollback module outputs rollback instructions and manages the recovery logic; the interface module outputs structured information such as "correction amount, trigger reason, and rollback level" for visualization and traceability.

[0191] 3. Methodology and Key Processes:

[0192] The following section provides a complete breakdown of the software's data processing and control procedures. (Reference) Figure 3 This includes: acquiring and aligning the states of multiple agents in time to obtain state data of personnel, robots, and platforms; calculating human perception confidence and communication confidence, and modulating conservative factors to adaptively adjust safety margins; generating a set of safety constraints including electrical clearance, personnel approach, mechanical limits, and collaborative interlocks based on voltage level, operation mode, and tool status; aggregating candidate control commands from the robot itself, platform scheduling, and personnel interaction, and uniformly representing them as control vectors; performing minimum change correction on candidate commands within the safety constraint set, and solving quadratic programming to obtain corrected commands; performing feasibility verification and multi-agent consistency verification on corrected commands to determine whether they pass; if they fail, triggering graded backoff, outputting deceleration, hold, evacuation, or emergency braking according to the grade, and defining recovery conditions; outputting a structured information package containing correction amount, triggering reason, backoff level, etc., for visual prompts and audit traceability.

[0193] S1. Multi-agent state acquisition and time alignment:

[0194] The control period is set to a fixed sampling period. The optimal timeframe is 10ms to 50ms. The system collects and generates a unified state in each control cycle. The personnel side includes at least the three-dimensional position and velocity information of key parts, denoted as... , Key areas preferably include the hands, head, and center of the torso. The robot's sides should at least include joint locations. Joint velocity , terminal pose With terminal velocity The platform side should at least include task phase identifiers. Voltage level (V), operating mode and tool status The system includes communication link quality indicators, etc. It adds timestamps to data from different sources and aligns it to a unified time base using nearest neighbor interpolation or linear interpolation to ensure that the data involved in constraint generation and instruction correction belong to the same control cycle. To avoid abnormal data affecting control, the system performs validity checks on input states. When missing, abrupt, or out-of-bounds conditions occur, the system lowers the corresponding confidence level and enters a conservative strategy.

[0195] S2. Confidence level calculation and determination of conservative factors:

[0196] Human perception confidence The confidence level is determined by detection stability and multi-source consistency. A sliding window statistical method is preferred, calculating the variance and frame drop rate of key location positions within a window of length N, and combining this with the consistency error of depth or millimeter-wave ranging to form the confidence level. Communication Confidence Level It is calculated from end-to-end latency, jitter, and packet loss rate. and Calculate the conservation factor Used to modulate the minimum safety gap and backoff threshold, preferably set as follows:

[0197] .

[0198] in and These are non-negative weighting parameters. This leads to the conservative minimum safety gap. It enables the automatic expansion of the safety margin when uncertainty increases.

[0199] S3, Set of Security Constraints generate:

[0200] This embodiment generates at least the following four types of constraints and uniformly enters them. .

[0201] The first category is electrical clearance constraints. The system is based on voltage levels. With tool status Query results And establish a distance function for each charged body or equivalent risk source. ,in This indicates the robot's joint configuration. The following and the first The closest distance to a charged body. In order to control the quantity The above forms a computable constraint. In this embodiment, the distance function is linearized to the first order at the current time, resulting in:

[0202] .

[0203] This creates linear inequality constraints:

[0204] .

[0205] This constraint ensures that the distance in the next control cycle is not less than the conservative minimum safety clearance.

[0206] The second type is personnel proximity constraints. The system calculates the closest distance between the critical parts of a person and the dangerous parts of the robot. Hazardous areas preferably include the end effector and the connecting rod surfaces that may come into contact with personnel. Similarly, by constructing constraints using first-order linearization, the robot's movements converge or decelerate in a specified direction when personnel approach, resulting in:

[0207] .

[0208] in Can be with The difference lies in the use of these terms to reflect personnel safety distance strategies.

