A multi-mode adaptive disinfection method and disinfection robot

Abstract

The present application belongs to the technical field of robot intelligent disinfection, and relates to a multi-mode adaptive disinfection method and a disinfection robot. When the disinfection robot first enters a disinfection space, a three-dimensional semantic map and a two-dimensional grid map of the disinfection space are automatically constructed. After receiving each round of disinfection task, the following steps are executed: planning a path based on the two-dimensional grid map and moving to the target disinfection area according to the planned path; collecting the environmental information of the target disinfection area and updating the three-dimensional semantic map in real time; generating a disinfection strategy by using a privatized large language model based on the received disinfection task and the updated three-dimensional semantic map, driving the mechanical arm to perform disinfection operation, monitoring the environmental information of the target disinfection area in real time during the disinfection process, and recording the disinfection data in the entire disinfection process; generating a disinfection result evaluation report of this disinfection task based on the recorded disinfection data; and returning to wait for receiving the next round of disinfection task. The present application can significantly improve the disinfection effect and work efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of robot intelligent disinfection, and in particular to a multi-mode adaptive disinfection method and a disinfection robot. Background Art

[0002] With the frequent occurrence of public health incidents, the application of disinfection technology in hospitals and various public places has become increasingly important. However, traditional disinfection methods, such as chemical spraying and photoelectric disinfection, have limitations such as heavy workload, high risk of cross-infection, and a single disinfection mode. They are difficult to flexibly adapt to the needs of different environments, and disinfection effectiveness is difficult to guarantee.

[0003] In order to solve the above problems, intelligent disinfection technology came into being. The prior art discloses an intelligent disinfection robot, which consists of a robot shell, a multi-joint robotic arm, a quick movement system, a core disinfection system and a visual recognition system. However, the robot adopts a single robotic arm design, and its operational flexibility is limited. It can only deal with a single type of accident and is difficult to adapt to the multi-task collaboration needs in complex scenarios. In terms of disinfection mode, the technology only supports two basic means: plasma and ultraviolet lamps, and lacks an aerosol chemical disinfection module, resulting in a disinfection coverage rate of as low as 55% for hidden areas (such as gaps and pipes), and is unable to cope with highly resistant pathogens. The design defects of the mobile system further restrict the adaptability of the scene: the height of the robot chassis from the ground is fixed at 80 to 120 mm, and it is not equipped with a suspension system, resulting in poor passability in low door frames or rugged ground.

[0004] Prior art also discloses a disinfection robot that uses three modes: ultraviolet light disinfection, aerosol chemical disinfection, and photocatalytic disinfection, achieving improved disinfection efficiency. However, this robot lacks a robotic arm collaboration unit, preventing it from performing physical intervention operations. This limits its ability to perform tasks in complex environments, reducing its disinfection coverage to below 60%, making it difficult to meet the requirements for efficient and safe disinfection.

[0005] In view of this, it is urgent to propose an intelligent disinfection technology with high disinfection efficiency, adaptability to various complex working environments, and high intelligence. Summary of the Invention

[0006] The purpose of the present invention is to address the shortcomings of the existing technology, such as the single disinfection mode, difficulty in adapting to complex scenarios, and low disinfection efficiency. A multi-mode adaptive disinfection method and a disinfection robot are proposed. The method can switch between three different disinfection modes according to disinfection needs, and can dynamically optimize the disinfection strategy according to environmental changes. It has the advantages of high adaptability, applicability to a variety of complex environments, and high disinfection efficiency.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] In a first aspect, the present application provides a multi-mode adaptive disinfection method, which is executed by a multi-mode adaptive disinfection robot. When the disinfection robot enters a disinfection space for the first time, it automatically constructs a three-dimensional semantic map and a two-dimensional grid map of the disinfection space, and then docks at a standby point to wait for receiving a disinfection task;

[0009] After receiving each round of disinfection task, the disinfection robot executes the following steps:

[0010] S10. Plan a path based on the two-dimensional grid map and move to the target disinfection area according to the planned path;

