Inspection method and inspection platform of inspection robot for multilayer caged chicken house

Abstract

The invention provides a routing inspection method and a routing inspection platform of a routing inspection robot for a multilayer caged chicken house. The routing inspection method comprises the following steps: S1, constructing an SLAM map and planning a motion track; s2, setting a queuing rule of the inspection robot and an activity rule of the camera group; s3, setting a coop number calculation model and a layer number mapping rule of the coop; and S4, transmitting one copy of the data acquired by the inspection robot to the cloud, and using the copy of the data for local backup. According to the inspection method and the inspection platform of the inspection robot for the multilayer caged chicken house, the problems of low efficiency, inaccurate positioning, inconvenience in multi-terminal interaction and the like of traditional inspection are solved, and the automation, precision and intelligent level of monitoring dead chickens due to diseases is remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of poultry breeding monitoring, and in particular relates to an inspection method and an inspection platform of an inspection robot for multi-layer caged chicken houses. Background Art

[0002] In modern poultry farming, especially in multi-layer caged chicken houses, timely and accurate detection of dead chickens is crucial for disease prevention and control, as well as ensuring farming profitability. However, traditional manual inspection methods are inefficient, prone to omissions, and difficult to operate in multi-layer cage structures. Some existing automated monitoring methods also suffer from incomplete data collection, inaccurate positioning, and inconvenient multi-terminal interaction, making them unable to meet the needs of efficient and accurate inspection and monitoring of dead chickens. Summary of the Invention

[0003] In view of this, the present invention aims to solve one of the technical problems in the related art to at least a certain extent. To this end, the present invention proposes an inspection method for an inspection robot for multi-layer caged chicken houses, so as to achieve efficient and accurate inspection and monitoring of sick and dead chickens in multi-layer caged chicken houses and convenient multi-terminal interactive control.

[0004] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0005] A patrol inspection method for a multi-layer cage chicken house using a patrol inspection robot comprises the following steps:

[0006] S1, build SLAM map and plan motion trajectory;

[0007] S2. Set the numbering rules for the inspection robots and the activity rules for the camera groups;

[0008] S3. Setting the cage number calculation model and layer mapping rules of the chicken cage;

[0009] S4. A copy of the data collected by the inspection robot is transmitted to the cloud, and the local VVR saves a copy to support fast playback or offline analysis. When a key event is detected, the inspection robot's system automatically captures the relevant frame image and stores it in conjunction with the environmental data and cage number information to facilitate subsequent tracing and report generation.

[0010] Furthermore, the method for setting the numbering rules of the inspection robots and the activity rules of the camera groups in S2 is:

[0011] S21. Define the global coordinate system

[0012] Formulate the rules for the numbering of chicken cages: the row numbers increase from west to east: row 1 is on the westernmost side, row R is on the easternmost side; the aisle number A increases from west to east: A=1 is the westernmost aisle, A=Amax is the easternmost aisle;

[0013] Specify the mapping direction of the inspection robot: NORTH: global coordinate system north direction Y+; SOUTH:

[0014] Global coordinate system south direction Y-;

[0015] The camera groups on the left side of the robot are numbered 1-4, and the right side of the camera groups on the right side of the robot are numbered 5-8;

[0016] S22. Based on the global coordinate system, determine the row number according to the aisle number A and the inspection direction (NORTH, SOUTH) uploaded by the SLAM chassis.

[0017] Furthermore, in S22, the specific method for determining the row number based on the global coordinate system and the aisle number A uploaded by the SLAM chassis and the inspection direction (NORTH, SOUTH) is as follows:

[0018] S221. Generate number function:

[0019]

[0020] Among them, A represents the aisle number uploaded by the SLAM chassis, Amax represents the last inspected aisle, and D represents the robot's inspection direction. The inspection direction is used to determine the robot's inspection direction and calculate the row number. NORTH means the robot inspects in the north direction, and SOUTH means the robot inspects in the south direction. It means that there is no inspection row to the right of the robot, and R is the last row of chicken houses;

[0021] S222. Setting camera activation rules

[0022]

[0023] Among them, A represents the aisle number uploaded by the SLAM chassis, D represents the robot's inspection direction, NORTH means the robot is inspecting in the north direction, and SOUTH means the robot is inspecting in the south direction.

[0024] Furthermore, the method for setting the cage number calculation model and layer mapping rules of the chicken cage in S3 is:

[0025] S31. Set SLAM points and define parameters

[0026] Cage width: W (the width of each cage in the X-axis direction);

[0027] Starting point: (x_start, y_start);

[0028] Current position: (x_current, y_current);

[0029] S32, calculation cage number model

[0030] Calculate the distance based on the uploaded start point (x_start, y_start) and the current inspection position (x_current, y_current), and calculate the cage number formula:

[0031]

[0032] S33. Set layer mapping rules

[0033] The camera number directly corresponds to the layer number:

[0034] Left camera No. 1-4: monitors the 1st to 4th floors of the left cage row;

[0035] Right cameras 5-8: monitor the 1st to 4th floors of the right cage row;

[0036] Physical meaning: Each cage column is divided into 4 layers in the vertical direction, and the robot monitors different layers through different cameras.

