Multi-robot task allocation and power management system for semi-closed complex environment

Abstract

The application discloses a kind of multi-robot task allocation and power management systems for semi-closed complex environment, the system is composed of cloud server, edge computing node and robot end;The cloud server is used to formulate global task planning according to task type and urgency and send to edge computing node;The edge computing node is used to split and distribute global task according to the real-time state of robot cluster in robot end, dynamically dispatch robot in robot end, and built-in computing gateway and local cache;The robot end is used to execute subtask and autonomous decision, and initiates task handover according to power prediction.The application fuses navigation adaptation and communication redundancy mechanism, so that the robot system has the ability to run continuously in complex, dynamic and weak communication tunnel environment, providing strong environmental adaptation support for multi-robot collaborative work.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent inspection robots, and particularly relates to a multi-robot task allocation and power management system for a semi-closed complex environment. BACKGROUND

[0002] The current inspection robot system has significant technical shortcomings in a semi-closed scene such as a tunnel: firstly, task allocation is rigid, and existing solutions rely on fixed priorities or preset paths, which cannot dynamically respond to sudden tasks, resulting in response delays and resource waste; secondly, the environment is not adaptable, and when the communication signal is intermittent, the robot relies too much on cloud instructions, and in a weak network area, task interruption or data loss may occur; thirdly, the computing resources are fragmented, and centralized processing in the cloud causes high latency (> 500 ms), while the computing power of a single machine is limited, making it difficult to support high-load tasks such as real-time image analysis; fourthly, energy management is extensive, and there is a lack of power awareness and intelligent handover mechanism, so the robot directly interrupts the task and returns when the power is low, causing an inspection blind area (data from a certain tunnel project shows that the repeated inspection rate is as high as 30%). The existing technology has not realized the integration of dynamic task allocation, local intelligent decision-making, and energy scheduling under the cloud-edge-end collaborative architecture, which restricts the inspection efficiency and continuity in complex scenes. Therefore, there is a need for an inspection robot cluster system and a task allocation method that can effectively solve the above problems to improve the overall efficiency and reliability of the inspection work. SUMMARY

[0003] The application aims to provide a multi-robot task allocation and power management system for a semi-closed complex environment to overcome the shortcomings of the prior art.

[0004] The purpose of the application is achieved by the following technical solution: a multi-robot task allocation and power management system for a semi-closed complex environment, the system is composed of a cloud server, an edge computing node, and a robot end;

[0005] The cloud server is used to formulate a global task plan according to the task type and urgency and send it to the edge computing node;

[0006] The edge computing node is used to split and allocate the global task according to the real-time state of the robot cluster in the robot end, dynamically assign robots in the robot end, and embed a computing gateway and local cache;

[0007] The robot end is used to execute sub-tasks and make autonomous decisions, and initiate task handover according to power prediction.

[0008] Further, the cloud server is configured to formulate a global task plan according to the task type and urgency and send it to the edge computing node, specifically: the cloud server formulates a total inspection scheduling and task load value calculation across tunnels and regions according to the task type, task urgency and historical inspection results, and sends it to each edge computing node through an encrypted channel; continuously aggregates massive sensing data, images and logs from the edge computing node and the robot end, generates a device health report, a fault prediction model and a resource optimization suggestion through an AI analysis engine; undertakes multi-data center disaster recovery, algorithm model OTA upgrade and unified identity authentication of robots in the entire network, ensuring that the entire inspection system has high reliability, high security and scalable top-level management and control capability in semi-closed complex environments such as tunnel groups and underground pipe gallery groups.

[0009] Further, the edge computing node is configured to split and assign the global task according to the real-time state of the robot cluster in the robot end, dynamically assign robots in the robot end, and has a built-in computing gateway and local cache, specifically: the edge computing node splits the global task plan issued by the cloud server into subtasks that can be executed locally and assigns them to the robots in the robot end according to the real-time state of the robot cluster in the robot end; continuously collects the remaining power, position, load and sensor state of all robots in the region, dynamically assigns tasks to robots using a matching degree formula and distance cost model, and transmits key data to the on-duty robot in the supply station when the power of the robot is lower than the lower threshold of the power; the edge computing node has an industrial-grade computing gateway and local cache.

[0010] Further, the robot end is configured to execute subtasks and make autonomous decisions, and initiate task handover according to power prediction, specifically: the robot end is configured to efficiently execute subtasks assigned by the edge computing node; and make simple autonomous decisions in real time according to the on-site environment; initiate a task handover request according to the power prediction of the robot.