[0209] The third category is mechanical and dynamic constraints, including upper limits on joint velocity, soft limits on joint position, and upper limits on end-effector velocity, which are uniformly represented as linear amplitude limiting constraints. And further tighten the cap when approaching risk or when confidence levels decrease.

[0210] The fourth category is constraints related to coordination consistency and conflict resolution. This embodiment establishes occupancy and interlocking relationships for multi-agent collaborative tasks. For example, when a tool is occupied, another agent is prohibited from initiating a conflicting action, or a process is prohibited from proceeding to the next stage until it is confirmed. This type of constraint can be expressed as an equality or inequality, such as constraints on synchronized actions. For interlocked actions, a prohibited set constraint is introduced and transformed into a speed direction limit or task phase limit. For the clock constraint generated by platform-side scheduling, the upper bound of the target speed or the priority weight can be incorporated into the subsequent modified objective function.

[0211] The above constraints form a matrix , It also includes a constraint clause number for subsequent interpretation, output, and audit records.

[0212] S4. Convergence and unified representation of candidate control commands:

[0213] The robot body planner provides candidate joint velocities. Alternatively, the candidate end-effector velocity is converted into a joint velocity representation via a Jacobian matrix. The platform agent can provide task-level velocity upper limit adjustments or target direction suggestions, which are uniformly converted into joint velocity biases. When a human agent expresses an intention to confirm, pause, or fine-tune via an interactive terminal, the system converts this into constraints or biases, ultimately resulting in a unified candidate instruction.

[0214] .

[0215] The percentage of each source is recorded for subsequent interpretable output.

[0216] S5. Safety constraint command correction calculation:

[0217] This embodiment uses the principle of minimum change to modify candidate instructions, and preferably constructs a quadratic optimization problem to solve the modified instructions. :

[0218] .

[0219] in The diagonal weight matrix represents the correction cost and coordination priority of different joints. This problem is a quadratic programming problem with linear constraints, which can be solved in real-time within each control cycle by a standard quadratic programming solver. To ensure real-time performance, the system preferably uses a warm-start method, i.e., starting from... As initial values, they accelerate convergence. If the number of constraints in the current cycle changes significantly, feasible initial values ​​are reconstructed and solved within a limited number of iterations.

[0220] Output the correction amount after solving:

[0221] .

[0222] Record the clause number that triggers the strongest constraint and the corresponding margin to explain "why it was modified and how much it was modified".

[0223] S6. Feasibility and Consistency Verification:

[0224] When the quadratic programming solver returns no feasible solution, or the margin of the solution satisfying the constraints is lower than a preset threshold, it is determined to be infeasible. To avoid misjudgment caused by numerical errors, this embodiment performs a verification calculation on the key constraints, that is, uses the corrected... Back-substitution distance function Perform a geometric check to confirm that electrical clearances and personnel spacing meet the requirements.

[0225] Consistency checks are used to handle conflicts when multiple agents simultaneously constrain the same actuator. The system performs compatibility checks on the constraints corresponding to platform instructions, robot instructions, and human intentions. For example, if the platform requires acceleration while the electrical clearance constraint requires deceleration, a conflict is identified and handled according to the principle of prioritizing safety constraints. At the same time, the conflict type and processing result are output to ensure that collaborative decision-making is deterministic and traceable.

[0226] S7. Tiered rollback and recovery control:

[0227] Rollback is triggered when any of the following conditions occur: the quadratic programming has no feasible solution; the verification fails; the confidence level decreases, leading to an increase in the conservative gap and insufficient current distance margin; the communication confidence level falls below the threshold, resulting in a failure to guarantee collaborative consistency; or the risk level suddenly increases. The system determines the rollback level LLL based on the risk level and the degree of infeasibility.