[0037] Compared with the prior art, the inspection method of the inspection robot for multi-layer cage chicken houses described in the present invention has the following advantages:

[0038] (1) The inspection method of the inspection robot for multi-layer caged chicken houses described in the present invention is that the inspection robot is equipped with a multimodal sensor group, an integrated environmental sensor, a dual-light camera and an infrared thermal imager, and collects chicken house environmental data, chicken images, sounds and body temperature information in real time; the data uploads environmental parameters through MQTT and LORA, and large-capacity data such as images are transmitted to the edge computing server via the streaming media protocol, which solves the problems of low efficiency, inaccurate positioning and inconvenience of multi-terminal interaction of traditional inspections, and significantly improves the automation, precision and intelligence level of diseased and dead chicken monitoring.

[0039] (2) The inspection method of the inspection robot for multi-layer caged chicken houses described in the present invention, the inspection robot and its supporting equipment build an efficient intranet data collection and transmission system to ensure the stable operation of environmental monitoring, video processing and remote collaborative work.

[0040] Another object of the present invention is to propose an inspection platform for an inspection robot in a multi-layer cage chicken house, so as to realize efficient and accurate inspection and monitoring of sick and dead chickens in the multi-layer cage chicken house and convenient multi-terminal interactive control.

[0041] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0042] An inspection platform for an inspection robot for multi-layer caged chicken houses, and an inspection method for an inspection robot for multi-layer caged chicken houses using the inspection platform, comprising:

[0043] Inspection robots, used to collect environmental data, images, sounds, and temperature data within the chicken house;

[0044] The edge computing server is used to perform preliminary processing and analysis on the data collected and uploaded by the inspection robot, and quickly screen out the relevant data information of suspected dead chickens.

[0045] Network equipment, used to build communication links between inspection robots and edge computing servers, and between servers and terminals;

[0046] Among them, the environmental data is uploaded through MQTT and LORA communication, and the image, sound and temperature data are uploaded through HTTP request.

[0047] Furthermore, the inspection robot includes a chassis and a multimodal sensor group installed above it, a lifting platform and a control cabinet. The SLAM function is integrated inside the chassis to form a SLAM chassis, which can construct a 2D grid map with a resolution of 5 cm in real time, and support dynamic obstacle avoidance and path planning. The multimodal sensor group is used to collect environmental data, images, sounds and temperature data in the chicken house and transmit them to the controller in the control cabinet. The lifting platform is a three-section telescopic square tube with a vertical travel range of 1.4 to 2.6 meters, which accurately adapts to the height distribution of the four-layer cages in the chicken house.

[0048] Furthermore, the inspection robot establishes a star connection structure with the edge computing server through the network device, and each terminal device is connected to the server step by step through the network device to establish a tree branch structure, thereby forming a topological structure combining star and tree structures with the edge computing server as the core, and the inspection robot, network equipment and each terminal device.

[0049] Furthermore, corresponding intranet penetration software or services are deployed on the server side, and intranet penetration technology is used to map the internal server port to an externally accessible public network address, so that external mobile and desktop devices can communicate and interact with the platform just like in a local area network without the need for complex network configuration.

[0050] Furthermore, the platform performs hierarchical management according to the urgency of the tasks and forms a dynamic priority queue, and the order of the priorities can be dynamically adjusted.

[0051] Furthermore, the platform also includes a multi-terminal command execution framework, and the multi-terminals include mobile terminals, desktop terminals, and local robot terminals. The command execution method is as follows:

[0052] T1, command reception phase: Each terminal device sends the command in a set format to the edge computing server through a preset user interface; the local robot generates the corresponding command based on its own preset automation rules or sensor trigger mechanism and sends it to the server;

[0053] T2, command judgment stage: After receiving the command, the edge computing server judges the legality and validity of the command;

[0054] T3, command queueing stage: After judging the legality and validity of commands, they are added to the queue of pending commands in the order of receipt or according to the priority of the commands to ensure that the commands can be executed in an orderly manner;

[0055] T4, Command Feedback Stage: When a command starts or completes execution, the edge computing server feeds back the corresponding execution status information to the command sender, allowing operators to promptly understand the execution status of the command and make subsequent operational decisions.

[0056] T5, command execution phase: According to the queue order, the edge computing server parses the command and sends it to the corresponding execution component, which completes the specific command operation. During the execution process, the execution component can also provide real-time feedback on execution progress and other information to the server based on actual conditions.

[0057] Compared with the prior art, the inspection platform of the inspection robot for multi-layer cage chicken houses described in the present invention has the following advantages:

[0058] (1) The inspection platform of the inspection robot for multi-layer caged chicken houses described in the present invention has a network topology with edge servers as the core, constructing a star-tree hybrid structure, supporting intranet penetration to achieve remote access; the multi-terminal command execution framework covers mobile terminals, desktop terminals and local robot terminals, and manages the entire process of command reception, verification, execution and feedback through dynamic priority queues; in addition, the video stream is collected through the RTSP protocol and optimized for transmission through H.264 encoding, and the AI model is combined to realize the detection of dead chickens, which solves the problems of low efficiency, inaccurate positioning and inconvenience of multi-terminal interaction in traditional inspections, and significantly improves the automation, precision and intelligence level of dead chicken monitoring.