[0011] Further, the edge computing node includes a task allocation and handover module, an on-board computing node module and a complex environment design module.

[0012] The task allocation and handover module is configured to divide the work area, allocate tasks, monitor power and hand over tasks.

[0013] The on-board computing node module is configured to temporarily store data when disconnected and synchronize incremental information after connection.

[0014] The complex environment design module is configured to build a multi-level communication guarantee system composed of fixed relay nodes, mobile communication nodes and local cache mechanisms.

[0015] Further, the task allocation and handover module is used for work area partitioning and task allocation, specifically:

[0016] The task allocation and handover module is used to divide the overall work area into a plurality of relatively independent sub-areas according to the layout of the inspection environment, the distribution of equipment, and the task demand; obtain the load value of each sub-area task from the cloud server; the cloud server calculates and determines the load value of the task in the corresponding sub-area according to the equipment importance, inspection difficulty, and urgency of each sub-area; and calculates the matching degree according to the task arranged by the cloud server, selects the robot with the highest matching value, determines the number of selected robots, and then performs task allocation: sends instructions to the selected robot and lets the robot perform inspection work; if there is a shortage of robots, the robot with the highest matching degree is selected first, then the robot with the second highest matching degree is selected and task allocation is performed.

[0017] Further, the task allocation and handover module is used for power monitoring and task handover, specifically:

[0018] The task allocation and handover module is used to monitor the remaining power of each robot in real time, and when the power of the robot being inspected is lower than the lower limit threshold, a task handover request is received from the robot, and after receiving the task handover request, a task handover operation is performed: notify the robot of the same level or adaptive level at the supply station in standby state to prepare to take over the task; package and arrange the key information of the current task execution of the robot, and transmit it to the supply station robot that will take over the task; issue a command to the robot that has sent a task handover request to go to the supply station for charging;

[0019] The task allocation and handover module is used for continuous data interaction with the supply station robot that receives the task handover information during the process of going to the task handover area.

[0020] Further, the on-board computing node module is used for temporarily storing data when communication is interrupted and synchronizing incremental information after communication is restored, specifically: the on-board computing node module is used to temporarily store task data and sensing information when communication is interrupted or in weak communication conditions by relying on a local caching mechanism; after communication is restored, automatically trigger the synchronization mechanism of task status and logs, and interface with the task allocation and handover module, the complex environment design module, or the cloud server for incremental information.

[0021] Further, the complex environment design module is used to construct a multi-level communication guarantee system composed of fixed relay nodes, mobile communication nodes, and local caching mechanisms, specifically:

[0022] The complex environment design module is used for constructing a multi-level communication guarantee system composed of fixed relay nodes, mobile communication nodes and local cache mechanisms, and the quality monitoring and optimization mechanism of the communication link is added on the basis of the fixed relay nodes and the mobile communication nodes; the mobile communication node monitors the signal strength, bandwidth and delay and other parameters of the communication link in real time, and automatically adjusts the communication parameters and transmission strategy according to the monitoring result; when the communication link quality decreases, the mobile communication node automatically reduces the data transmission rate or uses a more robust coding scheme; in the case of unreachable relay or temporary link breakage, the local cache and breakpoint resume mechanism is enabled, and the key task state data and sensing information are saved and sequentially marked in the off-network state.

[0023] The beneficial effects of the present application are: the present application fuses navigation adaptation and communication redundancy mechanism, so that the robot system has the ability to continuously run in a complex, dynamic and weak communication tunnel environment, and provides strong environmental adaptation support for multi-robot collaborative operation. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 It is a structural diagram of a multi-robot task allocation and power management system for a semi-closed complex environment. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical scheme and advantages of the present application more clear and clear, the present application is further described in detail in combination with the drawings and examples, and it should be understood that the specific examples described here are only used to explain the present application, not all examples. Based on the examples in the present application, all other examples obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0026] The present application provides a multi-robot task allocation and power management system for a semi-closed complex environment, characterized in that the system comprises a cloud server, an edge computing node and a robot end to form a cloud-edge-end three-level collaborative architecture. Through the cloud-edge-end three-level collaborative architecture, the problems of imperfect cloud-edge-end collaboration of traditional inspection robot systems and lack of onboard intelligent capability are solved.