[0011] S11. Collect environmental information of the target disinfection area and update the three-dimensional semantic map in real time; the environmental information includes: spatial structure, pollution area, pathogen content, object category and pose;

[0012] S12. Based on the received disinfection task and the updated three-dimensional semantic map, generate a disinfection strategy using a private large language model; the disinfection strategy includes: selecting a disinfection mode, determining a disinfection sequence and a disinfection duration;

[0013] S13. The disinfection robot drives the mechanical arm carried by it to execute disinfection operation based on the disinfection strategy, and monitors the environmental information of the target disinfection area in real time during the disinfection process, and records the disinfection data during the entire disinfection process;

[0014] S14. After the disinfection is completed, generate a disinfection result evaluation report for this disinfection task based on the recorded disinfection data;

[0015] S15. Return to the standby point and wait for receiving the next round of disinfection task.

[0016] As a possible implementation manner, the three-dimensional semantic map and the two-dimensional grid map of the disinfection space are automatically constructed by the following method:

[0017] S20. Obtain the environmental point cloud data of the disinfection space by using a laser radar, and collect the color and depth information of the disinfection space by using a depth camera;

[0018] S21. Perform space-time calibration and registration on the color and depth information to obtain registered color and depth information;

[0019] S22. Colorize and density complete the environmental point cloud data by using the registered color and depth information to obtain dense color point cloud;

[0020] S23. Perform semantic segmentation on the dense color point cloud by using a lightweight RandLA-Net network to identify pollution areas, fixed obstacles and passable areas, and obtain a three-dimensional semantic map;

[0021] S24. Project the fixed obstacles and traversable areas in the three-dimensional semantic map onto a two-dimensional plane to obtain a two-dimensional grid map.

[0022] As a possible implementation method, an improved hybrid A* algorithm is used to plan the path based on a two-dimensional grid map. The hybrid A* algorithm is:

[0023] C total =α·C path +β·C obstacle +γ·C task +δ·C kinematic

[0024] Where C total represents the global path planning cost function, C path represents the path length cost, C obstacle represents the obstacle avoidance cost, C task represents the cost of the disinfection task, C kinematic represents the kinematic constraint cost, α is the path length weight, β is the obstacle avoidance weight, γ is the elimination task weight, and δ is the kinematic constraint weight;

[0025] Improved disinfection task cost C task Expressed as:

[0026]

[0027] Among them, P(p k ) indicates p k The pollution probability of the location is obtained based on the 3D semantic map and has a value range of (0, 1); unvisted (p k ) represents the indicator function, where the undisinfected area is 1 and the disinfected area is 0); ω priority Priority of controlling pollution areas, ω coverage Balance global coverage, with the optimization goal of prioritizing high-pollution areas while maximizing disinfection coverage;

[0028] During the movement of the disinfection robot, near-field dynamic obstacles are detected in real time to generate local distance field data; ultrasonic ranging data and lidar data are fused into multi-sensor data, and the occupancy probability of the two-dimensional grid map is dynamically corrected using Bayesian updating; when a sudden obstacle is detected, local path replanning is triggered, and a local real-time obstacle avoidance path is generated based on the TEB algorithm.

[0029] As a possible implementation method, the disinfection modes include: plasma disinfection mode, ultraviolet lamp disinfection mode and aerosol chemical disinfection mode; the plasma disinfection mode adopts the atmospheric pressure cold plasma jet method, and uses an all-solid-state fast pulse generator to drive the dielectric barrier discharge to generate active free radicals, destroy the microbial membrane structure, and achieve disinfection.

[0030] As a possible implementation, the following method is used to train the large language model:

[0031] S120. Construct a corpus of disinfection tasks, clean and label the corpus content, and form training samples for supervised learning;

[0032] S121. Select a general large language model as the base model and load pre-trained weights;

[0033] S122. Based on the provided training samples, use supervised fine-tuning methods to fine-tune the selected base model for the target domain, enabling it to generate a standardized disinfection strategy based on the input structured / unstructured data;

[0034] S123. Introduce human feedback reinforcement learning mechanism to optimize model output.