[0059] (2) The inspection platform of the inspection robot for multi-layer caged chicken houses described in the present invention not only realizes efficient and reliable full-link collaboration, but also reserves sufficient space for the subsequent introduction of higher-level communication encryption, multi-path redundancy and service quality optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] The accompanying drawings, which constitute part of the present invention, are provided to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are provided to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings:

[0061] Figure 1 This is a functional block diagram of an inspection platform of an inspection robot for multi-layer caged chicken houses according to an embodiment of the present invention;

[0062] Figure 2 This is a schematic diagram of the state changes of the inspection robot according to an embodiment of the present invention;

[0063] Figure 3 This is a principle block diagram of the inspection robot according to an embodiment of the present invention;

[0064] Figure 4 This is a dead chicken image transmission display diagram according to an embodiment of the present invention;

[0065] Figure 5 The network topology and data transmission architecture described in the embodiment of the present invention;

[0066] Figure 6 This is a schematic diagram of the chicken house SLAM map and row number calculation according to an embodiment of the present invention;

[0067] Figure 7 This is a flowchart of video stream processing and transmission according to an embodiment of the present invention.

[0068] Description of reference numerals:

[0069] 1- Chassis; 2- Multimodal sensor group; 3- Lifting platform; 4- Control cabinet. DETAILED DESCRIPTION

[0070] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0071] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, features defined as "first", "second", etc. may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0072] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0073] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0074] Glossary:

[0075] MQTT: Message Queuing Telemetry Transport, a lightweight IoT communication protocol based on the publish / subscribe model.

[0076] LORA: The full name is Long Range, which is a long-distance wireless communication technology based on spread spectrum technology;

[0077] HTTP: The full name is HyperText Transfer Protocol, which is an important basic protocol for information transmission and interaction on the Internet.

[0078] SLAM: The full name is Simultaneous Localization and Mapping, which means simultaneous positioning and map construction.

[0079] IMU: The full name is Inertial Measurement Unit, which is a sensor device used to measure the motion state of an object.

[0080] FRP: Fast Reverse Proxy. It is an open-source, high-performance reverse proxy application that allows devices on the intranet or behind a firewall to provide services to the external network in a public network environment.

[0081] ROS2: The full name is Robot Operating System 2, which is the next generation version of ROS and provides a software framework and tool set for robot application development.

[0082] GUI: Graphical User Interface, a graphical user interface, is an important tool for interacting with and visualizing robotic systems.

[0083] JSON: JavaScript Object Notation, is a lightweight data exchange format that is easy for humans to read and write, and also easy for machines to parse and generate.

[0084] Server: server.

[0085] Action: Action is an important communication mechanism that is suitable for tasks that need to be executed for a period of time, can be canceled, and require feedback on the execution status.

[0086] Inspection method of inspection robot for multi-layer cage chicken house, such as Figures 1 to 7 As shown, the following steps are included:

[0087] S1. Build SLAM map and plan motion trajectory

[0088] like Figure 5 As shown, in the initial stage, the inspection robot scans the chicken house to generate a 2D grid map, plans the inspection path, and presets the chicken cage spacing parameter (S=1.2m).

[0089] Dynamic positioning correction: When the inspection robot moves, it combines IMU data with multi-line lidar points for cloud matching to correct position deviations in real time (accuracy ±2cm) to ensure the accuracy of number calculation.

[0090] S2. Set the numbering rules for the inspection robots and the activity rules for the camera groups

[0091] The row number is determined based on the aisle number A and the inspection direction (NORTH, SOUTH) uploaded by the SLAM chassis.

[0092] S21. Define the global coordinate system

[0093] like Figure 5 As shown, the rules for cage numbering are formulated: the row numbers increase from west to east: the 1st row is on the westernmost side, and the Rth row is on the easternmost side; the aisle number A increases from west to east: A=1 is the westernmost aisle, and A=A_max is the easternmost aisle.

[0094] Specify the mapping direction of the inspection robot:

[0095] NORTH: The north direction of the global coordinate system (Y+); SOUTH: The south direction of the global coordinate system (Y-).

[0096] Table 1. Schematic diagram of inspection robot inspection routes and effective cameras.

[0097]

[0098]

[0099] S22. Generate row number coordinates based on the global coordinate system

[0100] S221. Determine the row number of the chicken cage according to the aisle number and inspection direction uploaded by the inspection robot;

[0101] Number generation function:

[0102]

[0103] Among them, A represents the aisle number uploaded by the SLAM chassis, Amax represents the last inspected aisle, and D represents the robot's inspection direction. The inspection direction is used to determine the inspection robot's orientation during inspection and calculate the row number.

[0104] NORTH means the robot is patrolling in the north direction, and SOUTH means the robot is patrolling in the south direction. It means there is no inspection row to the right of the robot, and R is the last row of chicken houses.

[0105] like Figure 5 As shown, Example 1:

[0106] When the robot is in aisle 1, A=1. At this time, the robot's inspection direction D is NORTH. The robot only inspects the chicken coop on the left, and the right side is against the wall. Therefore, the robot's inspection number on the left side is 1, and there is no inspection on the right side (against the wall). When the robot enters aisle 2 according to the training path, A=2. At this time, the robot's inspection direction D is SOUTH. Therefore, the robot's inspection number on the left side is 2, and the right side is 3. When the robot enters aisle 3 according to the inspection path, A=3. At this time, the robot's inspection direction D is NORTH. The robot's inspection number on the left side is 5, and the right side is 4. When the robot enters the last aisle, A=Amax=4. At this time, the inspection direction D is NORTH. The robot's inspection number on the left side is R, and there is no inspection on the right side (against the wall).

