Public object warehouse-oriented robot patrol mode adaptive switching and voice interaction system

Abstract

The application discloses a public article warehouse-oriented robot patrol mode adaptive switching and voice interaction system, which comprises a mobile terminal, a robot body, a public article warehouse map and task configuration module, a state acquisition module, a patrol mode adaptive switching module, a voice interaction module, an abnormality identification and review module, an internet of things equipment sensing module, an environmental linkage control module and a communication and log module. The public article warehouse-oriented robot patrol mode adaptive switching and voice interaction system disclosed by the application constructs a state vector, determines a target patrol mode in combination with mode scoring, preemption arbitration, path, power, communication state and no-entry area safety verification, and generates a corresponding path, an acquisition action and a back transmission strategy. The voice interaction module analyzes user voice into a structured instruction, triggers a patrol, review, lending or adjustment task after permission verification, and realizes abnormality review, weak network back transmission, entity warehouse / virtual warehouse synchronization and closed loop processing.

Description

Technical Field

[0001] This invention belongs to the field of public warehouse asset management and mobile robot intelligent inspection technology. Specifically, it relates to a public warehouse robot dog inspection system that establishes a standardized cross-module bidirectional linkage link, distinguishes instruction relay transmission paths, and can realize adaptive switching of inspection modes and a closed-loop natural language voice interaction process. Background Technology

[0002] Public property warehouses in administrative and public institutions centrally store idle, borrowed, returned, transferred, and pending disposal physical assets, simultaneously supported by an online virtual asset ledger. The core management requirements balance on-site environmental safety, comprehensive physical inventory checks, and full traceability of asset transfer processes. Existing public property warehouse inspection plans have several structural deficiencies: 1. There are serious data silos among various business modules. There is a lack of standardized interaction links between modules. Data from map ledgers, environmental sensors, voice interaction, inspection decision-making, and anomaly handling units are fragmented and cannot participate in joint decision-making. The command transmission path is unclear, and the logic of inspection collection strategies, abnormal parameters, and alarm data flow is chaotic, resulting in command issuance gaps and data asynchrony. 2. Manual inspections are costly, pose safety risks during nighttime operations, and result in fragmented inspection records; fixed surveillance cameras have numerous blind spots at the bottom of shelves, in narrow aisles, and in temporary storage areas, making it impossible to verify assets at close range; traditional wheeled inspection robots lack obstacle-crossing and terrain-adapting capabilities, making them unsuitable for the complex aisle environment of public warehouses. 3. Existing inspection robots only support fixed routes and a single inspection mode, and cannot dynamically adjust inspection strategies based on multi-source data such as warehouse risk, asset lifecycle, robot dog battery level, network quality, and temporary voice service commands; when multiple tasks are concurrent, there is a lack of cross-module collaborative preemption arbitration and layered security verification mechanisms, which can easily lead to security risks such as crossing restricted areas, low battery shutdown, and task conflicts. 4. The voice interaction unit is not deeply integrated with the public property warehouse map asset ledger. It only supports simple start and stop commands and cannot convert natural language into standardized business tasks such as inventory, adjustment, and anomaly review. It also lacks cross-module collaboration logic for information completion and follow-up questions, and multi-level permission verification. 5. The environmental perception, anomaly identification, inspection decision-making, and environmental control modules are independent of each other. Environmental and asset anomalies are only pushed to alarms in one direction, without a complete closed loop of "perception, classification, mode switching, equipment linkage, manual review, and ledger update". 6. In scenarios with weak network, low power, and high priority task preemption, there is no unified caching and relay mechanism, the inspection task is directly interrupted, the breakpoint data and collection records are lost, and the data on the physical warehouse site is inconsistent with the virtual warehouse ledger. 7. The interaction, operation, and disposal records of the entire process modules are stored in a scattered manner, without a unified log collection bus, which cannot meet the requirements of full-chain traceability management of assets of administrative and public institutions. Summary of the Invention

[0003] The main objective of this invention is to provide a robotic dog-based inspection mode adaptive switching and voice interaction system for public warehouses. The system constructs a state vector based on asset number, storage location number, borrowing status, return status, transfer status, ledger synchronization status, and information such as temperature, humidity, water immersion, smoke detection, and illuminance. It combines mode scoring, preemption arbitration, path selection, battery level, communication status, and restricted area security checks to determine the target inspection mode and generates corresponding paths, data collection actions, and feedback strategies. The voice interaction module parses user voice into structured commands, triggering inspection, review, borrowing / returning, or transfer tasks after authorization verification, achieving anomaly review, weak network feedback, physical / virtual warehouse synchronization, and closed-loop processing.

[0004] To achieve the above objectives, this invention provides a robotic dog inspection mode adaptive switching and voice interaction system for public storage facilities, comprising a mobile terminal, a robotic dog body, a public storage facility map and task configuration module, a status acquisition module, an inspection mode adaptive switching module, a voice interaction module, an anomaly recognition and verification module, an IoT device sensing module, an environmental linkage control module, and a communication and logging module, wherein: The mobile terminal is bidirectionally connected to the voice interaction module and the communication and log module, and is used to issue voice / touch commands, receive inspection feedback, abnormal alarms and ledger synchronization results; the robot dog body communicates bidirectionally with all back-end business modules through the communication and log module, performs inspection actions and transmits back the first relevant sensor data. The public property warehouse map and task configuration module respectively output a standardized electronic map, a basic inspection task to status collection module, an inspection mode adaptive switching module, a voice interaction module, and an anomaly recognition and verification module, providing a spatial benchmark for full-process path planning, point location, and area permission verification. The IoT device sensing module collects second relevant sensor data in real time and pushes environmental risk values ​​to the status acquisition module, anomaly identification and verification module, and environmental linkage control module simultaneously to achieve full-domain sharing of environmental data. The status acquisition module connects to the public property warehouse map and task configuration module, the robot dog itself, the IoT device sensing module, the voice interaction module, and the anomaly recognition and verification module. It aggregates multi-dimensional discrete raw data and completes normalization processing, and pushes the normalized standardized parameters one-way to the inspection mode adaptive switching module. The inspection mode adaptive switching module is used to construct a public warehouse state vector based on the data obtained by the state acquisition module, calculate the priority values ​​of multiple candidate inspection modes, and determine the target inspection mode after performing security verification on the relevant parameters corresponding to the inspection mode to be switched. The voice interaction module receives the voice data stream uploaded by the mobile terminal, combines the public warehouse map with the relevant information provided by the task configuration module to complete semantic parsing and permission verification, and generates structured instructions which are then output in two channels. The anomaly identification and verification module receives the collected data transmitted back by the robot dog, the environmental sensing data output by the IoT device sensing module, and the anomaly identification threshold adjustment command issued by the inspection mode adaptive switching module. It identifies various anomalies and determines the level of the anomaly event. The anomaly level is synchronously transmitted back to the inspection mode adaptive switching module to trigger mode preemption. At the same time, it is pushed to the environmental linkage control module to start environmental handling and to the communication and log module to forward the alarm generated by the mobile terminal. After the anomaly verification task is completed, the verification result is reverse-synchronized to the public property warehouse map and task configuration module to update the ledger status. The environmental linkage control module receives real-time environmental data from the IoT device sensing module, anomaly level signals from the anomaly identification and verification module, and regional risk levels synchronized by the inspection mode adaptive switching module. When the indicators exceed the threshold, it outputs device control commands to drive relevant devices and sends the linkage execution results back to the communication and log module for archiving.