[0228] When the rollback level is level 1, the output slows down and rolls back, specifically for... Apply proportional scaling and tighten the speed limit to ensure continuous motion while reducing risk. At level two, the output maintains the retraction, i.e., maintaining the current joint position or stabilizing at zero speed, and locking the task phase, awaiting personnel confirmation or confidence recovery. At level three, the output initiates evacuation retraction, i.e., evacuating to a safe configuration along a predefined safe direction or predefined safe attitude trajectory. The evacuation direction is preferably determined by the electrical clearance gradient or the personnel's departure direction. At level four, the output initiates emergency braking and triggers a hardware emergency stop or safety shutdown procedure.

[0229] The recovery control regulations specify the recovery conditions and methods. Recovery conditions must include at least: a critical distance margin greater than the recovery threshold; and If the error exceeds the threshold for several consecutive periods, the platform task phase allows recovery, and personnel have completed confirmation. The preferred recovery method is a smooth ramp transition from rollback instructions to normal correction instructions, avoiding mechanical shocks caused by sudden instruction changes.

[0230] S8, Structured Output and Visualization Interface:

[0231] To meet the requirements of on-site visualization and acceptance traceability, this embodiment outputs a structured information package in each cycle, which includes at least a timestamp and candidate instructions. Correction instructions Correction amount The triggered constraint clause number, constraint margin, and confidence level. , Conservative factors , rollback level The information packet includes a rollback reason code. This packet can be published to a visualization interface via message queue or industrial Ethernet to display the security status in real time, explain the reasons for the correction, and generate an auditable log.

[0232] 4. Explanation of the correspondence between technical effects:

[0233] By uniformly generating a set of safety constraints and modifying candidate instructions with minimal changes, this embodiment maintains the continuity of coordinated actions while meeting electrical clearance and personnel safety distance requirements, reducing excessive downtime and instruction oscillations common in the prior art. Through feasibility and consistency verification and tiered rollback and recovery control, this embodiment provides a definite closed-loop solution for infeasible and conflict scenarios, reducing emergency response costs and improving risk controllability. By introducing sensing and communication confidence to modulate constraint strength and rollback triggers, this embodiment exhibits higher robustness under conditions of occlusion, reflection, and weak network, reducing false positives and false negatives. By outputting structured correction and rollback reason information, this embodiment facilitates visual prompts and engineering acceptance traceability, improving cross-project reuse and parameter configuration consistency.

[0234] 1. Core idea:

[0235] The core of this case lies in: for multi-agent collaboration in live-line work, the requirements such as electrical safety clearance, personnel access safety, mechanical limiting and collaborative interlocking are uniformly formalized into a computable set of safety constraints, and the candidate control commands generated by the multi-agents are modified with the minimum change of safety constraints under a unified time reference; when the modification is not feasible or the risk increases suddenly, a graded backoff is triggered and recovery conditions are defined to achieve closed-loop collaborative control of "constraint unification - command modification - verification - backoff - recovery".

[0236] 2. Other key points:

[0237] (1) Multi-agent state acquisition and time alignment mechanism.

[0238] The system requires the collection of at least personnel status, robot status, environmental and electrical parameters, and timestamp alignment at a fixed sampling period to create a unified state for constraint generation and command correction. This key point distinguishes it from solutions that only perform local obstacle avoidance within a single controller.

[0239] (2) Unified generation mechanism for the set of safety constraints for live-line work.

[0240] The minimum safe clearance is determined based on the voltage level and operating conditions, and a set of safety constraints is generated by combining personnel approach risk, mechanical limit, and interlocking relationships. Furthermore, this set of constraints is shared by all agents within the same control cycle. This point highlights the "constraints specific to live-line work" and "cooperative consistency."

[0241] (3) The convergence and unified representation of candidate control instructions.

[0242] The requirement is to obtain candidate control commands from at least two agents and convert them into a unified control vector representation, which serves as input for command correction. This key point is used to cover collaborative models involving "platform / edge / robot" collaboration.

[0243] (4) Minimum change amount safety constraint instruction correction mechanism.