[0027] The cloud server is connected to the edge computing nodes through 4G / 5G network or satellite communication, responsible for global task planning and data aggregation, and supervises the global task from a macro perspective, plans and aggregates big data: (1) According to the task type (daily inspection, emergency repair, etc.), urgency and historical inspection results, the overall inspection scheduling and task load value calculation are made across tunnels and regions, and sent to each edge computing node through an encrypted channel; (2) Continuously gather massive sensing data, images and logs from edge computing nodes and robot ends, generate device health reports, fault prediction models and resource optimization suggestions through AI analysis engine, and further feed back to global strategy iteration; (3) Assume multi-data center disaster recovery, algorithm model OTA upgrade and unified identity authentication of network robots, ensure that the entire inspection system has high reliability, high security and extensible top-level control capability in semi-closed complex environments such as tunnel groups and underground pipe corridors.

[0028] The cloud server adopts distributed architecture and is deployed in multiple data centers to improve the reliability and disaster recovery capability of the system. At the same time, the communication between the cloud and the edge computing nodes adopts encrypted transmission protocol to ensure the security of the data; the selection and deployment position of the edge computing nodes are carefully planned, considering the layout of the inspection area, communication coverage and equipment distribution, etc. For example, in the tunnel, the edge computing nodes can be deployed at the entrance, middle section and exit of the tunnel to ensure that the communication distance with the robots does not exceed the effective range. The edge computing nodes also have environmental adaptability such as waterproof, dustproof and shock resistance, which can meet the long-term stable operation requirements in complex environments such as tunnels. At the same time, they use industrial-grade edge computing gateways with strong computing and storage capabilities, communicate with the cloud through wired or wireless networks, and use wireless communication technology to communicate with the robot end; the robots execute the assigned tasks and rely on the wireless communication module to maintain contact with the edge computing nodes.

[0029] In terms of data interaction, the data flow forms a closed loop from the cloud server to the edge computing nodes and then to the robot end. The cloud server formulates a global task plan according to the task type and urgency, and then sends it to the edge computing nodes. The edge computing nodes decompose and distribute the tasks according to the real-time state of the robot cluster in the robot end, and the robots in the robot end receive the tasks and start executing them, and then feed back the execution results and real-time data to the edge computing nodes, which are aggregated and then uploaded to the cloud server.

[0030] The cloud server is mainly responsible for macro task planning and data summary analysis, for example, according to the priority and urgency of the task, the overall arrangement of the inspection task is made, and various inspection data are analyzed and processed to provide support for subsequent decision-making. The robot end executes the inspection work according to the task allocated by the edge computing node, and in this process, the robot end can make real-time autonomous decisions according to the on-site environment. It is particularly worth mentioning that the inspection robot starts from the beginning or the end of the tunnel, and in order to ensure the endurance of the robot, multiple charging stations are set in the tunnel, which are reasonably distributed and convenient for the robot to supplement energy in time, so that the robot does not need to consider the problem of power supply, thereby better completing the inspection task.

[0031] The edge computing node plays a role in that it undertakes the function of "tactical level command" in the three-level cooperative architecture of cloud-edge-end: (1) the global task planning issued by the cloud server is real-time split into sub-tasks that can be executed locally and allocated to robots in the robot end; (2) the residual power, position, load and sensor state of all robots in the region are continuously collected, and the matching degree formula and distance cost model are used to dynamically assign tasks to the most suitable A / B / C level robots, and when the power of the robot is lower than the lower limit threshold of the power (in this embodiment, the lower limit threshold of the power is 30%), the task relay is triggered - the key data such as the current inspection route and the state of the inspected equipment are packaged and transmitted to the standby robot of the supply station, realizing "millisecond level" seamless handover; (3) the built-in industrial computing gateway and local cache in the node can still operate offline in the weak network or offline scene of the tunnel, and after the communication is restored, the state is incrementally synchronized to ensure that the task is continuous, the data is not lost, and the calculation is autonomous, the task is continuous, the data is synchronized, and the scheduling is relayed.

[0032] The robot end as "soldier level execution" has the following functions: (1) it has certain autonomy and simple computing ability, and can efficiently execute the sub-tasks allocated by the edge computing node; (2) when needed, it makes real-time simple autonomous decisions according to the on-site environment; (3) it predicts the power of the robot and initiates a task handover request.

[0033] The cloud server, edge computing node and robot end form a three-level cooperative architecture of cloud-edge-end through a heterogeneous communication network, which ensures efficient cooperative transmission of task instructions, state information and scheduling strategies, and supports seamless switching between online and offline modes. Even if the communication fails temporarily, the robot can still continue to execute the task relying on the edge computing node, and after the communication is restored, the state information is synchronized in time to ensure the continuous progress of the task.