[0005] As a further preferred technical solution to the above technical solution, the inspection mode adaptive switching module includes a state vector construction unit, a mode scoring unit, a mode arbitration unit, a security verification unit, and a strategy generation unit. The inspection mode adaptive switching module receives normalized parameters output by the state acquisition module, structured business instructions issued by the voice interaction module, and abnormal event level signals returned by the abnormal identification and review module. The built-in state vector construction unit assembles and generates a public warehouse state vector based on the normalized parameters, and then generates a target inspection acquisition execution strategy through mode weighted scoring, preemptive arbitration, and multi-dimensional security verification.

[0006] As a further preferred technical solution to the above technical solution, the candidate inspection modes include at least three of the following: regular inspection mode, key asset enhanced inspection mode, asset entry verification mode, in-warehouse inventory mode, anomaly review inspection mode, voice real-time inspection mode, remote takeover mode, outbound review mode, allocation and escort mode, and safe return mode. The differentiated travel speed, sensor collection frequency, point dwell time, and data transmission priority configuration corresponding to different inspection modes are issued by the inspection mode adaptive switching module to the communication and log module for forwarding to the robot dog itself. At the same time, the inspection mode adaptive switching module synchronously sends the corresponding anomaly recognition threshold parameters to the anomaly recognition and review module to dynamically adjust the anomaly recognition sensitivity. The public warehouse status vector includes at least five of the following: current or target public warehouse area, area risk level, current task type, abnormal event level, abnormal duration, robot dog remaining battery power, network quality, obstacle density or path accessibility, personnel or forklift interference status, user command source, user command priority, asset lifecycle status, physical warehouse / virtual warehouse synchronization status, and environmental risk status.

[0007] As a further preferred technical solution to the above technical solution, the mode scoring unit performs weighted calculation on each state item in the public warehouse state vector according to the weight parameters corresponding to different candidate inspection modes, and combines the abnormal event level, voice command priority, asset life cycle status, physical warehouse / virtual warehouse synchronization status and environmental risk status to form the priority value of each candidate inspection mode.

[0008] As a further preferred technical solution to the above technical solution, the mode arbitration unit presets a preemption priority rule, which is as follows: when at least one security event is detected, the security return mode preempts the current inspection mode; when the level of the abnormal event reaches the review threshold, the abnormal review inspection mode preempts the regular inspection mode or the key asset enhanced inspection mode; when a voice command that has passed the permission verification is marked as an immediate task, the voice immediate inspection mode is written into the task queue and preempted or queued for execution according to the task priority; when asset entry, return, transfer or inventory tasks are marked as high priority, the asset entry verification mode, exit review mode, transfer escort mode or warehouse inventory mode participate in the preemption arbitration in a preset order.

[0009] As a further preferred technical solution to the above technical solution, the safety verification unit verifies at least three of the following before switching the inspection mode: robot dog pose stability, target path reachability, whether the path passes through a restricted area, whether the remaining power is sufficient for task execution and return, whether the target area communication status is sufficient for data transmission, and whether the physical / virtual warehouse synchronization status is sufficient for task execution. When the safety verification fails, the mode arbitration unit maintains the current inspection mode, enters a waiting avoidance state, requests confirmation from the mobile terminal, or enters a safe return mode.

[0010] As a further preferred technical solution of the above technical solution, the inspection mode adaptive switching module maintains the task queue and task breakpoint information; when the target inspection mode preempts the current inspection task, it records the completed inspection points, incomplete inspection points, current location and collected data of the current inspection task, and generates a breakpoint continuation inspection task after the preempted task is completed, the power is restored or the network is restored.

[0011] As a further preferred technical solution to the above technical solution, the communication and logging module connects the mobile terminal, the robot dog itself, and all background functional modules to complete the issuance of instructions, the transmission of inspection data, and the push of abnormal alarms; under weak network conditions, data is transmitted in a hierarchical cache, and after the network is restored, the interrupted inspection records, physical warehouse data collection data, and virtual warehouse ledger are synchronized; at the same time, all module interaction signals, mode switching records, voice operations, abnormal handling, and equipment linkage behaviors are collected to uniformly generate an unalterable log.

[0012] As a further preferred technical solution of the above technical solution, the mobile terminal is also used to display images, videos, locations, abnormality levels, review results and processing status of abnormal events, and to receive user input via touch or voice to confirm abnormality, false alarm, re-review, transfer to manual processing, mark as processed, confirm adjustment or confirm return.

[0013] As a further preferred technical solution to the above technical solution, the system's workflow is as follows: S1: System initialization, the public property warehouse map and task configuration module directly pushes the electronic map, asset ledger and regional risk benchmark data to the status acquisition module, inspection mode adaptive switching module, voice interaction module and anomaly recognition and review module to generate basic routine inspection tasks; S2: The status acquisition module directly connects to various business modules to aggregate multi-dimensional status data, constructs a public warehouse status vector, and pushes it unidirectionally to the inspection mode adaptive switching module. S3: The inspection mode adaptive switching module retrieves the public property warehouse map and task configuration module parameters to complete mode weighted scoring, preemption arbitration, and multi-dimensional security verification, and generates the target inspection collection strategy and corresponding anomaly identification threshold. The complete collection strategy is sent to the communication and log module, which forwards it to the robot dog for execution. The anomaly threshold parameters are directly synchronized to the anomaly identification and review module. S4: The robot dog completes the point inspection according to the collection strategy. The visual, environmental and radio frequency collection data are relayed back to the anomaly identification and verification module through the communication and log module. S5: The IoT device sensing module pushes environmental data directly to the anomaly identification and verification module and the environmental linkage control module in real time; the anomaly identification and verification module determines the anomaly level based on the received identification threshold, and the anomaly level is directly pushed to the inspection mode adaptive switching module to trigger mode preemption, and directly pushed to the environmental linkage control module to start device handling; anomaly alarms are pushed to the mobile terminal via the communication and log module; S6: Managers issue voice commands via mobile terminals. The data flows through the communication and log module to the voice interaction module. The voice interaction module directly connects to the public property warehouse map and task configuration module to verify asset and area information. If key parameters are missing, the communication and log module issues voice prompts. After the permission verification is passed, the structured command is directly pushed to the inspection mode adaptive switching module to participate in the arbitration. S7: In the event of low power, weak network, or high-priority task preemption, the inspection mode adaptive switching module will store the breakpoint location and data collection cache in the communication and log module; after power / network recovery, the cached breakpoint will be read to generate a continued inspection strategy, which will be sent to the robot dog via the communication and log module; after network recovery, the communication and log module will synchronize all cached data collection data to the public property warehouse map and task configuration module to update the virtual warehouse ledger. S8: All system module interactions, mode switching, voice operation, anomaly handling, and device linkage behaviors are uniformly collected and archived by the communication and log module to generate complete and traceable inspection business logs.