[0244] Within a set of safety constraints, the candidate control commands are solved to obtain a correction command that satisfies the constraints and minimizes the deviation from the candidate commands, and the correction amount is output. This key point is the crucial technical approach that distinguishes this case from "trigger-based shutdown / simple speed limiting".

[0245] (5) Feasibility verification and multi-agent consistency verification mechanism.

[0246] The requirement stipulates that at least one feasibility check be performed on the modified instructions, and a consistency check be performed on the interlocking of multi-agent cooperation or the compatibility of task phases. If the check fails, a rollback is triggered. This point is used to cover "conflict resolution" and "executability guarantee".

[0247] (6) Hierarchical rollback control and recovery control mechanism.

[0248] The guidelines stipulate that tiered rollbacks should be triggered only under conditions such as impracticality of correction, a sudden increase in risk, or a decline in confidence. They also specify the corresponding control actions, recovery conditions, and recovery methods for different rollback levels. This aspect aims to cover the certainty and verifiability of emergency response.

[0249] (7) Confidence modulation constraint strength or threshold mechanism.

[0250] The conservative factor is calculated based on human perception confidence and communication confidence, and the minimum safety gap or backoff threshold is adaptively modulated. This point is used to enhance robustness and creativity, avoiding being perceived as a simple superposition of rules.

[0251] 3. Key points for devices / systems:

[0252] It includes at least the following functional modules and defines the data interaction relationships between the modules.

[0253] (1) Status acquisition and time alignment module, used to acquire and align personnel, robot, environment and electrical parameters, and output a unified status.

[0254] (2) Safety constraint generation module, used to generate a set of safety constraints based on a unified state, and output the constraint clause number and margin.

[0255] (3) Candidate instruction aggregation module, used to collect and uniformly represent candidate control instructions of multi-agent.

[0256] (4) Instruction correction module, used to calculate correction instructions within the set of safety constraints and output correction amounts.

[0257] (5) Verification module, used to verify the feasibility and consistency of the correction instructions.

[0258] (6) Rollback and recovery module, used to trigger hierarchical rollback and recover collaborative control according to conditions.

[0259] (7) Interface and logging module, used to output structured information and form logs, which can be visualized and traced.

[0260] Other key points:

[0261] (1) The minimum safety gap is determined by querying the mapping table of “voltage level - working mode - tool insulation status” and modulating it by the confidence level conservative factor.

[0262] (2) The electrical clearance constraint is obtained by linearizing the distance function into a linear inequality constraint, and the distance in the next period is predicted to meet the threshold with a fixed sampling period.

[0263] (3) The instruction correction adopts a quadratic programming solution with linear constraints, supports hot start and iteration number limit to meet real-time requirements.

[0264] (4) Feasibility verification includes a recalculation of the distance function back-substituted by the corrected state to avoid relying solely on linearized approximation.

[0265] (5) Consistency verification includes compatibility checks of task phase interlocks, tool occupancy interlocks, or synchronous action constraints, and specifies that the priority of conflict handling is security constraints first.

[0266] (6) The retreat levels include at least deceleration, maintaining, retreating to a safe posture and emergency braking, and the recovery conditions are specified as distance margin, confidence level continuously meeting the threshold and personnel confirmation.

[0267] (7) The structured output should include at least the candidate instruction, the correction instruction, the correction amount, the trigger constraint clause number, the constraint margin, the confidence level, the rollback level and the rollback reason code to support visualization and traceability.

[0268] Technical effects:

[0269] This embodiment achieves the following technical effects through the technical means of "unified generation of security constraint set - correction of minimum change amount of candidate instruction - feasibility and consistency verification - hierarchical rollback and recovery - structured output", and overcomes the defects and deficiencies pointed out in the background technology respectively.