[0034] The edge computing node includes a task allocation and handover module, an on-board computing node module and a complex environment design module.

[0035] I. Task allocation and handover module

[0036] 1.1 Task allocation

[0037] Robot inspection level division: The inspection robots are divided into three inspection levels A, B and C. Among them, the A-level robot has high-precision sensors and powerful data processing capability, which is suitable for high-complexity and high-accuracy inspection tasks; the B-level robot has moderate performance and can undertake routine inspection work; the C-level robot focuses on basic inspection and is suitable for simple tasks such as preliminary screening of equipment status.

[0038] Work area partitioning: According to the layout of the inspection environment, the distribution of equipment and the task requirements, the whole work area is divided into several relatively independent sub-areas. Each sub-area is calculated and determined by the cloud server according to its important degree, inspection difficulty, emergency degree and other factors to match the corresponding level of robots for work.

[0039] The cloud server needs to calculate the load value for each task when formulating the global task plan , which is determined by the three-dimensional weighting of equipment importance, inspection difficulty and emergency degree: , wherein represents the load value of the task in the th sub-area; is the equipment importance in the th sub-area, and the value can be 1, 2 and 3, 1 for ordinary equipment, 2 for important equipment, and 3 for core equipment; is the inspection difficulty of the th sub-area, and the value can be 1 and 2, 1 for simple environment and 2 for complex environment; is the emergency degree of the th sub-area, and the value can be 1 and 2, 1 for daily inspection and 2 for emergency inspection; , and are weight coefficients, which need to satisfy ; in this embodiment, takes 0.4, takes 0.3, takes 0.3.

[0040] Edge computing node task allocation method:

[0041] The task allocation and handover module in the edge computing node calculates the matching degree according to the task arranged by the cloud server. Because the robots have three levels A, B and C, their work efficiency is different for different equipment and different environments, so when allocating tasks, robots with high matching degree are required. The matching degree formula is as follows:

[0042] ;

[0043] wherein, is the matching degree; represents three levels of equipment importance, which are 1, 2 and 3 respectively; represents three levels of robots, which are A, B and C respectively; represents two levels of inspection difficulty, which are 1 and 2 respectively. represents the work efficiency of the robot of level and the equipment of equipment importance ; in this embodiment, , , , , , , , , . represents the work efficiency of the robot of level and the environment of inspection difficulty ; in this embodiment, , , , , , .

[0044] According to the calculated matching degree, the robot with the highest matching degree is applied, and the number of selected robots is determined according to the formula, and if the number is not an integer, it needs to be rounded up: wherein, represents the number of selected robots with the highest matching degree; is the task completion amount of a single robot. The edge computing node calculates the required number of robots to perform task allocation, sends instructions to the selected robots and lets the robots perform inspection work. If there is a shortage of robots, the robots with the highest matching degree are selected first, then the robots with the second highest matching degree are selected, and the required number of another level of robots is calculated by substitution. The formula is as follows:

[0045] wherein, represents the number of selected robots with the second highest matching degree; represents the task load value completed by the selected robot with the highest matching degree; the task load value is .

[0046] ​2.2 Task handover

[0047] Power monitoring and threshold setting: Real-time monitoring of the remaining power of each robot, the power of the robot below the lower limit threshold of power (30%) is the trigger threshold of task handover.

[0048] Task handover process: When the power of any one of the robots on patrol is lower than the lower limit threshold of power (30%), the robot automatically sends a task handover request to the edge computing node in the cluster, and also backs up its task execution status and key data locally to prevent data loss. After receiving the task handover request, the edge computing node performs task handover operations: (1) notifies the robots of the same level or adaptive level in the standby state of the supply station to prepare to take over the task; (2) packages and organizes the key information of the current task execution (such as the status data of the inspected equipment, the inspection route progress, etc.) and transmits it to the supply station robot that will take over the task to ensure the continuity and integrity of the task; (3) issues a command to the robot that has sent a task handover request to go to the supply station for charging. The robot arrives at the supply station for charging, and when the power reaches the working threshold (80%), it can participate in the task handover work. If no handover request is received, it will continue to charge at the charging station or stand by.