[0107] S222: Set the camera activation rules according to Table 1

[0108]

[0109] Among them, A represents the aisle number uploaded by the SLAM chassis, D represents the robot's inspection direction, NORTH means the robot is inspecting in the north direction, and SOUTH means the robot is inspecting in the south direction.

[0110] like Figure 5As shown, when the robot is in aisle 1, A = 1, and the inspection direction D is NORTH, only the robot's left camera, numbered 1-4, is turned on. The right camera is against the wall, and the camera is disabled. When the robot is in the last aisle, the inspection direction D is SOUTH, and only the robot's left camera, numbered 1-4, is turned on. The right camera is against the wall, and the camera is disabled. When the robot is in aisle 1 or any other aisle number other than the maximum aisle number, since chicken coops are on both sides of the robot's camera, cameras 1-4 on the left and cameras 5-8 on the right are turned on simultaneously.

[0111] S3. Set the cage number calculation model and layer mapping rules for the chicken cage

[0112] S31. Set SLAM points and define parameters

[0113] Cage width: W (width of each cage in the X-axis direction)

[0114] Starting point: (x_start, y_start)

[0115] Current position: (x_current, y_current)

[0116] S32, calculation cage number model

[0117] Calculate the distance between the uploaded start point (x_start, y_start) and the current inspection position (x_current, y_current), divide it by the width W of a single breeding cage, and round it up.

[0118] Cage number calculation formula:

[0119]

[0120] like Figure 5 As shown in the figure, there may be a deviation between the actual farm coordinates and the SLAM dynamic positioning corrected coordinates, so this algorithm is used to calculate the point information uploaded by the SLAM chassis 1.

[0121] S33. Set layer mapping rules

[0122] The camera number directly corresponds to the layer number:

[0123] Left camera (No. 1-4): monitors the 1st to 4th floor of the left cage row

[0124] Right camera (No. 5-8): monitors the 1st to 4th floor of the right cage row

[0125] Physical meaning: Each cage is divided into 4 layers in the vertical direction (Z axis), and the robot monitors different layers through different cameras.

[0126] S4. Transmit and process data collected by the inspection robot

[0127] like Figure 6 As shown, this solution uses the RTSP protocol to enable multiple P cameras to collect raw video streams, and effectively reduces bandwidth usage through H.264 encoding compression. It also uses the GStreamer pipeline to perform optimization processing such as decoding, filtering, and format conversion on the video. At the same time, it combines OpenCV to extract key frames for real-time target detection and frame compression to reduce the amount of single-frame data. In terms of network transmission, the video stream penetrates through the first and second level NAT (with the help of STUN / TURN servers or port mapping to achieve connectivity under multi-level NAT), and is encrypted by the HTTPS client and transmitted to the cloud in the form of HTTP requests. The cloud server is responsible for data storage, AI analysis, and visualization, while the local VVR saves video copies to support fast playback or offline analysis, thereby achieving a balance and optimization of the overall system in terms of performance, latency, and security. When a key event is detected (such as a sick chicken alarm), the system automatically captures the relevant frame images (such as Figure 3 The image of a dead chicken is transmitted back as shown in the example above) and is stored together with the environmental data and cage number information to facilitate subsequent traceability and report generation.

[0128] Inspection platform for inspection robots in multi-layer cage chicken houses, such as Figures 1 to 7 Shown, including

[0129] Inspection robots are used to collect multiple data such as environmental data, images, sounds, and temperature in the chicken house;

[0130] The edge computing server is used to perform preliminary processing and analysis on the data collected and uploaded by the inspection robot, and quickly screen out the relevant data information of suspected dead chickens.

[0131] Network equipment is used to establish communication links between inspection robots and edge computing servers, and between servers and terminals, ensuring stable and efficient data transmission between all components of the entire platform.

[0132] Among them, the environmental data is uploaded through MQTT and LORA communication, and the image, sound and temperature data are uploaded through HTTP request.

[0133] The data transmission methods of network devices include MQTT communication method, LORA communication method and HTTP request method.

[0134] Preferably, the environmental data collected by the inspection robot is implemented through MQTT and LoRa communication technologies. The MQTT protocol is based on a publish / subscribe model and is lightweight and low-bandwidth, making it suitable for efficiently uploading data collected by various environmental sensors in the chicken house (such as temperature and humidity) to the edge computing server. LoRa technology, with its long-distance transmission advantages, ensures that environmental data can be stably transmitted from the inspection robot to the edge server in the relatively complex spatial layout of multi-layer cage chicken houses.

[0135] Since images, sounds, and other data typically have large data volumes, HTTP requests are well-suited to this data transmission requirement. By constructing a suitable request message format, these multiple data channels can be accurately sent from the inspection robot to the edge computing server for subsequent analysis and processing. Therefore, the inspection robot uploads the collected image, sound, and temperature data to the edge server via HTTP requests.