[0270] 1. Achieve unified safety criteria for multi-agent collaboration, eliminating instruction conflicts and oscillations caused by decentralized decision-making:

[0271] This embodiment unifies the minimum safety clearance corresponding to voltage levels, tool insulation status, personnel access constraints, and collaborative interlocking relationships into a unified set of safety constraints. Within the same control cycle, all agents share this same set of constraints for instruction processing. Therefore, robot planning, platform scheduling strategies, and personnel interaction intentions no longer employ different thresholds or triggering rules. Instead, they generate or modify execution instructions based on the same set of constraints, fundamentally reducing conflicts, mutual rejections, and repeated switching caused by multiple agents giving different safety judgments at the same time. This directly overcomes the problems of inconsistent safety criteria, instruction convergence relying on simple priorities or manual coordination, and the susceptibility to instruction oscillations and control lags in related technologies.

[0272] 2. Maintaining operational continuity while meeting live-line safety constraints significantly reduces accidental shutdowns and work interruptions:

[0273] This embodiment employs the principle of minimum change to modify candidate control commands within safety constraints, ensuring that the modified execution command minimizes deviation from the original candidate command while meeting electrical clearance and personnel safety distance requirements. Compared to common "trigger-to-stop" or coarse-grained speed-limiting strategies in related technologies, this embodiment prioritizes "correction over shutdown" to maintain continuous collaborative operation when risks are controllable, thereby reducing cycle time fluctuations caused by frequent shutdowns, waiting, and recovery. This directly overcomes the problems of low efficiency and uncontrollable cycle time resulting from excessive reliance on shutdowns / speed limits in related technologies.

[0274] 3. Provide verifiable feasibility and consistency verification mechanisms to improve instruction executability and reduce the risk of execution failure:

[0275] This embodiment performs feasibility and consistency checks before outputting execution instructions. Firstly, it verifies whether the modified instructions meet key safety constraints to prevent numerical errors or local approximations from violating these constraints. Secondly, it checks the consistency of interlocking relationships, task phase compatibility, and conflict types under the simultaneous action of multiple agents, ensuring that the final issued instructions are executable and do not conflict with other agent strategies. This verification mechanism identifies and handles situations where "locally feasible but globally infeasible" or "individually satisfied but collectively conflicting" occur before issuance, significantly reducing execution failures, action interruptions, or repeated alarm triggers. This overcomes the shortcomings of related technologies where a lack of systematic verification leads to frequent unexecutable instructions and conflicts.

[0276] 4. Establish a tiered retreat and recovery closed loop to reduce emergency response costs and improve risk controllability:

[0277] When candidate commands become infeasible under constraints, perception or communication uncertainties increase, or the risk level suddenly escalates, this embodiment can output reversal commands such as deceleration, maintaining, evacuating to a safe posture, or emergency braking according to the risk level, and specify the recovery conditions and recovery sequence to achieve closed-loop control from anomaly to recovery. Compared with related technologies that rely on manual takeover or a single emergency stop strategy, this embodiment provides deterministic and repeatable handling logic, avoiding unnecessary emergency stops and the risk diffusion caused by slow handling or inconsistent recovery procedures. This effect directly overcomes the problems of related technologies, such as the lack of unified reversal and recovery logic, high emergency handling costs, and difficult recovery.

[0278] 5. Improve robustness under occlusion, reflection, and weak mesh conditions by modulating constraint strength and backoff triggering based on confidence level:

[0279] This embodiment establishes a confidence index for the human perception and communication link, and uses the confidence index to modulate a conservative factor to adaptively expand the minimum safety gap and tighten control limits, while adjusting the backoff trigger threshold. In cases of strong occlusion, strong reflection, or increased communication delay, the system can enter conservative control early and back off promptly, reducing the risk of missed detections. After the confidence level recovers, the system converges constraints according to rules and smoothly resumes collaborative control, reducing efficiency losses caused by accidental shutdowns and prolonged conservative control. This effect directly overcomes the problems in related technologies where the lack of a quantification mechanism for uncertainty and reliance on fixed thresholds leads to misjudgments or missed detections.