[0049] Task state migration and synchronization: After receiving the task handover information, the supply station robot quickly goes to the task handover area and continuously interacts with the edge computing node during the approach to the task handover position, updating its understanding and grasp of the task in real time. After arriving at the designated handover position, it performs the final task state confirmation and calibration with the low-power robot that originally executed the task to ensure that the task can be seamlessly and accurately transferred from the original robot to the supply station robot. If there is a problem with the robot state confirmation and calibration, the supply station robot executes the task information packaged and sent by the edge computing node, and continues to complete the subsequent inspection task, ensuring information transmission and task handover through double insurance. Here, the setting of multiple charging stations in the tunnel is particularly important, as they not only provide energy replenishment for the robots, but also provide convenient conditions for task handover, enabling the robots to perform relay more efficiently when executing tasks and avoiding task interruption due to insufficient power.

[0050] II. Onboard computing node module

[0051] The present application constructs an onboard edge computing node mechanism for tunnel inspection complex environment, aiming to improve the autonomous operation ability and task continuity of the inspection robot system under weak communication or temporary network interruption conditions, and enhance the collaborative efficiency and robustness of the entire multi-robot system.

[0052] During system operation, the cloud server is mainly responsible for global task planning and data aggregation and analysis, while the airborne edge node undertakes real-time on-site computation and decision-making required in the specific task execution process, reducing the problem of response lag caused by data transmission delay or network interruption. The system supports adaptive adjustment and strategy switching according to task priority, current state of the robot such as power, load, position and environmental changes, thereby realizing the organic combination of dynamic scheduling and local autonomy. The airborne edge computing node has certain local computing capability and can preliminarily analyze and process the images, sensor data and the like collected in the inspection process locally, quickly identify abnormal conditions and respond without waiting for instructions from the cloud, which greatly improves the efficiency and real-time performance of task execution. For example, in a tunnel environment with limited communication, the robot can rely on local computing capability to first filter and analyze the collected data, extract key information, and then transmit important data to the edge node or the cloud through limited communication bandwidth.

[0053] Considering the communication-limited scenarios such as tunnels, underground spaces or post-disaster ruins, the local mechanism particularly introduces local caching and delay synchronization technology. When the robot enters a communication blind area (communication interruption occurs or is in a weak communication condition), the local caching mechanism can temporarily store task data and sensor information to ensure uninterrupted operation; after the communication is restored, the airborne edge node will automatically trigger the task state and log synchronization mechanism to interface with the non-airborne module (task allocation and handover module or complex environment design module) of the edge computing node or the cloud to realize seamless transition and state consistency of the system. This design ensures the continuity of the task, so that the robot can continue to execute the task even in the case of unstable signal in the tunnel, and will not fail or lose data due to communication problems. At the same time, this also provides strong support for the autonomous operation capability of the robot, enabling the robot to more flexibly cope with various situations in complex and variable tunnel environments.

[0054] In summary, the present application provides reliable, stable and efficient operation support for multi-robot systems in high-risk, high-dynamic and weak-communication application scenarios by constructing an edge intelligent control system with autonomous computing, continuous tasks, synchronized data and relayable scheduling, which has good engineering feasibility and promotional value.

[0055] III. Complex environment design module

[0056] The present application is directed to the positioning accuracy decline, signal instability, task obstruction and other problems faced by multi-robot system operation in semi-closed complex environments such as tunnels, underground passages, converter valve halls and the like, and proposes a comprehensive adaptation mechanism with environment perception, task adaptation and communication guarantee capabilities. Based on fine map construction and dynamic division of task area, the mechanism fuses inertial navigation, laser SLAM and scene recognition technology to enable the inspection robot to achieve stable autonomous navigation and accurate positioning in a tunnel environment with complex spatial structure and unavailable GNSS signal. At the same time, the system supports an adaptive task area division strategy based on environmental features and task density, which can divide the overall operation area into multiple logical sub-areas according to the tunnel geometry, equipment distribution and task requirements, and automatically match the most suitable robot for deployment, thereby improving task coverage efficiency and reducing resource redundancy.