[0014] The beneficial effects of this invention are as follows: 1. Strictly distinguish between two types of data transmission links and clarify logical boundaries: The robot dog must transmit the complete inspection and data collection strategy through the communication and log module; the module collaboration uses the threshold, risk level, and map benchmark data backend module for direct transmission and interaction, which solves the problems of confusion and logical conflict in the original technical command transmission path; 2. Build a standardized, direct-connection architecture for all back-end modules to eliminate data silos. Multi-source data, including map ledgers, environmental sensors, voice commands, and abnormal signals, collaborate to participate in joint decision-making in the inspection mode, adapting to diverse public warehouse business scenarios such as inventory, allocation, anomaly review, and safe return. 3. The voice interaction module is deeply integrated with the warehouse map and asset ledger, enabling cross-module collaboration for semantic verification, information inquiry, and multi-level permission verification. Managers can directly issue standardized asset inspection tasks using natural language, and the entire process of human-computer interaction and back-end inspection decision-making is seamlessly connected. 4. Construct a complete closed loop for anomalies, including "environmental perception, anomaly classification, mode switching, equipment linkage, terminal alarm, review and archiving, and ledger update", and coordinate the handling of asset and environmental risks by multiple modules to solve the defects of traditional systems that only provide alarms, lack linkage, and have outdated ledgers. 5. Using the communication and log module as a unified cache relay bus, it enables breakpoint caching, automatic patrol resumption, and batch ledger synchronization in scenarios such as weak network, low power, and task preemption, ensuring the continuity of inspection and the consistency of physical / virtual warehouse ledgers. 6. The robot dog has excellent obstacle-crossing and low-space passage capabilities, and can cover fixed monitoring blind spots such as the bottom of shelves, narrow passages, and temporary storage areas, greatly improving the accuracy of asset inventory and on-site verification; 7. All system operations, interactions, and disposal behaviors are uniformly logged, and the entire asset inspection process is traceable, fully meeting the standardized and compliant management requirements for public property assets of administrative and public institutions. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the overall structure of the public property warehouse robot dog inspection system of the present invention; Figure 2 This is a schematic diagram of the structure of the inspection mode adaptive switching module of the present invention; Figure 3 This is a schematic diagram illustrating the public warehouse state vector construction, pattern scoring, and preemption arbitration process of the present invention. Figure 4 This is a schematic diagram of the voice interaction and inspection task generation process of the present invention; Figure 5 This is a schematic diagram of the anomaly identification, verification, and mobile terminal closed-loop processing flow of the present invention; Figure 6 This is a schematic diagram of the breakpoint resumption process under weak network, low power, and task preemption scenarios of the present invention.

[0016] The attached diagram includes the following labels: 100 - mobile terminal; 200 - robot dog body; 300 - server or edge computing terminal; 310 - public property warehouse map and task configuration module; 320 - status acquisition module; 330 - inspection mode adaptive switching module; 340 - voice interaction module; 350 - anomaly recognition and verification module; 360 - communication and log module. Detailed Implementation

[0017] The following description is intended to disclose the present invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art. The basic principles of the invention defined in the following description can be applied to other embodiments, modifications, improvements, equivalents, and other technical solutions that do not depart from the spirit and scope of the invention.

[0018] In the preferred embodiments of the present invention, those skilled in the art should note that the robot dog and the like involved in the present invention can be considered as prior art.

[0019] Preferred embodiment.

[0020] like Figure 1-6 As shown, this invention discloses a robot dog inspection mode adaptive switching and voice interaction system for public storage areas, including a mobile terminal, a robot dog body, and a server or edge computing terminal. The server or edge computing terminal includes a public storage area map and task configuration module, a status acquisition module, an inspection mode adaptive switching module, a voice interaction module, an anomaly recognition and verification module, an IoT device sensing module, an environmental linkage control module, and a communication and logging module, wherein: The mobile terminal is bidirectionally connected to the voice interaction module and the communication and logging module, used to issue voice / touch commands, receive inspection feedback, anomaly alarms, and synchronize ledger results. The robot dog itself communicates bidirectionally with all backend business modules through the communication and logging module, executes inspection actions, and transmits back the first relevant sensor data (visual, environmental, and positioning data, etc.). The mobile terminal can be a mobile phone, tablet, handheld terminal, or public property warehouse management terminal, used to display the public property warehouse map, inspection points, robot dog's real-time location, battery level, network status, task progress, anomaly alarms, asset transfer records, and inspection reports. Users can select inspection areas, set inspection frequencies, view anomaly details, remotely take over, or input voice commands through the mobile terminal.

[0021] The mobile terminal is also used to display images, videos, locations, anomaly levels, review results, and processing status of abnormal events, and to receive user input via touch or voice to confirm anomalies, false alarms, re-review, transfer to manual processing, mark as processed, confirm reassignment, or confirm return.

[0022] The robot dog is a mobile inspection robot with a quadrupedal locomotion mechanism. It is equipped with a motion control unit, posture control unit, positioning unit, communication unit, audio acquisition and playback unit, and one or more inspection sensors. These sensors include, but are not limited to, visible light cameras, infrared thermal imaging cameras, depth cameras, LiDAR, ultrasonic sensors, temperature and humidity sensors, water immersion / leak detectors, networked smoke / temperature detectors, illuminance sensors, gas sensors, RFID readers, and barcode readers. When the robot dog detects personnel, forklifts, shelf edges, temporary stacks of goods, or narrow passages during inspections, it will perform actions such as deceleration, stopping, detouring, low-profile passage, waiting to avoid obstacles, or providing voice prompts, depending on the obstacle type, distance, and direction of movement.

[0023] The public warehouse map and task configuration module outputs a standardized electronic map, a basic inspection task to status acquisition module, an inspection mode adaptive switching module, a voice interaction module, and an anomaly recognition and verification module, providing a spatial benchmark for full-process path planning, point location, and area permission verification. The public warehouse map and task configuration module is used to maintain the public warehouse electronic map and generate inspection tasks. The electronic map includes area numbers, asset numbers, storage location numbers, passage boundaries, restricted areas, speed limit zones, borrowing areas, return areas, adjustment areas, inventory areas, temporary storage areas, charging points, environmental linkage points, and safe parking points. The public warehouse map and task configuration module generates inspection tasks based on mobile terminal configuration, public warehouse management system data, physical warehouse ledgers, virtual warehouse ledgers, or preset rules, and configures risk levels and asset importance levels for different areas.

[0024] The IoT device sensing module collects second-related sensor data (temperature and humidity, water immersion, smoke detection, light, gas, etc. in the warehouse) in real time, and simultaneously pushes environmental risk values ​​to the status acquisition module, anomaly identification and verification module, and environmental linkage control module to achieve full-domain sharing of environmental data; the IoT device sensing module is used to collect environmental data from temperature and humidity sensors, water immersion / leakage detectors, networked smoke / temperature detectors, light intensity sensors, and gas sensors, and outputs the environmental risk status to the anomaly identification and verification module and the environmental linkage control module.

[0025] The status acquisition module interfaces with the public asset warehouse map and task configuration module, the robot dog itself, the IoT device sensing module, the voice interaction module, and the anomaly recognition and verification module. It aggregates multi-dimensional discrete raw data including area risk, robot dog battery / position, network quality, obstacle congestion, asset lifecycle, ledger synchronization deviation, anomaly level, and voice command priority, and performs normalization processing. The normalized standardized parameters are then pushed unidirectionally to the inspection mode adaptive switching module. The status acquisition module periodically or event-triggeredly acquires data on the public asset warehouse area risk level, inspection task type, robot dog operating status, network status, obstacle status, environmental perception data, asset lifecycle status, physical / virtual warehouse synchronization status, and anomaly event level. The status acquisition module can also receive asset category, asset status, borrowing status, return status, transfer status, work plan, and asset approval information from the public asset warehouse management system.