[0280] 6. Output structured correction and rollback information to enhance interpretability, visual prompts, and project acceptance traceability:

[0281] This embodiment outputs structured data such as candidate instructions, correction instructions, correction amounts, trigger constraint clause numbers, constraint margins, confidence levels, rollback levels, and rollback reasons in each control cycle. This output enables on-site operators to understand the reasons for changes in safety boundaries and the current handling strategy, facilitating visual prompts and human-machine collaborative decision-making. Simultaneously, it generates auditable logs, supporting post-event traceability and parameter optimization, improving cross-scenario reuse and consistency in project delivery. This effect overcomes the shortcomings of related technologies that only provide alarm or shutdown status, lacking explanations of causes and traceability information, making acceptance evaluation and problem localization difficult.

[0282] In summary, this embodiment, through the aforementioned technical means, not only ensures electrical safety and personal safety during live-line work, but also improves the consistency, continuity, and robustness of multi-agent collaborative control, reduces the cost of accidental shutdowns and emergency response, and enhances explainability and traceability, thereby effectively overcoming major and minor technical problems.

[0283] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0284] Based on the same inventive concept, this application also provides a safety constraint command correction and rollback control device for implementing the aforementioned method for correcting and rolling back safety constraints for multi-agent collaboration in live-line work. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the safety constraint command correction and rollback control device for multi-agent collaboration in live-line work provided below can be found in the limitations of the method for correcting and rolling back safety constraints for multi-agent collaboration in live-line work described above, and will not be repeated here.

[0285] In one exemplary embodiment, such as Figure 4 As shown, a safety constraint command correction and rollback control device for multi-agent collaboration in live-line working is provided. This safety constraint command correction and rollback control device 400 for multi-agent collaboration in live-line working may include:

[0286] The data acquisition module 401 is used to collect the status data of multiple intelligent agents participating in live-line work according to a preset sampling period, and to perform timestamp alignment processing on the status data to obtain the unified status of multiple intelligent agents under a unified time reference; the status data includes at least personnel status data, robot status data, and environmental and electrical parameters.

[0287] The set generation module 402 is used to determine the minimum safe clearance for live-line work based on the voltage level and working conditions, and generate a set of safety constraints for live-line work by combining the personnel approach risk, mechanical limit and cooperative interlock relationship corresponding to the unified state; the set of safety constraints can be shared by multiple intelligent agents within the same control cycle;

[0288] The instruction acquisition module 403 is used to acquire candidate control instructions from at least two agents and convert the candidate control instructions into unified candidate instructions represented by a unified control vector.

[0289] The instruction processing module 404 is used to solve the unified candidate instruction based on the set of security constraints to obtain the corrected instruction that satisfies the set of security constraints and has the smallest deviation from the unified candidate instruction.

[0290] The instruction verification module 405 is used to verify the feasibility of the correction instruction and to verify the consistency of the cooperative interlocking or task phase compatibility among multiple intelligent agents, and to obtain the verification result.

[0291] The rollback control module 406 is used to trigger the hierarchical rollback control corresponding to the correction instruction when the verification result indicates that the verification has failed.

[0292] In an exemplary embodiment, the set generation module 402 is further configured to: determine the distance function information between the robot's joint configuration and the charged body based on the unified state; perform first-order linearization on the distance function information according to the minimum safety clearance to obtain electrical clearance constraints; determine the shortest distance between the critical parts of personnel and the dangerous parts of the robot based on the unified state; perform first-order linearization on the shortest distance to obtain personnel approach constraints corresponding to personnel approach risks; determine mechanical and dynamic constraints based on mechanical limits; determine cooperative consistency and conflict resolution constraints based on cooperative interlocking relationships; and generate a safety constraint set based on electrical clearance constraints, personnel approach constraints, mechanical and dynamic constraints, and cooperative consistency and conflict resolution constraints.

[0293] In an exemplary embodiment, the instruction processing module 404 is further configured to construct a quadratic optimization problem with the minimum deviation between the unified candidate instruction and the instruction to be solved as the optimization objective and the instruction to be solved satisfying the set of safety constraints as the constraint condition; wherein the deviation is determined based on a preset weight matrix, and the weight matrix is ​​used to characterize the correction cost and collaborative priority of different joints; and the quadratic optimization problem is solved to obtain the correction instruction.