[0057] To cope with the actual problems of limited communication link and unstable network in the tunnel, the application further constructs a multi-level communication guarantee system composed of fixed relay nodes, mobile communication nodes and local cache mechanism. In the tunnel environment, the system uses the mobile characteristics of part of the inspection robots as temporary communication fixed relay nodes, constructs a flexible and adjustable communication grid structure, and guarantees the data backhaul and instruction reception of various robots during task execution. In the execution of complex tunnels and other semi-closed environments, the system will dispatch communication robots (inspection robots that do not perform tasks act as communication robots) together with the robot cluster to perform tasks: the communication robots automatically dock in the weak signal section according to the system instructions, start the vehicle-mounted Wi-Fi / LoRa relay mode, and relay the data packets sent by the nearby robots to the edge computing node; when the network topology changes due to tunnel bending or device shielding, the system recalculates and directs the robots to move to new optimal relay positions according to the real-time link quality, forming a dynamic communication grid that "stretches and contracts" with the scene changes, thereby ensuring that all robots can always smoothly backhaul data and receive instructions. At the same time, in the case of unreachable relay or temporary link interruption, the edge computing module inside the robot will enable the local cache mechanism to temporarily store task data and the breakpoint resume mechanism to save and sequentially mark critical task state data and sensor information, ensuring that task execution does not be interrupted and data is not lost due to temporary communication failure. The breakpoint resume mechanism takes the "task segment-time stamp-checksum" triplet as the core: when the communication is interrupted, the on-board computing node immediately cuts the current task into several "task segments" according to the completed devices, remaining route and sensor data, generates a unique checksum with a timestamp for each segment and writes it to the local cache; after the communication is restored, the mobile communication node sends the last successfully uploaded checksum to the edge side, the edge side compares the missing segment list according to the checksum and returns the "gap list", and the on-board computing node only needs to transmit the data packets corresponding to the gap list, without retransmitting the entire content, the whole process is like the reverse operation of "breakpoint download", which not only saves bandwidth but also ensures zero loss of inspection records and seamless connection of task status.

[0058] After the communication is restored, the system will automatically perform data incremental comparison and task state synchronization, and the data cached during the network interruption will be completely and accurately backhauled to the edge computing node or cloud server, thereby maintaining the consistency and continuity of the system task state. This "network interruption operation, network synchronization" design greatly improves the robustness and reliability of the robot system in complex closed environments, and provides a solid data support for subsequent task scheduling optimization and state evaluation.

[0059] In summary, the application integrates navigation adaptation and communication redundancy mechanism, enabling the robot system to have continuous operation capability in complex, dynamic and weak communication tunnel environment, and providing strong environmental adaptation support for multi-robot collaborative operation.

[0060] The present application has the following three aspects of innovation advantage:

[0061] 1) Intelligent mechanism of task allocation and handover: The present application innovatively constructs a cloud-edge-end three-level collaborative architecture composed of a cloud server, an edge computing node and a robot end in the aspects of task allocation and optimization and handover process, to realize dynamic task allocation and intelligent handover. The cloud server is responsible for global task planning, the edge computing node performs localized task decomposition and intelligent allocation, and the robot end is equipped with an on-board edge computing node and has certain autonomy and computing ability. The edge computing node collects task information, calculates the matching degree of the task and different levels of robots by using a matching degree formula, allocates the task to the optimal category of robot that is not occupied in turn after sorting the task according to the matching degree from high to low, and ensures that each task is executed by the most suitable robot. At the same time, the remaining power of the inspection robot is monitored in real time, and when the power is lower than the lower limit threshold, the robot automatically sends a task handover request to the edge computing node, the edge computing node notifies the standby robot of the supply station to prepare to take over the task, and the key information of the task execution is packaged and transmitted to the robot of the supply station, the robot of the supply station goes to the handover area, continuously interacts with the edge computing node data, and finally confirms and calibrates the task state with the low-power robot that originally executes the task, to ensure seamless and accurate migration of the task and continue to complete the subsequent inspection task.

[0062] 2) Local decision and cache mechanism of on-board computing node: The on-board computing node mechanism for tunnel inspection complex environment is constructed, the cloud server is responsible for global task planning and strategy optimization during system operation, and the on-board edge node undertakes real-time calculation and decision in the specific task execution process, to reduce the response lag problem caused by data backhaul delay or network interruption. In the communication limited scene, local cache and delay synchronization technology is introduced, when the robot enters the communication blind area, the local cache mechanism is relied on to temporarily store task data and sensing information, to ensure uninterrupted work; after the signal is restored, the task state and log synchronization mechanism is automatically triggered, to interface with the task allocation and handover module, the complex environment design module or the cloud server for incremental information, to realize seamless transition of the system and state consistency guarantee.