[0026] The inspection mode adaptive switching module is used to construct a public warehouse state vector based on the data obtained by the state acquisition module, calculate the priority values ​​of multiple candidate inspection modes, and determine the target inspection mode after performing security verification on the path, power, communication status, pose stability and restricted area status related parameters corresponding to the inspection mode to be switched. The inspection mode adaptive switching module is the core decision-making hub of the system, and internally includes a state vector construction unit, a mode scoring unit, a mode arbitration unit, a security verification unit, and a policy generation unit, wherein: The system receives normalized parameters from the status acquisition module, structured business instructions from the voice interaction module, and abnormal event level signals from the anomaly identification and review module. The built-in status vector construction unit assembles and generates a public warehouse status vector based on the normalized parameters, and then generates a target inspection and collection execution strategy through pattern weighted scoring, preemption arbitration, and multi-dimensional security verification. First, the complete path, travel parameters, sensor configuration, point acquisition rules, and data feedback strategy are sent to the communication and logging module, which then forwards them to the robot dog for execution. Secondly, the anomaly identification threshold adjustment command corresponding to this inspection is synchronized to the anomaly identification and review module to dynamically adjust the anomaly judgment sensitivity; Third, synchronize the regional risk level to the environmental linkage control module and match the environmental response priority; fourth, when the preemptive task is triggered, store the breakpoint location and collected cache information in the communication and log module for retention.

[0027] The candidate inspection modes include at least three of the following: regular inspection mode, key asset enhanced inspection mode, asset entry verification mode, in-warehouse inventory mode, anomaly review inspection mode, real-time voice inspection mode, remote takeover mode, outbound review mode, transfer and escort mode, and safe return mode. The differentiated travel speed, sensor collection frequency, point dwell time, and data transmission priority configuration corresponding to different inspection modes are issued by the inspection mode adaptive switching module to the communication and log module for forwarding to the robot dog. At the same time, the inspection mode adaptive switching module synchronously sends the corresponding anomaly recognition threshold parameters to the anomaly recognition and review module to dynamically adjust the anomaly recognition sensitivity. The public warehouse status vector includes at least five of the following: current or target public warehouse area, area risk level, current task type, abnormal event level, abnormal duration, robot dog remaining battery power, network quality, obstacle density or path accessibility, personnel or forklift interference status, user command source, user command priority, asset lifecycle status, physical warehouse / virtual warehouse synchronization status, and environmental risk status.

[0028] The mode scoring unit performs weighted calculations on each state item in the public warehouse state vector according to the weight parameters corresponding to different candidate inspection modes, and combines the abnormal event level, voice command priority, asset life cycle status, physical warehouse / virtual warehouse synchronization status and environmental risk status to form the priority value of each candidate inspection mode.

[0029] Specifically, the state vector expression of the public warehouse is as follows: ; Where: A is the target area identifier, R is the area risk level, T is the current task type, E is the abnormal event level, D is the range of continuous impact of the abnormality, B is the remaining battery power of the robot dog, N is the network quality, O is the obstacle passage status, H is the human / forklift interference status, U is the voice command priority, L is the synchronization deviation of the physical and virtual warehouse ledgers, and C is the environmental risk level. The status acquisition module retrieves the public property warehouse map and the electronic map from the task configuration module, environmental data from the IoT device sensing module, and the robot dog's own operating data. After normalizing all discrete parameters, it outputs the data to the inspection mode adaptive switching module. The mode scoring unit is based on a weighted formula. ; Calculate the priority value of each mode. For the normalized status item, E represents the abnormal event level, U represents the voice command priority, L represents the ledger consistency penalty or synchronization deviation item, C represents the environmental risk item, and B represents the power consumption penalty item. This is a penalty for network instability. It represents the weighted coefficient corresponding to the k-th dimension state vector term under the i-th inspection mode; the priority value can be normalized to 0.0 to 1.0, the mode switching threshold T_switch can be set to 0.65 to 0.85, and the safe return threshold is higher than the abnormal review threshold.

[0030] The mode scoring unit scores the public warehouse state vector based on the weight parameters corresponding to the candidate inspection modes. Scoring can be implemented using weighted summation, rule tables, state machines, or machine learning models. For example, for the anomaly review inspection mode, the anomaly level, anomaly duration, and risk level of the area have higher weights; for the safe return mode, remaining battery power, path accessibility, and communication status have higher weights; for the voice real-time inspection mode, voice command priority, user permissions, and target area accessibility have higher weights; and for the asset entry verification mode and in-warehouse inventory mode, asset lifecycle status, ledger consistency, and target location accessibility have higher weights.

[0031] The mode arbitration unit determines the inspection mode to be switched based on priority values ​​and preset preemption rules. Preset preemption rules may include: security events take precedence over regular inspection tasks; high-level anomalies take precedence over low-level anomalies; real-time voice tasks that pass permission verification can preempt regular inspection tasks; remote takeover commands must meet user permission and communication quality requirements; enhanced inspections of key assets can adjust the collection frequency, dwell time, and travel speed of regular inspections without necessarily interrupting the current inspection task; asset warehousing, return, allocation, and inventory tasks can be inserted into the task queue according to preset priorities and preemption arbitration will be executed. Specifically, the preemption priority rule is: when at least one of the following security events is detected—low battery, hazardous gas, smoke, water immersion, water leakage, or unreachable path—the safety return mode preempts the current inspection mode. Before switching the inspection mode, the safety verification unit verifies at least three of the following: the robot dog's pose stability, the reachability of the target path, whether the path passes through a restricted area, whether the remaining battery power meets the requirements for task execution and return, whether the communication status of the target area meets the requirements for data transmission, and whether the synchronization status of the physical warehouse / virtual warehouse meets the task execution conditions. When the safety verification fails, the mode arbitration unit maintains the current inspection mode, enters a waiting avoidance state, requests confirmation from the mobile terminal, or enters a safe return mode.

[0032] The strategy generation unit is used to generate execution strategies based on the target inspection mode. The execution strategy includes the inspection path, target location, travel speed, dwell time, posture and actions, activated sensor type, acquisition angle, keyframe acquisition rules, data transmission priority, and voice feedback content.

[0033] The inspection mode adaptive switching module maintains the task queue and task breakpoint information; when the target inspection mode preempts the current inspection task, it records the completed inspection points, incomplete inspection points, current location and collected data of the current inspection task, and generates a breakpoint continuation inspection task after the preempted task is completed, the power is restored or the network is restored.

[0034] The voice interaction module receives the voice data stream uploaded by the mobile terminal, and combines it with the public warehouse map and the area, storage location, and asset basic information provided by the task configuration module to complete semantic parsing and permission verification (including remote takeover permission verification). After generating structured instructions, it outputs them in two ways: one way is written to the task queue and pushed to the inspection mode adaptive switching module to participate in mode preemption arbitration; the other way is to synchronize the user's identity and operation permissions to the communication and log module to retain the interaction record. If the voice information is missing key parameters, the voice interaction module sends voice follow-up instructions to the mobile terminal and the robot dog to complete the information. The voice interaction module parses the user's voice into a structured instruction containing at least four of the following: intent type, public warehouse area, target asset or storage location, execution action, sensor type, task priority, and data feedback requirements. When the structured instruction is missing the public warehouse area, target asset, storage location number, or execution action, the voice interaction module initiates voice follow-up through the mobile terminal or the robot dog and updates the structured instruction based on the user's supplementary voice. The voice interaction module performs permission verification before executing voice commands. The permission verification includes at least one of the following: user identity verification, area permission verification, operation type permission verification, secondary confirmation of dangerous actions, remote takeover permission verification, and asset transfer approval status verification.