[0294] In an exemplary embodiment, the instruction verification module 405 is further configured to perform feasibility verification on the correction instruction to obtain a feasibility verification result; the feasibility verification includes reviewing the electrical clearance between energized bodies and the distance between personnel; performing consistency verification on the collaborative interlocking or task phase compatibility among multiple intelligent agents to obtain a consistency verification result; the consistency verification includes compatibility checking of task phase interlocking, tool occupancy interlocking, or synchronous action constraints; and determining the verification result based on the feasibility verification result and the consistency verification result.

[0295] In an exemplary embodiment, the rollback control module 406 is further configured to determine the risk level and infeasibility degree when the verification result indicates that the verification fails; determine the rollback level based on the risk level and infeasibility degree; the rollback level includes deceleration, maintaining, retreating to a safe posture and emergency braking; trigger graded rollback control based on the rollback level; the recovery conditions corresponding to the graded rollback control include distance margin and confidence level continuously meeting thresholds and personnel confirmation.

[0296] In an exemplary embodiment, the device 400 further includes: an information generation module, configured to generate and output a structured information package in each control cycle; the structured information package includes at least a unified candidate instruction, a correction instruction, a correction amount, a triggered constraint clause number, a constraint margin, a confidence level, a rollback level, and a rollback reason code.

[0297] In an exemplary embodiment, the set generation module 402 is further configured to determine a conservative factor based on perception confidence and communication confidence; and to determine a minimum safety gap based on voltage level, operating conditions, and the conservative factor.

[0298] The modules in the aforementioned safety constraint command correction and rollback control device for multi-agent collaboration in live-line working can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0299] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements a safety constraint instruction correction and rollback control method for multi-agent collaborative operation in live-line work. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0300] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0301] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0302] In one exemplary embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above-described method embodiments.

[0303] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0304] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0305] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0306] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for correcting and backing down safety constraint commands for multi-agent collaborative live-line work, characterized in that, The method includes: According to a preset sampling period, the status data of multiple intelligent agents participating in live-line work are collected, and the status data is timestamped to obtain the unified status of the multiple intelligent agents under a unified time reference; the status data includes at least personnel status data, robot status data, and environmental and electrical parameters. The minimum safe clearance for live-line work is determined based on the voltage level and operating conditions. Combined with the personnel approach risk, mechanical limit, and collaborative interlocking relationships corresponding to the unified state, a set of safety constraints for live-line work is generated. This set of safety constraints is shared by the multiple intelligent agents within the same control cycle. Acquire candidate control instructions from at least two of the agents, and convert the candidate control instructions into unified candidate instructions represented by a unified control vector; Based on the set of security constraints, the unified candidate instruction is solved to obtain the corrected instruction that satisfies the set of security constraints and has the smallest deviation from the unified candidate instruction. The feasibility of the correction instructions is verified, and the consistency of the cooperative interlocking or task phase compatibility among the multiple intelligent agents is verified to obtain the verification results. If the verification result indicates that the verification failed, the hierarchical rollback control corresponding to the correction instruction is triggered.

2. The method according to claim 1, characterized in that, The method combines the personnel approach risk, mechanical limit, and collaborative interlocking relationships corresponding to the unified state to generate the safety constraint set for live-line work, including: Based on the unified state, the distance function information between the robot's joint configuration and the charged body is determined. According to the minimum safety clearance, the distance function information is linearized to the first order to obtain the electrical clearance constraint. Based on the unified state, the shortest distance between the key parts of the personnel and the dangerous parts of the robot is determined, and the shortest distance is linearized to obtain the personnel approach constraint corresponding to the personnel approach risk. Based on the aforementioned mechanical limits, determine the mechanical and dynamic constraints; Based on the aforementioned collaborative interlocking relationship, determine the collaborative consistency and conflict resolution constraints; The set of safety constraints is generated based on the electrical clearance constraints, the personnel access constraints, the mechanical and dynamic constraints, and the coordination and conflict resolution constraints.