[0063] 3) Adaptive design in complex semi-closed environment: In terms of adaptive design in semi-closed environment, the application proposes a comprehensive adaptive mechanism with environmental perception, task adaptation and communication guarantee capability. Through fine map construction and dynamic division of task area, the fusion of inertial navigation, laser SLAM and scene recognition technology enables the inspection robot to realize stable autonomous navigation and accurate positioning in the tunnel environment with complex spatial structure and unavailable GNSS signal, improves the task coverage efficiency and reduces resource redundancy. At the same time, a multi-level communication guarantee system composed of fixed relay nodes, mobile communication nodes and local cache mechanism is constructed. The mobile characteristics of the communication robot itself are used as temporary communication relay nodes to build a flexible and adjustable communication grid structure, guaranteeing the data backhaul and instruction reception of the robot during task execution. When the relay is unreachable or temporarily disconnected, the edge computing module in the robot enables the local cache and breakpoint resume mechanism to save and sequentially mark the key task state data and sensing information. After the communication is restored, the data incremental comparison and task state synchronization are automatically performed to ensure the complete and accurate data backhaul.

[0064] This paper proposes a multi-robot task allocation and power management system for semi-closed complex environment. By constructing a cloud-edge-end three-level collaborative architecture composed of cloud server, edge computing node and robot, the system effectively solves the shortcomings of traditional inspection robot system in task allocation, environmental adaptability, computing resource utilization and energy management. The system realizes dynamic allocation and intelligent handover of tasks, improves the inspection efficiency and continuity, and enhances the autonomous operation capability of the robot through on-board edge computing node, ensuring the continuity of the task under weak communication or temporary network interruption. In addition, the adaptability design for complex environment enables the robot to operate stably under adverse conditions such as reduced positioning accuracy and unstable signal, complete the scheduled inspection task, and improve the efficiency and reliability of the inspection work as a whole, with good engineering feasibility and promotion value.

[0065] The above only describes the preferred embodiments of the application and is not intended to limit the application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the application shall be included in the scope of protection of the application.

Claims

1. A multi-robot task allocation and power management system for semi-closed complex environments, characterized by, The system is composed of a cloud server, an edge computing node and a robot end; The cloud server is configured to formulate a global task plan according to a task type and an urgency and send the global task plan to the edge computing node; The edge computing node is configured to split and allocate the global task according to a real-time state of a robot cluster in the robot end, dynamically assign robots in the robot end, and internally build a computing gateway and a local cache; The edge computing node comprises a task allocation and handover module; The task allocation and handover module is configured to divide a working area into sub-areas and allocate tasks, specifically as follows: The task allocation and handover module is configured to divide the whole working area into a plurality of relatively independent sub-areas according to a layout of an inspection environment, a distribution of equipment and a task demand; the cloud server needs to calculate a load value for each task when formulating the global task plan, and the load value is determined by a three-dimensional weighting of an equipment importance, an inspection difficulty and an urgency; the load value of each sub-area task is obtained from the cloud server; the cloud server calculates and determines the load value of the task in each sub-area according to the equipment importance, the inspection difficulty and the urgency; The matching degree is calculated according to the task arranged by the cloud server, the robot with the highest matching value is selected, the number of the selected robots is determined, and then the task allocation is performed: instructions are sent to the selected robots and the robots perform the inspection work; If there is a shortage of robots, the robot with the highest matching degree is selected first, then the robot with the second highest matching degree is selected and the task allocation is performed; The calculation formula of the matching degree is as follows: ； wherein, is a matching degree; represents three levels of equipment importance, which are 1, 2 and 3 respectively; represents three levels of robots, which are A, B and C respectively; represents two levels of inspection difficulty, which are 1 and 2 respectively; represents the working efficiency of the robot of level and the equipment of equipment importance ; represents the working efficiency of the robot of level and the environment of inspection difficulty of level According to the calculated matching degrees, the highest matching degree is taken The number of selected robots is determined according to the formula, and if the number is not an integer, it is rounded up: Wherein, Indicates the number of selected robots with the highest matching degree; The single robot task completion amount; The robot end is configured to execute sub-tasks and make autonomous decisions, and initiate task handover according to power prediction.