[0136] As the inspection robot moves along a preset trajectory within the chicken coop, it uses SLAM technology to construct a real-time map of the coop's interior and determine its precise position within the map. Based on this constructed map and the robot's planned trajectory, it calculates the distance between its starting point (typically the coop entrance or a pre-set initial inspection position) and the current point. Combined with preset parameters such as the cage spacing, the robot can accurately determine the row and specific cage number corresponding to its current position, providing critical information for subsequently pinpointing the location of dead chickens.

[0137] The SLAM positioning technology constructs a real-time map of the internal environment of the chicken house to determine the precise position of the inspection robot in the map, and calculates the row number and cage number of the chicken cage based on the preset chicken cage arrangement spacing parameters.

[0138] like Figure 1 As shown, the inspection robot consists of a chassis 1, a multimodal sensor assembly 2 mounted above it, a lifting platform 3, and a control cabinet 4. Chassis 1 is a four-wheel drive differential chassis with large-diameter tires that can overcome 4cm obstacles, ensuring stable movement and flexible adaptability in complex chicken house environments. Chassis 1 integrates SLAM functionality, forming a SLAM chassis 1. It is equipped with a multi-line LiDAR and IMU sensor, which can construct a 2D grid map with a resolution of 5cm in real time and support dynamic obstacle avoidance and path planning.

[0139] The multimodal sensor group 2 includes an environmental acquisition card, a camera group, and a microphone array. The number of camera groups is 8, and they are arranged in four layers and two rows. Each row contains a dual-light camera and an infrared thermal imaging sensor (±2°C accuracy) installed back to back. The dual-light camera is an IP camera composed of a visible light and infrared light dual-light camera (marked as "IP camera" in the accompanying drawings to match other components of the platform). The resolution is 1280×720 resolution. The camera group and the microphone array are both installed on the lifting platform 3. In one or more embodiments, the lifting platform 3 is a three-stage telescopic high-strength aluminum square tube with a vertical travel range of 1.4 to 2.6m, which accurately adapts to the height distribution of the four-layer cage frame of the chicken house. The aluminum alloy square tube has a built-in synchronous belt transmission mechanism, which is driven by a servo motor to achieve vertical lifting and lowering. It cooperates with an absolute encoder to achieve a repeatable positioning accuracy of ±0.5mm. The servo motor signal is connected to the edge computing server. The aluminum alloy square tube includes a base tube and a first telescopic tube and a second telescopic tube arranged in sequence above it. A row of camera groups is respectively installed on the middle and top of the base tube, the top of the first telescopic tube, and the top of the second telescopic tube.

[0140] like Figure 2 As shown, control cabinet 4 houses the Jetson Orin NX master embedded microcontroller, an industrial switch, relays, stepper actuator drivers, an environmental data acquisition card, an RFID reader, and electrical wiring. A touch screen is also mounted on the wall of control cabinet 4. The industrial switch integrates data from the eight dual-light cameras and the SLAM chassis 3, transmitting it to the Jetson Orin NX core computing unit via the LAN port. The environmental data acquisition card monitors temperature, humidity, wind speed, light intensity, ammonia, and CO2 concentrations in real time, uploading the data to the Jetson Orin NX (6-core ARM CPU, 16GB video memory) via the RS485 bus and USB4 interface for fusion analysis. The relay module receives commands via the 485 protocol, controlling the stepper actuators to adjust the platform height, start and stop the operating lights, and capture images with the camera. Simultaneously, an RFID reader reads the chicken coop identification information via the USB2 interface and binds it to the SLAM positioning data, the dual-light camera images (encoded in H.264 and uploaded via HTTP), and environmental parameters, forming a multi-source data association. The touchscreen provides a human-machine interface via HDMI / USB1, displaying real-time monitoring data and equipment status. It allows operators to issue inspection instructions or adjust parameters, ultimately achieving fully automated, high-precision collaborative monitoring of the chicken house environment and sick and dead chickens. The core computing unit, Jetson Orin NX, is responsible for SLAM positioning and image decoding, ensuring efficient system operation and meeting the requirements of intelligent inspection.

[0141] The inspection robot uses a multimodal sensor set 2 to collect, process, and transmit environmental, image, sound, and infrared temperature data in real time, ensuring precise monitoring and intelligent inspections. Environmental data is collected in real time by an environmental acquisition card, encapsulated in JSON format, and uploaded to the edge server via the MQTT protocol (LoRa module). Abnormal conditions (such as ammonia concentration exceeding a threshold) trigger local alarms and initiate focused inspections. Image data is captured by the camera at a frame rate of 0.5 seconds. The vision function package decodes the data and superimposes SLAM coordinates (x, y) on it. The data is then transmitted to the server via HTTP streaming. Simultaneously, a microphone array collects voiceprint data, processes it for noise reduction, and uploads it simultaneously. Infrared temperature data is generated by an infrared thermal imaging sensor, which generates a temperature distribution map for each bird. This data is then linked and stored with the corresponding cage number information via the IoT function package, enabling precise individual temperature monitoring. This integrated solution ensures efficient data transmission, anomaly warnings, and precise tracking, enhancing the intelligent level of environmental monitoring in the poultry house.