[0035] As a possible implementation method, the disinfection tasks include daily disinfection tasks and emergency disinfection tasks;

[0036] After receiving daily disinfection tasks, the disinfection robot generates the following disinfection strategy:

[0037] Determine whether there are people in the target disinfection area. If there are people, select the plasma disinfection mode; if there are no people, select a composite disinfection mode including ultraviolet lamp disinfection mode, plasma disinfection mode and aerosol chemical disinfection mode;

[0038] Calculate the disinfection time based on the pathogenic microorganism content in the target disinfection area;

[0039] Use a single robotic arm to perform surface disinfection operations;

[0040] After receiving the emergency disinfection task, the disinfection robot generates the following disinfection strategy:

[0041] A vision-guided dual-arm collaborative control algorithm is used to drive the collaborative operation of two robotic arms to perform emergency treatment in the spill area, including laying paper, wiping, and cleaning.

[0042] After the emergency treatment is completed, select the aerosol chemical disinfection mode for disinfection;

[0043] Calculate the disinfection time based on the pathogenic microorganism content in the spill area.

[0044] As a possible implementation method, a dual-arm collaborative control algorithm based on vision guidance is used to drive the collaborative operation of two robotic arms, including the following steps:

[0045] S130 real-time collection of RGB images and depth information of the spill area, and converted into point cloud data for three-dimensional reconstruction of the spill area;

[0046] S131. Discretize the point cloud data into a voxel grid. Use a 3D convolution-based UNet 3D network to process the voxel grid and predict the Q value of each voxel. The Q value represents the long-term benefit of choosing a certain action in the current state.

[0047] S132. Perform local voxelization on the area surrounding the voxel index with the largest Q value, construct a new Q value prediction network, and determine the translation motion of the robotic arm based on the local voxelized maximum Q value index;

[0048] S133. Discretize the rotation of the manipulator into a rotation index every 5°, use a fully connected neural network to predict the rotation and gripper motion of the manipulator, and generate a 7D pose for controlling the motion of the manipulator;

[0049] S134. Transmit the 7D pose command to the robotic arm controller through the ROS framework.

[0050] In a second aspect, the present invention provides a multi-mode adaptive disinfecting robot for executing the multi-mode adaptive disinfecting method provided in the first aspect, the multi-mode adaptive disinfecting robot comprising: a mobile unit, a robot body, an environment sensing unit, a disinfecting unit, a robotic arm collaboration unit, and a control unit;

[0051] The environmental perception unit collects environmental information in real time and transmits it to the control unit. The environmental information includes: spatial structure, contaminated area, pathogen content, object type and position;

[0052] The control unit includes a host computer and a slave computer. The host computer generates corresponding disinfection strategies and autonomous navigation paths based on environmental information, and converts them into control instructions and sends them to the slave computer. The slave computer coordinates and controls the mobile unit, disinfection unit, and robotic arm collaboration unit to perform disinfection tasks based on the control instructions.

[0053] The robotic arm collaboration unit includes two six-degree-of-freedom robotic arms. Each six-degree-of-freedom robotic arm is equipped with a quick-release component to quickly replace the modular end effector. The modular end effector has two replaceable working tools, including: a multi-finger gripper and an adaptive suction cup.

[0054] As a possible implementation method, the environmental perception unit includes: pathogen detection components, depth cameras, lidar, distance sensors, and temperature and humidity sensors;

[0055] The lidar is used to perform three-dimensional reconstruction of the target disinfection area, the depth camera is used to collect RGB images and depth information of the target disinfection area, the distance sensor is used for obstacle detection and distance measurement, the pathogen detection component is used to detect the concentration of pathogens in the target disinfection area in real time, and the temperature and humidity sensor is used to collect temperature and humidity data of the target disinfection area.