[0035] The voice interaction module receives voice commands from mobile terminals or the robot itself, and performs speech recognition, semantic parsing, target object extraction, permission verification, and confirmation feedback. The module converts natural language commands into structured commands. These structured commands can include intent type, warehouse area, target asset or storage location, action to be performed, sensor type, task priority, and data feedback requirements. For example, if a user says "Go to storage location A3 to check for temperature and humidity anomalies" or "Move office equipment from area A to area B," the system will parse the command to obtain the intent type, target object, action to be performed, sensor type, task priority, and data feedback strategy.

[0036] If the voice command lacks a target area, target asset, storage location number, or action to be performed, the voice interaction module will follow up with voice questions via a mobile terminal or the robot itself. If the voice command involves remote takeover, entering a restricted area, stopping the current task, asset transfer, or inspection of a dangerous area, the system will perform user identity verification, area permission verification, operation type permission verification, secondary confirmation of dangerous actions, or verification of asset transfer approval status.

[0037] The anomaly identification and verification module receives visual / RF acquisition data from the robot dog itself, environmental sensing data output by the IoT device sensing module, and anomaly identification threshold adjustment instructions issued by the inspection mode adaptive switching module. It identifies various asset, channel, and environmental anomalies and determines the anomaly event level. The anomaly level is synchronously transmitted back to the inspection mode adaptive switching module to trigger mode preemption, and simultaneously pushed to the environmental linkage control module to initiate environmental handling and to the communication and log module to forward alarms generated by mobile terminals. After the anomaly verification task is completed, the verification results are reverse-synchronized to the public warehouse map and task configuration module to update the ledger status. The anomaly identification and verification module is used to identify at least one of the following anomaly events: asset tilting, pallet offset, shelf obstruction, aisle blockage, personnel intrusion, forklift obstruction, abnormal temperature and humidity, water immersion, water leakage, smoke, gas, lighting, inconsistency in asset ledgers, and cold chain temperature anomaly.

[0038] The anomaly identification and verification module determines the anomaly event level based on the anomaly type, location, duration, impact range, risk level of the area, sensor confidence level, and asset importance, and sends the anomaly event level to the inspection mode adaptive switching module and the environmental linkage control module. When the anomaly event is humidity anomaly, water immersion anomaly, water leakage anomaly, temperature anomaly, or smoke anomaly, the environmental linkage control module controls the intelligent dehumidifier / humidifier, exhaust equipment, or audible and visual alarm equipment to perform linkage actions.

[0039] When an anomaly level reaches the review threshold, the anomaly review inspection mode takes over the regular inspection mode or the key asset enhanced inspection mode. When a voice command that has passed permission verification is marked as an immediate task, the voice immediate inspection mode writes it into the task queue and executes it according to task priority (voice commands marked as immediate high priority can take over the task, while ordinary voice commands enter the queue). When asset entry, return, transfer, or inventory tasks are marked as high priority, the asset entry verification mode, exit review mode, transfer escort mode, or warehouse inventory mode participate in the arbitration in a preset order. When the anomaly level reaches the review threshold, the inspection mode adaptive switching module automatically generates an anomaly review task. The anomaly review task includes the anomaly location, review path, target acquisition angle, dwell time, activated sensor type, keyframe acquisition rules, voice prompt content, and data feedback strategy. After the robot dog executes the anomaly review task, it sends the review image, video, environmental data, and anomaly judgment result to the mobile terminal, and the user can confirm the processing result via voice or mobile terminal.

[0040] The anomaly verification task includes at least four of the following: anomaly location, verification path, target acquisition angle, dwell time, activated sensor type, keyframe acquisition rules, voice prompt content, and data feedback strategy.

[0041] The environmental linkage control module receives real-time environmental data from the IoT device sensing module, anomaly level signals from the anomaly identification and verification module, and regional risk levels synchronized by the inspection mode adaptive switching module. When the indicators exceed the threshold, it outputs device control commands to drive the dehumidifier / humidifier, exhaust fan, and related audible and visual alarm devices, and sends the linkage execution results back to the communication and log module for archiving.

[0042] The communication and logging module connects the mobile terminal, the robot dog itself, and all backend functional modules to complete command issuance, inspection data feedback, and anomaly alarm push. Under weak network conditions, it performs tiered caching and transmission of data, and synchronizes breakpoint inspection records, physical warehouse data collection data, and virtual warehouse ledgers after network recovery. Simultaneously, it collects all module interaction signals, mode switching records, voice operations, anomaly handling, and device linkage behaviors to generate an unalterable log. When network quality is below a preset threshold, the communication and logging module prioritizes inspection data according to the following transmission order: anomaly alarm data, robot dog status data, keyframe images, environmental sampling data, and ordinary video streams. After network recovery, it synchronizes any unuploaded inspection records, physical warehouse inspection results, and virtual warehouse ledger update records. The communication and logging module serves as the central hub for the entire system's interconnected transmission. It interfaces with all hardware and business modules to set up a hierarchical data transmission strategy. Under weak network conditions, the transmission priority from high to low is as follows: abnormal alarm messages, robot dog positioning status, abnormal key frame images, environmental sampling data, and ordinary inspection video streams. When the network is interrupted, it locally caches all inspection data uploaded by the robot dog and interaction commands from various modules. After the network is restored, it synchronizes physical warehouse collection data and virtual warehouse ledger update records to the public property warehouse map and task configuration module according to priority. Simultaneously, it collects all interaction signals, mode switching records, voice operations, abnormal handling, and device linkage behaviors between the public property warehouse map and task configuration module, status collection module, inspection mode adaptive switching module, voice interaction module, abnormal identification and verification module, IoT device sensing module, environmental linkage control module, robot dog, and mobile terminal in real time, and archives them uniformly to generate traceable inspection logs.

[0043] Specifically, the communication and logging module is used for data transmission between the mobile terminal, the robot dog itself, and the server. Communication methods can include Wi-Fi, 5G, Bluetooth, LAN, or a dedicated wireless network. When network quality is below a preset threshold, the communication and logging module performs tiered data transmission, prioritizing the uploading of abnormal alarms, robot dog status, location information, keyframe images, and ledger synchronization results, while temporarily suspending the uploading of ordinary video streams and non-critical environmental sampling data. After the network is restored, the system synchronizes the unuploaded data and adds incomplete inspection points to the breakpoint resume inspection task, while simultaneously synchronizing the physical warehouse inspection results and virtual warehouse ledger update records.