[0142] The inspection robot and its supporting equipment have built an efficient intranet data collection and transmission system to ensure the stable operation of environmental monitoring, video processing and remote collaborative work.

[0143] In order to enable external terminal devices to easily access platform-related components (such as edge computing servers) located within the chicken house's internal LAN, intranet penetration technology is used. By deploying the corresponding intranet penetration software or services on the server side, specific ports in the internal network are mapped to externally accessible public network addresses. This allows external mobile and desktop devices to communicate and interact with the platform just as they would within the LAN, without the need for complex network configuration.

[0144] The network topology combines a star and tree structure, with the edge computing server at the core, and inspection robots, network devices, and various terminal devices built around it. Inspection robots establish a direct connection to the edge computing server via the network device (the star portion), while each terminal device connects to the server through network devices in a step-by-step manner, forming a tree-like branching structure. This ensures efficient data transmission while also balancing network scalability and stability, facilitating subsequent platform upgrades and device additions.

[0145] The platform uses an intranet penetration configuration. By deploying the frp service, the server ports (such as HTTP 8080 and MQTT 1883) are mapped to public domain names (such as chickenfarm.example.com:8000), enabling external devices to access them through HTTPS encryption, ensuring the security of remote control, and using the TLS 1.3 protocol for encrypted data transmission.

[0146] like Figure 4As shown, the overall network architecture adopts a hybrid model, with a star-shaped core layer consisting of inspection robots and Jetson Orin NX directly connected via an industrial switch (1Gbps bandwidth) to ensure low-latency data transmission (end-to-end latency <50ms). The tree-shaped expansion layer supports access from external devices (such as mobile devices and NVRs) via 4G / 5G gateways and can dynamically expand new nodes (such as temperature and humidity monitors), improving system flexibility and scalability. In addition, to enhance system stability, disaster recovery and caching mechanisms are deployed. The edge servers adopt a dual-machine hot standby master-slave architecture, ensuring that if the master node (IP 192.168.1.107) fails, the slave node (IP 192.168.1.X) can automatically take over the service and maintain business continuity. At the same time, the robot has a built-in E57A storage module, which can cache data locally (for 24 hours) in the event of a network interruption and automatically retransmit it to the cloud via the intranet penetration service after the network is restored, ensuring data integrity and system reliability.

[0147] In one embodiment, the inspection robot is assigned IP addresses 10.10.1.11-18 and connected to the core computing unit Jetson Orin NX IP 10.10.1.1 via the industrial switch IP 192.168.1.104 to enable environmental data collection and real-time control. The intranet video stream pull address of the 8-channel dual-light camera is 192.168.1.107:6001-6008. The video data is encoded in H.264 and uploaded to the NVR network video recorder for storage and management. To achieve remote access, the edge computing server IP 192.168.1.107 deploys the FRP intranet penetration service, mapping key ports such as HTTP 8080 and MQTT 1883 to public network addresses such as chickenfarm.example.com 8000, allowing external network devices such as cloud servers and laboratory workstations to securely access intranet resources via the HTTPS protocol. In addition, the cloud-based collaborative mechanism enables external network streaming services to receive video streams through cloud servers and store them in association with environmental data such as temperature and humidity, so that remote users can monitor and analyze them in real time, thereby improving the system's intelligence level and remote operation and maintenance capabilities.

[0148] The platform can realize multi-terminal command execution and priority scheduling, specifically including: (1) Mobile and desktop users can input commands such as "emergency stop" or "local inspection" through the APP or Web interface. The system encapsulates the command into a JSON message with a priority identifier and transmits it to the server for processing through the MQTT protocol. At the same time, the local robot has the ability to make independent decisions. When the battery power is lower than 20%, it will automatically trigger the "return to charge" command and push it to the task queue with the help of the ROS event processing function package to ensure the autonomy of the robot operation and the stability of task execution. (2) The system adopts a strict legality verification mechanism to verify the authority and parameter range of user commands, such as determining whether the inspection path is out of bounds. For illegal commands, the corresponding error code (such as ERR_403) is returned to ensure the compliance and safety of the command. At the same time, the command execution adopts a dynamic priority queue and performs hierarchical management according to the urgency of the task. For example, "dead chicken processing" is set as priority 1 and "routine inspection" is set as priority 3. The execution order is dynamically adjusted based on the ROS task scheduling module to optimize the task response efficiency. In addition, the system has built a complete feedback closed-loop mechanism to push the execution status to the sender in real time, including status such as "executing", "success" or "failure", and record all execution logs in the database to ensure the traceability of tasks and the reliability of the system.

[0149] The platform has a multi-terminal command execution framework, which includes mobile terminals, desktop terminals, and local robot terminals. The command execution method is as follows:

[0150] T1. Command reception phase: The operator inputs corresponding commands (such as controlling the inspection robot to start inspections, querying data in specific areas, etc.) through the preset user interface of each terminal device (mobile and desktop). These commands are sent to the edge computing server through the network in a specific message format. The local robot can also generate corresponding commands according to its own preset automation rules or sensor trigger mechanism (such as actively requesting local key inspections when an abnormal situation is detected) and send them to the server.