[0056] As a possible implementation method, the host computer is used to process environmental information, formulate navigation paths and disinfection strategies, and generate motion control instructions, which include robotic arm motion control instructions and mobile unit motion control instructions; the robotic arm motion control instructions are used to control the robotic arm to perform surface disinfection and emergency treatment of spills, and the mobile unit motion control instructions are used for autonomous navigation and dynamic obstacle avoidance;

[0057] The lower computer receives motion control instructions issued by the upper computer, drives the mobile unit, disinfection unit and robotic arm collaboration unit to perform corresponding tasks, collects multi-source data in real time and transmits it back to the upper computer to support the system's closed-loop control and dynamic response.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] 1. Existing pest control equipment often relies on manual operation or simple path planning technology and lacks the ability to adapt to complex environments. The multi-mode adaptive pest control method and pest control robot provided by this invention can achieve precise autonomous navigation and dynamic obstacle avoidance, eliminating manual intervention and improving the independence and safety of pest control equipment.

[0060] 2. The multi-mode adaptive disinfection method and disinfection robot proposed in the present invention integrate three modes: aerosol chemical disinfection, ultraviolet lamp disinfection and plasma disinfection. They can automatically switch or simultaneously enable multiple modes according to the environment and task requirements with the help of a private large language model, thereby significantly improving the disinfection effect and work efficiency.

[0061] 3. Traditional disinfection robots are typically equipped with only a single robotic arm, making them inefficient when handling complex tasks. In contrast, the disinfection robot provided by this invention can achieve dual-arm collaboration, more efficiently handling spills and leaks, and significantly improving the robot's performance and flexibility in complex tasks.

[0062] 4. The multi-mode adaptive disinfection robot proposed in the present invention adopts a multi-task adaptive design, which can simultaneously perform multiple tasks such as disinfection, detection, path planning, etc., and has autonomous decision-making and task scheduling capabilities, which significantly improves the robot's work efficiency in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] The accompanying drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0064] Figure 1 This is a flow chart of the multi-mode adaptive disinfecting method in an embodiment of the present invention;

[0065] Figure 2 Schematic diagram of the network structure of the improved YOLO algorithm in an embodiment of the present invention;

[0066] Figure 3 This is a schematic diagram of the structure of a multi-mode adaptive disinfection robot in an embodiment of the present invention;

[0067] Figure 4 This is an exploded diagram of the structure of the multi-mode adaptive disinfection robot in an embodiment of the present invention;

[0068] Figure 5 Schematic diagram of the structure of the ultraviolet lamp disinfection device in an embodiment of the present invention;

[0069] Figure 6 This is a schematic structural diagram of a plasma disinfection device according to an embodiment of the present invention;

[0070] Figure 7 Schematic diagram of the structure of the aerosol chemical disinfection device in an embodiment of the present invention.

[0071] Reference numerals

[0072] 1-mobile unit, 10-chassis, 11-drive motor, 12-wheel, 13-transmission assembly, 14-suspension assembly, 15-power supply;

[0073] 2- robot body, 20- disassembly assembly, 21- protective cover, 22- main assembly, 23- storage compartment;

[0074] 3-Environmental perception unit, 30-Pathogen detection component, 31-Depth camera, 32-LiDAR, 33-Distance sensor, 34-Temperature and humidity sensor;

[0075] 4- disinfection unit, 40- ultraviolet lamp disinfection device, 41- plasma disinfection device, 42- aerosol chemical disinfection device;

[0076] 5-robotic arm collaboration unit, 50-six-degree-of-freedom robotic arm, 500-quick-release assembly, 501-modular end effector;

[0077] 6-control unit, 60-upper computer, 61-lower computer. DETAILED DESCRIPTION

[0078] To facilitate a clear description of the technical solutions of the embodiments of the present invention, the words "first" and "second" are used in the embodiments of the present invention to distinguish between identical or similar items with substantially the same functions and effects. For example, the first threshold and the second threshold are merely used to distinguish between different thresholds and do not limit their order. Those skilled in the art will understand that the words "first" and "second" do not limit the quantity or execution order, and the words "first" and "second" do not necessarily mean different.