[0044] It is worth mentioning that the system's workflow is as follows: S1: System initialization, the public property warehouse map and task configuration module directly pushes the electronic map, asset ledger and regional risk benchmark data to the status acquisition module, inspection mode adaptive switching module, voice interaction module and anomaly recognition and review module to generate basic routine inspection tasks; S2: The status acquisition module directly connects to various business modules to aggregate multi-dimensional status data, constructs a public warehouse status vector, and pushes it unidirectionally to the inspection mode adaptive switching module. S3: The inspection mode adaptive switching module retrieves the public property warehouse map and task configuration module parameters to complete mode weighted scoring, preemption arbitration, and multi-dimensional security verification, and generates the target inspection collection strategy and corresponding anomaly identification threshold. The complete collection strategy is sent to the communication and log module, which forwards it to the robot dog for execution. The anomaly threshold parameters are directly synchronized to the anomaly identification and review module. S4: The robot dog completes the point inspection according to the collection strategy. The visual, environmental and radio frequency collection data are relayed back to the anomaly identification and verification module through the communication and log module. S5: The IoT device sensing module pushes environmental data directly to the anomaly identification and verification module and the environmental linkage control module in real time; the anomaly identification and verification module determines the anomaly level based on the received identification threshold, and the anomaly level is directly pushed to the inspection mode adaptive switching module to trigger mode preemption, and directly pushed to the environmental linkage control module to start device handling; anomaly alarms are pushed to the mobile terminal via the communication and log module; S6: Managers issue voice commands via mobile terminals. The data flows through the communication and log module to the voice interaction module. The voice interaction module directly connects to the public property warehouse map and task configuration module to verify asset and area information. If key parameters are missing, the communication and log module issues voice prompts. After the permission verification is passed, the structured command is directly pushed to the inspection mode adaptive switching module to participate in the arbitration. S7: In the event of low power, weak network, or high-priority task preemption, the inspection mode adaptive switching module will store the breakpoint location and data collection cache in the communication and log module; after the power / network is restored, the cached breakpoint will be read to generate a continued inspection strategy, which will be sent to the robot dog via the communication and log module; after the network is restored, the communication and log module will synchronize all cached data collection data to the public property warehouse map and task configuration module to update the virtual warehouse ledger. S8: All system module interactions, mode switching, voice operation, anomaly handling, and device linkage behaviors are uniformly collected and archived by the communication and log module to generate complete and traceable inspection business logs.

[0045] For further details, Example 1 is provided.

[0046] I. System Composition: In this embodiment, as Figure 1As shown, the system includes a mobile terminal 100, a robot dog 200, and a server or edge computing terminal 300. The public warehouse is divided into three zones: A, B, and C. Each zone has several shelves, aisles, and environmental control points. The system pre-creates an electronic map of the public warehouse, recording zone numbers, asset numbers, storage location numbers, aisle widths, restricted areas, speed-limited areas, borrowing areas, return areas, exchange areas, inventory areas, temporary storage areas, charging points, safe parking points, and environmental linkage points. Risk levels and asset importance levels are configured for different zones. The public warehouse map and task configuration module 310, status acquisition module 320, inspection mode adaptive switching module 330, voice interaction module 340, anomaly recognition and verification module 350, and communication and logging module 360 ​​are deployed on the server or edge computing terminal 300, or partially deployed on the robot dog 200 or mobile terminal 100 depending on network and computing power conditions.

[0047] The robot dog itself is equipped with a visible light camera, depth camera, LiDAR, temperature and humidity sensor, water immersion / leakage detector, network-connected smoke / temperature detector, light sensor, microphone, speaker, and wireless communication unit. A mobile terminal with an inspection application is installed to view the robot dog's real-time location, task list, abnormal events, video footage, asset movement status, and inspection reports.

[0048] The server or edge computing terminal runs the public property warehouse map and task configuration module, status acquisition module, inspection mode adaptive switching module, voice interaction module, anomaly recognition and verification module, and communication and logging module. In another implementation, the above modules can also be partially deployed on the robot dog itself or a mobile terminal. The IoT device sensing module and environmental linkage control module can be deployed independently on the server or edge, or they can be integrated into the anomaly recognition and verification module, communication and logging module, or public property warehouse map and task configuration module.

[0049] II. Adaptive switching of inspection mode: After the system starts up, the public property warehouse map and task configuration module generates routine inspection tasks according to the preset plan. The robot dog then passes through the inspection points in areas A, B, and C in sequence according to the routine inspection mode, and collects images and environmental data.

[0050] During the inspection, the status acquisition module periodically acquires public asset warehouse status data, and the inspection mode adaptive switching module constructs a public asset warehouse status vector S. If the robot dog enters a high-value asset area, and the risk level of this area is higher than that of ordinary areas, the priority value of the key asset enhanced inspection mode is increased. The system controls the robot dog to reduce its travel speed, increase the number of stopping points, increase the image acquisition frequency, and lower the anomaly recognition threshold.

[0051] If the anomaly identification and verification module detects a passageway blockage near cargo location A3, and the blockage lasts for more than a preset time, the anomaly level reaches the verification threshold. The mode arbitration unit determines that the anomaly verification inspection mode will preempt the regular inspection mode. After the safety verification unit confirms that the path from the current location to cargo location A3 is reachable, the remaining power is sufficient for verification and return, the target area is not in a restricted state, and the physical / virtual warehouse synchronization status is normal, the strategy generation unit generates the verification path, collection angle, dwell time, and keyframe feedback strategy from the current location to cargo location A3.

[0052] After the robot dog reaches the target location, it acquires images and videos at preset angles and sends abnormal locations, verification results, and keyframes to the mobile terminal. Once the verification task is completed, the system resumes the preempted routine inspection task based on the task breakpoint information.

[0053] If the robot dog's remaining battery power is below the safety threshold, and the current task cannot be completed within the remaining battery power, the priority of the safe return mode will be increased. The system records completed inspection points, incomplete inspection points, and collected data, controls the robot dog to return to the charging point, and generates a resume inspection task based on the incomplete inspection points after charging is completed.

[0054] III. Voice Command Generation Inspection Tasks: The user issues a voice command on their mobile device: "Go to storage location A3 to verify temperature and humidity abnormalities." The voice interaction module recognizes and semantically parses this voice command, resulting in the following structured instruction: Intent type: Exception review; Target object: Storage location A3; Action performed: Temperature and humidity verification; Enable sensors: temperature and humidity sensor and thermal imaging camera; Task priority: Temporary high priority; Data return method: Keyframes and temperature and humidity data are returned first.

[0055] The voice interaction module further verifies user permissions based on their account. If the user has both area A inspection and anomaly review permissions, the system displays confirmation information on the mobile terminal and announces via voice, "We are about to proceed to storage location A3 to review temperature and humidity anomalies." After user confirmation, the structured instructions are written to the task queue. The inspection mode adaptive switching module determines whether to immediately switch to the real-time voice inspection mode or execute it after the current high-priority task is completed, based on task priority and the current inspection status.

[0056] If the voice command is "transfer office equipment from area A to area B", the voice interaction module will further analyze the transfer object, target area, execution action and approval status, and generate a transfer escort task or prompt the user for approval and confirmation based on the ledger synchronization results.

[0057] IV. Abnormal closed-loop handling and environmental linkage: During the inspection and verification process, the robot dog detected that the temperature in cargo location A3 was higher than a preset threshold and acquired a corresponding thermal imaging image. The anomaly identification and verification module marked this anomaly as a temperature anomaly and determined the anomaly level based on the temperature value, duration, risk level of the area, asset importance, and thermal imaging confidence level.