3. The method according to claim 1, characterized in that, The process of solving the unified candidate instruction based on the set of security constraints to obtain the corrected instruction that satisfies the set of security constraints and has the smallest deviation from the unified candidate instruction includes: The optimization objective is to minimize the deviation between the unified candidate instruction and the instruction to be solved, and the constraint condition is that the instruction to be solved satisfies the set of safety constraints. A quadratic optimization problem is constructed. The deviation is determined based on a preset weight matrix, which is used to characterize the correction cost and cooperative priority of different joints. The correction instruction is obtained by solving the quadratic optimization problem.

4. The method according to claim 1, characterized in that, The feasibility verification of the modified instructions and the consistency verification of the cooperative interlocking or task phase compatibility among the multiple intelligent agents, to obtain the verification results, include: The feasibility of the correction instruction is verified to obtain the feasibility verification result; the feasibility verification includes a review of the electrical clearance between live parts and the distance between personnel. A consistency check is performed on the cooperative interlocks or task phase compatibility among the multiple intelligent agents to obtain a consistency check result; the consistency check includes compatibility checks and processing of task phase interlocks, tool occupancy interlocks, or synchronous action constraints. The verification result is determined based on the feasibility verification result and the consistency verification result.

5. The method according to claim 1, characterized in that, The step of triggering the hierarchical rollback control corresponding to the correction instruction when the verification result indicates that the verification failed includes: If the verification result indicates that the verification fails, the risk level and degree of infeasibility are determined. The retreat level is determined based on the risk level and the degree of infeasibility; the retreat level includes deceleration, maintaining, retreating to a safe position, and emergency braking; The graded rollback control is triggered based on the rollback level; the recovery conditions corresponding to the graded rollback control include distance margin and confidence level continuously meeting thresholds, as well as personnel confirmation.

6. The method according to claim 1, characterized in that, The method further includes: Within each control cycle, a structured information packet is generated and output; the structured information packet includes at least the unified candidate instruction, the correction instruction, the correction amount, the triggered constraint clause number, the constraint margin, the confidence level, the rollback level, and the rollback reason code.

7. The method according to any one of claims 1 to 6, characterized in that, Determining the minimum safe clearance for live-line work based on voltage level and operating conditions includes: Conservative factors are determined based on perceived confidence and communication confidence. The minimum safety clearance is determined based on the voltage level, the operating conditions, and the conservative factor.

8. A safety constraint command correction and rollback control device for multi-agent collaborative live-line work, characterized in that, The device includes: The data acquisition module is used to collect the status data of multiple intelligent agents participating in live-line work according to a preset sampling period, and to perform timestamp alignment processing on the status data to obtain the unified status of the multiple intelligent agents under a unified time reference; the status data includes at least personnel status data, robot status data, and environmental and electrical parameters. The set generation module is used to determine the minimum safe clearance for the live-line operation based on the voltage level and working conditions, and to generate a set of safety constraints for the live-line operation by combining the personnel approach risk, mechanical limit and cooperative interlock relationship corresponding to the unified state; the set of safety constraints is shared by the multiple intelligent agents within the same control cycle; The instruction acquisition module is used to acquire candidate control instructions from at least two of the intelligent agents and convert the candidate control instructions into unified candidate instructions represented by a unified control vector. The instruction processing module is used to solve the unified candidate instruction based on the set of security constraints to obtain the corrected instruction that satisfies the set of security constraints and has the smallest deviation from the unified candidate instruction. The instruction verification module is used to verify the feasibility of the correction instruction and to verify the consistency of the cooperative interlocking or task phase compatibility among the multiple intelligent agents, and to obtain the verification result. The rollback control module is used to trigger the hierarchical rollback control corresponding to the correction instruction when the verification result indicates that the verification has failed.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images