[0079] It should be noted that, in the present invention, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the present invention should not be construed as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0080] In the present invention, "at least one" refers to one or more, and "more" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: the existence of A alone, the existence of A and B at the same time, and the existence of B alone, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. The following at least one item (item) or similar expressions thereof refer to any combination of these items, including any combination of single items (items) or plural items (items). For example, at least one item (item) of a, b or c can represent: a, b, c, the combination of a and b, the combination of a and c, the combination of b and c, or the combination of a, b and c, where a, b, c can be single or multiple.

[0081] The embodiments of the present invention aim to provide a multi-mode adaptive disinfection method and disinfection robot, which can switch between three different disinfection modes according to disinfection needs, and can dynamically optimize the disinfection strategy according to environmental changes. It has the advantages of high adaptability, applicability to a variety of complex environments, and high disinfection efficiency.

[0082] In a first aspect, an embodiment of the present invention provides a multi-mode adaptive disinfection method, which is performed by a multi-mode adaptive disinfection robot. When the disinfection robot first enters a disinfection space, it automatically constructs a three-dimensional semantic map and a two-dimensional grid map of the disinfection space. After the construction is completed, it docks at a standby point and waits to receive disinfection tasks.

[0083] See also Figure 1 After receiving each round of disinfection tasks, the disinfection robot performs the following steps:

[0084] S10. Plan a path based on a two-dimensional grid map and move to the target disinfection area according to the planned path;

[0085] As a possible implementation method, an improved hybrid A* algorithm is used to plan the path based on a two-dimensional grid map. The hybrid A* algorithm is:

[0086] C total =α·C path +β·C obstacle +γ·C task +δ·C kinematic

[0087] Where C total represents the global path planning cost function, C path represents the path length cost, C obstacle represents the obstacle avoidance cost, C task represents the cost of the disinfection task, C kinematic represents the kinematic constraint cost, ɑ is the path length weight, β is the obstacle avoidance weight, γ is the elimination task weight, and δ is the kinematic constraint weight;

[0088] Improved disinfection task cost C task Expressed as:

[0089]

[0090] Among them, P(p k ) indicates p k The pollution probability of the location is obtained based on the 3D semantic map and has a value range of (0, 1); unvisted (p k ) represents the indicator function, where the undisinfected area is 1 and the disinfected area is 0); ω priority Priority of controlling pollution areas, ω coverage Balance global coverage, with the optimization goal of prioritizing high-pollution areas while maximizing disinfection coverage;

[0091] During the movement of the disinfection robot, near-field dynamic obstacles are detected in real time to generate local distance field data; ultrasonic ranging data and lidar data are fused into multi-sensor data, and the occupancy probability of the two-dimensional grid map is dynamically corrected using Bayesian updating; when a sudden obstacle is detected, local path replanning is triggered, and a local real-time obstacle avoidance path is generated based on the TEB algorithm.

[0092] This application combines a three-dimensional semantic map with dynamic obstacle avoidance capabilities and local path replanning to not only calculate the shortest path, but also dynamically adjust the path based on task priority and regional characteristics. For example: in a hospital environment, if a room is marked as a high-contamination area, an optimal path is calculated to prioritize that area, rather than just choosing the shortest path. In laboratories with heavy traffic, detours can be selected to reduce the risk of contact with personnel. In chemical storage areas, disinfection strategies will be adjusted to avoid the use of disinfectants that may produce chemical reactions.

[0093] Existing pest control equipment often relies on manual operation or simple path planning technology, lacking the ability to adapt to complex environments. The method provided by the present invention can achieve precise autonomous navigation and dynamic obstacle avoidance, avoiding manual intervention and improving the independence and safety of pest control equipment.

[0094] S11. Collect environmental information about the target disinfection area and update the 3D semantic map in real time. Environmental information includes: spatial structure, contaminated area, pathogen content, object type, and position.