[0058] The communication and logging module sends anomaly alarms, thermal imaging keyframes, visible light images, location coordinates, anomaly level, asset number, and robot dog status to the mobile terminal. The mobile terminal displays anomaly details and provides options such as "Confirm Anomaly," "False Alarm," "Review Again," "Transfer to Manual Processing," "Confirm Return," "Confirm Allocation," and "Mark as Processed." Users can also say "Review A3 Storage Location Again" or "Mark as Processed" via voice. The system updates the anomaly handling status based on voice commands and permission verification results.

[0059] When the IoT device's sensing module detects a continuous increase in humidity or a water leak, the environmental linkage control module automatically controls the smart dehumidifier / humidifier, exhaust equipment, and audible and visual alarm devices to work together, and writes the linkage results into the log to achieve a closed loop of environmental anomaly detection, linkage handling, verification, and confirmation.

[0060] From discovery, review, alarm, user confirmation to processing results, all abnormal events are recorded by the communication and logging module, forming a traceable inspection log.

[0061] V. Data backhaul in weak network environments: When the robot dog enters the depths of the public storage area, causing a decline in network quality, the communication and logging module detects that the network quality is below a preset threshold. The system then enters a weak network transmission strategy, prioritizing the uploading of abnormal alarm data, robot dog location, battery level, task status, asset synchronization status, and keyframe images, while reducing the bitrate of the regular video stream or suspending the uploading of the regular video stream.

[0062] If the network is completely interrupted, the robot dog continues to cache inspection data locally and continues to execute inspection points that can be safely completed based on the local map and issued tasks. When the network is restored, the communication and logging module synchronizes the cached data according to time order and data priority. For inspection points that were not completed during the network interruption, the system generates a resume inspection task and sends it to the mobile terminal, while simultaneously synchronizing the physical warehouse inspection results and the virtual warehouse ledger update results.

[0063] VI. Scenarios of borrowing, returning, adjusting, and inventorying: In public warehouse borrowing or returning scenarios, users initiate asset entry and exit applications via mobile terminals. The voice interaction module parses commands such as "Return office equipment in Area A" and "Relocate assets from storage location A3 to Area B" into structured tasks. The public warehouse map and task configuration module generates entry verification, exit review, or transfer escort tasks based on asset number, storage location number, and approval status, and controls the robot dog to go to the corresponding storage location to scan, photograph, conduct environmental checks, and verify ledger records.

[0064] In inventory scenarios, the robot dog inspects each storage location in a designated area according to inventory mode, collecting asset numbers, locations, statuses, and storage location occupancy information, and comparing this data with the virtual warehouse ledger. When inconsistencies are found in the asset ledger, inconsistent storage location labels, or obscured goods, the system automatically triggers an anomaly review and notifies management personnel.

[0065] It is worth mentioning that the comparative example of this invention uses a fixed-route inspection robot combined with an independent voice remote control terminal, which can only perform inspections and issue simple action commands according to a preset route. Although this solution can provide basic alarms, manual selection of a verification route is still required after an anomaly occurs, and it cannot automatically complete mode preemption and breakpoint resumption of inspection based on anomaly level, asset lifecycle status, physical / virtual warehouse synchronization status, environmental risks, power consumption, and network quality.

[0066] Compared to the comparative example, the present invention forms a closed loop by linking state vectors, pattern scoring, preemption arbitration, security verification, environmental linkage, and structured voice commands to switch inspection modes, review anomalies, verify asset transfers, and restore tasks, making it more suitable for the complex environment of public warehouses.

[0067] This invention can be applied to scenarios such as public property warehouses in administrative and public institutions, asset warehouses in public organizations, temporary turnover warehouses, warehouses for conference and event support, shared warehouses, and asset management centers. Through robot dog inspection, mobile terminal interaction, voice command parsing, adaptive switching of inspection modes, environmental perception, environmental linkage control, anomaly verification, and breakpoint resumption of inspection, this invention can improve the automation level of public property warehouse inspections, reduce manual inspection costs, and enhance the efficiency of asset transfer verification, anomaly detection, and handling, demonstrating good industrial applicability.

[0068] It is worth mentioning that the technical features such as the robot dog involved in this patent application should be regarded as prior art. The specific structure, working principle, and possible control methods and spatial arrangement of these technical features can be adopted using conventional choices in the field, and should not be regarded as the inventive point of this patent. This patent will not be further elaborated in detail.

[0069] For those skilled in the art, modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should be included within the protection scope of this invention.

Claims

1. A robot dog inspection mode adaptive switching and voice interaction system for public storage warehouses, characterized in that, It includes a mobile terminal, the robot dog itself, a public property warehouse map and task configuration module, a status acquisition module, an inspection mode adaptive switching module, a voice interaction module, an anomaly recognition and verification module, an IoT device sensing module, an environmental linkage control module, and a communication and logging module, among which: The mobile terminal is bidirectionally connected to the voice interaction module and the communication and log module, and is used to issue voice / touch commands, receive inspection feedback, abnormal alarms and ledger synchronization results; the robot dog body communicates bidirectionally with all back-end business modules through the communication and log module, performs inspection actions and transmits back the first relevant sensor data. The public property warehouse map and task configuration module respectively output a standardized electronic map, a basic inspection task to status collection module, an inspection mode adaptive switching module, a voice interaction module, and an anomaly recognition and verification module, providing a spatial benchmark for full-process path planning, point location, and area permission verification. The IoT device sensing module collects second relevant sensor data in real time and pushes environmental risk values ​​to the status acquisition module, anomaly identification and verification module, and environmental linkage control module simultaneously to achieve full-domain sharing of environmental data. The status acquisition module connects to the public property warehouse map and task configuration module, the robot dog itself, the IoT device sensing module, the voice interaction module, and the anomaly recognition and verification module. It aggregates multi-dimensional discrete raw data and completes normalization processing, and pushes the normalized standardized parameters one-way to the inspection mode adaptive switching module. The inspection mode adaptive switching module is used to construct a public warehouse state vector based on the data obtained by the state acquisition module, calculate the priority values ​​of multiple candidate inspection modes, and determine the target inspection mode after performing security verification on the relevant parameters corresponding to the inspection mode to be switched. The voice interaction module receives the voice data stream uploaded by the mobile terminal, combines the public warehouse map with the relevant information provided by the task configuration module to complete semantic parsing and permission verification, and generates structured instructions which are then output in two channels. The anomaly identification and verification module receives the collected data transmitted back by the robot dog, the environmental sensing data output by the IoT device sensing module, and the anomaly identification threshold adjustment command issued by the inspection mode adaptive switching module. It identifies various anomalies and determines the level of the anomaly event. The anomaly level is synchronously transmitted back to the inspection mode adaptive switching module to trigger mode preemption. At the same time, it is pushed to the environmental linkage control module to start environmental handling and to the communication and log module to forward the alarm generated by the mobile terminal. After the anomaly verification task is completed, the verification result is reverse-synchronized to the public property warehouse map and task configuration module to update the ledger status. The environmental linkage control module receives real-time environmental data from the IoT device sensing module, anomaly level signals from the anomaly identification and verification module, and regional risk levels synchronized by the inspection mode adaptive switching module. When the indicators exceed the threshold, it outputs device control commands to drive relevant devices and sends the linkage execution results back to the communication and log module for archiving.