2. The multi-robot task allocation and power management system for semi-closed complex environments of claim 1, wherein, The cloud server is configured to formulate a global task plan according to a task type and an urgency and send the global task plan to the edge computing node, specifically as follows: the cloud server formulates a total inspection scheduling and task load value calculation across tunnels and areas according to a task type, a task urgency and historical inspection results, and sends the total inspection scheduling and task load value calculation to each edge computing node through an encrypted channel; continuously collects massive sensing data, images and logs from the edge computing nodes and the robot end, generates an equipment health report, a fault prediction model and a resource optimization suggestion through an AI analysis engine; undertakes multi-data center disaster recovery, algorithm model OTA upgrade and unified identity authentication of robots in the whole network, and ensures that the whole inspection system has high reliability, high security and extensible top-level control capability in a semi-closed complex environment such as a tunnel group or an underground pipe gallery group.

3. The multi-robot task allocation and power management system for semi-closed complex environments of claim 1, wherein, The edge computing node is used for splitting and distributing the global task according to the real-time state of the robot cluster in the robot end, dynamically assigning the robots in the robot end, and internally building a computing gateway and a local cache, specifically: the edge computing node splits the global task planned by the cloud server into subtasks that can be executed locally in real time according to the real-time state of the robot cluster in the robot end, and distributes the subtasks to the robots in the robot end for execution; the remaining power, position, load and sensor state of all robots in the region are continuously collected, the task is dynamically assigned to the robots by using a matching degree formula and a distance cost model, and when the power of the robot is lower than the lower limit threshold of the power, the key data is packaged and transmitted to the robot on standby in the supply station; the edge computing node internally builds an industrial-grade computing gateway and a local cache.

4. The multi-robot task allocation and power management system for semi-closed complex environments of claim 1, wherein, The robot end is used for executing subtasks and making autonomous decisions, and initiating task handover according to power prediction, specifically: the robot end is used for efficiently executing the subtasks distributed by the edge computing node; and making simple autonomous decisions in real time according to the on-site environment; task handover request is initiated according to power prediction of the robot.

5. The multi-robot task allocation and power management system for semi-closed complex environments of claim 1, wherein, The edge computing node further includes an on-board computing node module and a complex environment design module; The task distribution and handover module is used for work area partitioning, task distribution, power monitoring and task handover. The on-board computing node module is used for temporarily storing data when the connection is interrupted and synchronizing incremental information after the connection is restored. The complex environment design module is used for constructing a multi-level communication guarantee system composed of fixed relay nodes, mobile communication nodes and local cache mechanisms.

6. The multi-robot task allocation and power management system for semi-closed complex environments of claim 5, wherein, The task distribution and handover module is used for power monitoring and task handover, specifically: The task distribution and handover module is used for monitoring the remaining power of each robot in real time, receiving a task handover request from a robot whose power is lower than the lower limit threshold of the power when the robot is on patrol, and performing a task handover operation after receiving the task handover request: notifying the robots of the same level or adaptive level in the standby state of the supply station to prepare to take over the task; packaging and arranging the key information of the current task execution of the robot, and transmitting the information to the supply station robot that will take over the task; issuing a command to the robot that has sent the task handover request to go to the supply station for charging; The task distribution and handover module is used for continuously interacting with the supply station robot that receives the task handover information during the process of going to the task handover area.

7. The multi-robot task allocation and power management system for semi-closed complex environments of claim 5, wherein, The on-board computing node module is used for temporarily storing data when the connection is interrupted and synchronizing incremental information after the connection is restored, specifically: the on-board computing node module is used for temporarily storing task data and sensing information by relying on the local cache mechanism when the communication is interrupted or in a weak communication condition; after the communication is restored, the synchronization mechanism of task state and log is automatically triggered, and incremental information is connected with the task distribution and handover module, the complex environment design module or the cloud server.

8. The multi-robot task allocation and power management system for semi-enclosed complex environments of claim 5, wherein, The complex environment design module is used for constructing a multi-level communication guarantee system composed of fixed relay nodes, mobile communication nodes and local cache mechanisms, specifically: The complex environment design module is used for constructing a multi-level communication guarantee system composed of fixed relay nodes, mobile communication nodes and local cache mechanism. On the basis of the fixed relay nodes and the mobile communication nodes, a quality monitoring and optimization mechanism of the communication link is added. The mobile communication nodes monitor parameters such as signal strength, bandwidth and delay of the communication link in real time, and automatically adjust the communication parameters and the transmission strategy according to the monitoring results. When the quality of the communication link decreases, the mobile communication nodes automatically reduce the data transmission rate or adopt a more robust coding scheme. In the case that the relay is unreachable or the communication link is temporarily disconnected, the local cache and the breakpoint resume mechanism are enabled, and the key task state data and the sensing information are saved and sequentially marked in the disconnected network.

Images