[0151] T2. Command Judgment Phase: After receiving a command, the edge computing server verifies its legitimacy and validity. For example, it determines whether the operations involved in the command are within the platform's permitted range and whether the parameter settings are reasonable. If the command does not meet the requirements, an error message is returned to the sender, requesting the correct command be resent.

[0152] T3. Command queuing stage: After judging the legality and validity of the commands, they are added to the queue of pending commands according to the order in which they are received or according to the priority of the commands (for example, commands related to the handling of dead chickens in an emergency have a higher priority), ensuring that the commands can be executed in an orderly manner.

[0153] T4. Command feedback phase: When a command starts or is completed, the edge computing server will feedback the corresponding execution status information (such as whether the command has started execution, execution success, execution failure, and the reason for failure, etc.) to the command sender, so that the operator can understand the execution status of the command in a timely manner and make subsequent operational decisions.

[0154] T5. Command execution phase: In queue order, the edge computing server parses the command and sends it to the corresponding execution component (such as an inspection robot). The execution component completes the specific command operation (such as controlling the robot to move to a specified location for data collection, etc.). During the execution process, the execution component can also provide real-time feedback on execution progress and other information to the server based on actual conditions, further ensuring the accuracy and reliability of command execution.

[0155] An inspection platform for inspection robots in multi-layer cage chicken houses, based on an implementation of ROS2 full-link collaborative instruction execution and status feedback.

[0156] like Figure 7 As shown in the figure, the MQTT and ROS2 asynchronous message-driven architecture are integrated to form an end-to-end collaborative link from command input, protocol conversion, task scheduling, hardware execution to status feedback: users input control commands and parameter settings through the IoT communication function package or GUI function package in the cloud or local QT5GUI interface, and the front end encapsulates them into JSON messages and publishes them to the chassis status topic, task status topic, meteorological data topic (IoT side) and lift control topic, relay control topic, and visual command topic (GUI side); each ROS2MQTT Client subscribes to the corresponding topic, completes JSON parsing, message verification and message conversion, and then publishes the command and status as ROS2 messages to the central ROS2Server node. ROS2Server has built-in format and permission filters and a message deduplication mechanism. It pushes legitimate messages into a blocking queue based on dynamic priority, and drives the microcontroller to route them sequentially to processes such as navigation, lift, visual acquisition, exception handling, and task completion. These processes asynchronously call specific functional modules through the ROS2Action client. The chassis action uses the SLAM chassis HTTP interface to obtain odometer and position in real time and perform navigation or manual movement. The lift action is driven by a servo motor and ball screw closed-loop to achieve millimeter-level lifting. The camera function package captures images at a preset frame rate and uploads them for visual algorithm review. During execution, each module continuously transmits feedback on the current position, motion state, payload, and environmental sensors. ROS2Server then repackages this feedback and pushes it to the cloud and GUI via MQTT, enabling simultaneous multi-terminal monitoring. The system also uses global variables to control topics to flexibly switch on and off streaming media transmission, target detection data transmission, and global emergency stops.

[0157] When the LiDAR detects an obstacle, the MCU receives the information and controls the chassis to trigger obstacle avoidance and re-plan the route. When the gas sensor detects that the concentration exceeds the limit, the MCU issues an emergency shutdown and activates the buzzer and warning light. In the dead chicken detection scenario, after the camera group transmits the signal to the MCU, the MCU controls the positioning of the chassis and the lifting platform and instructs the camera to re-inspect. Finally, the results are fed back to the cloud and cleaning or isolation instructions are automatically issued, completing the "monitoring → verification → disposal" closed loop. With MQTT+ROS2 dual-channel coupling, dynamic priority scheduling, Action asynchronous closed loop and state machine management, this solution not only achieves efficient and reliable full-link collaboration, but also reserves sufficient space for the subsequent introduction of higher-level communication encryption, multi-path redundancy and service quality optimization. The MCU is the MCU that comes with the inspection robot and is located in the control cabinet.

[0158] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An inspection method for a multi-layer cage chicken house using an inspection robot, characterized in that: The steps include: S1, build SLAM map and plan motion trajectory; S2. Set the numbering rules for the inspection robots and the activity rules for the camera groups; S3. Setting the cage number calculation model and layer mapping rules of the chicken cage; S4. A copy of the data collected by the inspection robot is transmitted to the cloud, and the local VVR saves a copy to support fast playback or offline analysis. When a key event is detected, the inspection robot's system automatically captures the relevant frame image and stores it in conjunction with the environmental data and cage number information to facilitate subsequent tracing and report generation.

2. The inspection method for a multi-layer cage chicken house using an inspection robot according to claim 1, characterized in that: The method for setting the numbering rules of the inspection robots and the activity rules of the camera groups in S2 is: S21. Define the global coordinate system Formulate the rules for the numbering of chicken cages: the row numbers increase from west to east: row 1 is on the westernmost side, row R is on the easternmost side; the aisle number A increases from west to east: A=1 is the westernmost aisle, A=Amax is the easternmost aisle; Specify the mapping direction of the inspection robot: NORTH: Y+ in the global coordinate system; SOUTH: Y- in the global coordinate system; The camera groups on the left side of the robot are numbered 1-4, and the right side of the camera groups on the right side of the robot are numbered 5-8; S22. Based on the global coordinate system, determine the row number according to the aisle number A and the inspection direction (NORTH, SOUTH) uploaded by the SLAM chassis.