2. The adaptive switching and voice interaction system for robot dog inspection modes in a public warehouse according to claim 1, characterized in that, The inspection mode adaptive switching module includes a state vector construction unit, a mode scoring unit, a mode arbitration unit, a security verification unit, and a strategy generation unit. The inspection mode adaptive switching module receives normalized parameters output by the state acquisition module, structured business instructions issued by the voice interaction module, and abnormal event level signals returned by the abnormal identification and review module. The public warehouse state vector is generated by assembling the built-in state vector construction unit based on normalized parameters, and then the target inspection and collection execution strategy is generated through pattern weighted scoring, preemption arbitration, and multi-dimensional security verification.

3. The adaptive switching and voice interaction system for robot dog inspection modes in a public warehouse according to claim 2, characterized in that, The candidate inspection modes include at least three of the following: regular inspection mode, key asset enhanced inspection mode, asset entry verification mode, in-warehouse inventory mode, anomaly review inspection mode, real-time voice inspection mode, remote takeover mode, outbound review mode, transfer and escort mode, and safe return mode. The differentiated travel speed, sensor collection frequency, point dwell time, and data transmission priority configuration corresponding to different inspection modes are issued by the inspection mode adaptive switching module to the communication and log module for forwarding to the robot dog. At the same time, the inspection mode adaptive switching module synchronously sends the corresponding anomaly recognition threshold parameters to the anomaly recognition and review module to dynamically adjust the anomaly recognition sensitivity. The public warehouse status vector includes at least five of the following: current or target public warehouse area, area risk level, current task type, abnormal event level, abnormal duration, robot dog remaining battery power, network quality, obstacle density or path accessibility, personnel or forklift interference status, user command source, user command priority, asset lifecycle status, physical warehouse / virtual warehouse synchronization status, and environmental risk status.

4. The adaptive switching and voice interaction system for robot dog inspection modes in a public warehouse according to claim 3, characterized in that, The mode scoring unit performs weighted calculations on each state item in the public warehouse state vector according to the weight parameters corresponding to different candidate inspection modes, and combines the abnormal event level, voice command priority, asset life cycle status, physical warehouse / virtual warehouse synchronization status and environmental risk status to form the priority value of each candidate inspection mode.

5. The adaptive switching and voice interaction system for robot dog inspection modes in a public warehouse according to claim 4, characterized in that, The mode arbitration unit has a preset preemption priority rule, which is: when at least one security event is detected, the security return mode preempts the current inspection mode. When the level of an abnormal event reaches the review threshold, the abnormal review inspection mode will take over the regular inspection mode or the key asset enhanced inspection mode. When a voice command that has passed the permission verification is marked as an immediate task, the voice immediate inspection mode writes it into the task queue and executes it in a preemptive or queued manner according to the task priority. When asset entry, return, transfer, or inventory tasks are marked as high priority, the asset entry verification mode, exit review mode, transfer escort mode, or inventory mode will participate in the arbitration in a preset order.

6. The adaptive switching and voice interaction system for robot dog inspection modes in a public warehouse according to claim 5, characterized in that, Before switching the inspection mode, the safety verification unit verifies at least three of the following: the robot dog's pose stability, the reachability of the target path, whether the path passes through a restricted area, whether the remaining power is sufficient for task execution and return, whether the communication status of the target area is sufficient for data transmission, and whether the synchronization status of the physical warehouse / virtual warehouse is sufficient for task execution. When the security verification fails, the mode arbitration unit maintains the current inspection mode, enters the waiting avoidance state, requests confirmation from the mobile terminal, or enters the safe return mode.

7. The adaptive switching and voice interaction system for robot dog inspection modes in a public warehouse according to claim 6, characterized in that, The inspection mode adaptive switching module maintains the task queue and task breakpoint information; When the target inspection mode preempts the current inspection task, it records the completed inspection points, incomplete inspection points, current location, and collected data of the current inspection task, and generates a resume inspection task after the preempted task is completed, the power is restored, or the network is restored.

8. The adaptive switching and voice interaction system for robot dog inspection modes in a public warehouse according to claim 1, characterized in that, The communication and logging module connects the mobile terminal, the robot dog itself, and all backend functional modules to complete command issuance, inspection data feedback, and abnormal alarm push. Under weak network conditions, data is transmitted in a hierarchical cache. After the network is restored, the interrupted inspection records, physical warehouse data collection data, and virtual warehouse ledger are synchronized. At the same time, all module interaction signals, mode switching records, voice operations, abnormal handling, and equipment linkage behaviors are collected to generate an unalterable log.

9. A robot dog inspection mode adaptive switching and voice interaction system for public warehouses according to claim 1, characterized in that, The mobile terminal is also used to display images, videos, locations, anomaly levels, review results, and processing status of abnormal events, and to receive user input via touch or voice to confirm anomalies, false alarms, re-review, transfer to manual processing, mark as processed, confirm reassignment, or confirm return.

10. A robot dog inspection mode adaptive switching and voice interaction system for public warehouses according to claim 7, characterized in that, The system's workflow is as follows: S1: System initialization, the public property warehouse map and task configuration module directly pushes the electronic map, asset ledger and regional risk benchmark data to the status acquisition module, inspection mode adaptive switching module, voice interaction module and anomaly recognition and review module to generate basic routine inspection tasks; S2: The status acquisition module directly connects to various business modules to aggregate multi-dimensional status data, constructs a public warehouse status vector, and pushes it unidirectionally to the inspection mode adaptive switching module. S3: The inspection mode adaptive switching module retrieves the public property warehouse map and task configuration module parameters to complete mode weighted scoring, preemption arbitration, and multi-dimensional security verification, and generates the target inspection collection strategy and corresponding anomaly identification threshold; the complete collection strategy is sent to the communication and log module, which forwards it to the robot dog itself for execution. The abnormal threshold parameters are directly synchronized to the abnormal identification and verification module; S4: The robot dog completes the point inspection according to the collection strategy. The visual, environmental and radio frequency collection data are relayed back to the anomaly identification and verification module through the communication and log module. S5: The IoT device sensing module pushes environmental data directly to the anomaly identification and verification module and the environmental linkage control module in real time; the anomaly identification and verification module determines the anomaly level based on the received identification threshold, and the anomaly level is directly pushed to the inspection mode adaptive switching module to trigger mode preemption, and directly pushed to the environmental linkage control module to start device handling; anomaly alarms are pushed to the mobile terminal via the communication and log module; S6: Managers issue voice commands via mobile terminals. The data flows through the communication and log module to the voice interaction module. The voice interaction module directly connects to the public property warehouse map and task configuration module to verify asset and area information. If key parameters are missing, the communication and log module issues voice prompts. After the permission verification is passed, the structured command is directly pushed to the inspection mode adaptive switching module to participate in the arbitration. S7: In the event of low power, weak network, or high-priority task preemption, the inspection mode adaptive switching module will store the breakpoint location and data collection cache in the communication and log module; after the power / network is restored, the cached breakpoint will be read to generate a continued inspection strategy, which will be sent to the robot dog via the communication and log module; after the network is restored, the communication and log module will synchronize all cached data collection data to the public property warehouse map and task configuration module to update the virtual warehouse ledger. S8: All system module interactions, mode switching, voice operation, anomaly handling, and device linkage behaviors are uniformly collected and archived by the communication and log module to generate complete and traceable inspection business logs.

Images