3. The inspection method for a multi-layer cage chicken house using an inspection robot according to claim 2, characterized in that: In S22, the specific method for determining the row number based on the global coordinate system and the aisle number A uploaded by the SLAM chassis and the inspection direction (NORTH, SOUTH) is as follows: S221. Generate number function: Among them, A represents the aisle number uploaded by the SLAM chassis, Amax represents the last inspected aisle, and D represents the robot's inspection direction. The inspection direction is used to determine the robot's inspection direction and calculate the row number. NORTH means the robot inspects in the north direction, and SOUTH means the robot inspects in the south direction. It means that there is no inspection row to the right of the robot, and R is the last row of chicken houses; S222. Setting camera activation rules Among them, A represents the aisle number uploaded by the SLAM chassis, D represents the robot's inspection direction, NORTH means the robot is inspecting in the north direction, and SOUTH means the robot is inspecting in the south direction.

4. The inspection method for a multi-layer cage chicken house using an inspection robot according to claim 2, characterized in that: The method for setting the cage number calculation model and layer mapping rules of the chicken cage in S3 is: S31. Set SLAM points and define parameters Cage width: W (the width of each cage in the X-axis direction); Starting point: (x_start, y_start); Current position: (x_current, y_current); S32, calculation cage number model Calculate the distance based on the uploaded start point (x_start, y_start) and the current inspection position (x_current, y_current), and calculate the cage number formula: S33. Set layer mapping rules The camera number directly corresponds to the layer number: Left side cameras 1-4: monitor the 1st to 4th floors of the left cage row; Right cameras 5-8: monitor the 1st to 4th floors of the right cage row; Physical meaning: Each cage column is divided into 4 layers in the vertical direction, and the robot monitors different layers through different cameras.

5. The inspection platform of the inspection robot for multi-layer cage chicken houses is characterized by: The inspection method using the inspection robot for multi-layer cage chicken houses according to any one of claims 1 to 4 comprises: Inspection robots, used to collect environmental data, images, sounds, and temperature data within the chicken house; The edge computing server is used to perform preliminary processing and analysis on the data collected and uploaded by the inspection robot, and quickly screen out the relevant data information of suspected dead chickens. Network equipment, used to build communication links between inspection robots and edge computing servers, and between servers and terminals; Among them, the environmental data is uploaded through MQTT and LORA communication, and the image, sound and temperature data are uploaded through HTTP request.

6. The inspection platform of the inspection robot for multi-layer cage chicken houses according to claim 5, characterized in that: The inspection robot consists of a chassis and a multimodal sensor group installed above it, a lifting platform and a control cabinet. The SLAM function is integrated inside the chassis to form a SLAM chassis, which can build a 2D grid map with a resolution of 5cm in real time, and support dynamic obstacle avoidance and path planning. The multimodal sensor group is used to collect environmental data, images, sounds and temperature data in the chicken house and transmit them to the controller in the control cabinet. The lifting platform is a three-section telescopic square tube with a vertical travel range of 1.4 to 2.6m, which accurately adapts to the height distribution of the four-layer cages in the chicken house.

7. The inspection platform of the inspection robot for multi-layer cage chicken houses according to claim 5, characterized in that: The inspection robot establishes a star-shaped connection structure with the edge computing server through network equipment, and each terminal device is connected to the server step by step through the network equipment to establish a tree-shaped branch structure, thus forming a topological structure that combines star and tree structures with the edge computing server as the core, and the inspection robot, network equipment and each terminal device.

8. The inspection platform of the inspection robot for multi-layer cage chicken houses according to claim 5, characterized in that: Deploy the corresponding intranet penetration software or service on the server side, and use intranet penetration technology to map the internal server port to an externally accessible public network address, so that external mobile and desktop devices can communicate and interact with the platform just like in a local area network without complex network configuration.

9. The inspection platform of the inspection robot for multi-layer cage chicken houses according to claim 5, characterized in that: Tasks are managed hierarchically according to their urgency and a dynamic priority queue is formed, and the order of priority can be adjusted dynamically.

10. The inspection platform of the inspection robot for multi-layer cage chicken houses according to claim 9, characterized in that: It also includes a multi-terminal command execution framework, which includes mobile terminals, desktop terminals, and local robot terminals. The command execution method is as follows: T1, command reception phase: Each terminal device sends the command in a set format to the edge computing server through a preset user interface; the local robot generates the corresponding command based on its own preset automation rules or sensor trigger mechanism and sends it to the server; T2, command judgment stage: After receiving the command, the edge computing server judges the legality and validity of the command; T3, command queueing stage: After judging the legality and validity of commands, they are added to the queue of pending commands in the order of receipt or according to the priority of the commands to ensure that the commands can be executed in an orderly manner; T4, Command Feedback Stage: When a command starts or completes execution, the edge computing server feeds back the corresponding execution status information to the command sender, allowing operators to promptly understand the execution status of the command and make subsequent operational decisions. T5, command execution phase: According to the queue order, the edge computing server parses the command and sends it to the corresponding execution component, which completes the specific command operation. During the execution process, the execution component can also provide real-time feedback on execution progress and other information to the server based on actual conditions.

Images