[0095] As a possible implementation method, sensors are used to collect environmental information of the target disinfection area. An improved YOLO algorithm is used to identify contaminated areas. Depth cameras are used to collect semantic information and label risk levels based on the degree of contamination. Spatial structures are classified into different scenarios, such as medical wards, offices, elevators, and public areas. The types of pollutants, such as bacteria, viruses, or organic pollutants, are determined. This environmental information is then converted into structured data. For example, the environmental information of a hospital ward can be converted into:

[0096] Location: Hospital; Spatial structure: Medical ward; Object type: Biological reagent bottle; Contamination type: Biological contamination (bacteria + viruses); Contaminated area size: 15 square meters; Airborne transmission risk: High.

[0097] As an example, the network structure of the improved YOLO algorithm is as follows Figure 2 As shown in the figure, the improvements to the YOLO algorithm in this embodiment include the following aspects: first, a lightweight phantom convolution layer is used to compress redundant features, which significantly reduces the amount of calculation while ensuring accuracy; second, a weighted bidirectional feature pyramid and global-local attention fusion structure is designed. By combining the bidirectional multi-scale feature pyramid with the global-local spatial attention mechanism, the defect localization capability under background interference is significantly enhanced; third, an innovative detection module with spatial and channel collaborative attention is introduced, and feature expression is optimized through spatial-channel collaborative attention.

[0098] S12. Based on the received disinfection tasks and the updated three-dimensional semantic map, a private large language model is used to generate a disinfection strategy. The disinfection strategy includes: selecting a disinfection mode, determining the disinfection sequence, and disinfection duration;

[0099] As a possible implementation, the following method is used to train the large language model:

[0100] S120. Construct a corpus of disinfection tasks, clean and label the corpus content, and form training samples for supervised learning;

[0101] As an example, based on authoritative standards, specifications, and industry case studies related to pathogen control, we compiled control procedures, environmental requirements, path planning rules, and optimal control mode conditions to construct a structured and unstructured corpus. This corpus was cleaned (de-duplicated, de-invalidated, and standardized terminology) and labeled according to different categories (such as region type, pathogen level, pathogenic microorganism, and presence of humans) to form training samples for supervised learning.

[0102] S121. Select a general large language model as the base model and load pre-trained weights;

[0103] As an example, select one of DeepSeek, ChatGLM, LLaMA, or BERT as the base model and load the pre-trained weights.

[0104] S122. Based on the provided training samples, use supervised fine-tuning methods to fine-tune the selected base model for the target domain, enabling it to generate a standardized disinfection strategy based on the input structured / unstructured data;

[0105] S123. Introduce human feedback reinforcement learning mechanism to optimize model output.

[0106] As an example, the model output is further optimized using a human feedback reinforcement learning mechanism (PPO algorithm) combined with the strategy pros and cons annotated by domain experts. After optimization, it is verified and evaluated using a test set, including accuracy, compliance, diversity, and generation robustness, to ensure that the output meets disinfection standards and meets actual usage needs. Finally, the fine-tuned large language model is encapsulated as an API interface or embedded service and integrated into the disinfection robot's control unit to generate disinfection strategies driven by environmental perception.

[0107] The specific implementation can be divided into four stages, the first of which is the data preparation stage: collect and organize the contents of the following authoritative standard documents: "Technical Standards for the Evaluation of National Standard Strains of Pathogenic Microorganisms (Viruses)",

[0108] Disinfection Technical Specification for Medical Institutions, Disinfection of Medical Waste, Disinfection Management Measures, and High-level Pathogenic Microorganism Laboratory Part 1: Disinfection Technical Specification. Extract the rules content such as disinfection mode selection principle, area division standard, air quality judgment basis, equipment usage condition, and personnel contact restriction. Build a disinfection task sample pair, for example: input: area type = laboratory, air quality = poor, pollution level = high, no personnel; output: select composite disinfection mode, ultraviolet + aerosol + plasma, duration 20 minutes. Store in JSON format or natural language, ensure machine learning format compatibility.

[0109] Secondly, the model training phase: select ChatGLM, LLaMA, Baichuan, etc. Open source language model to load corresponding basic weight, use training set to fine tune the model, training content includes: generating disinfection strategy text description, matching mode and pollution environment and space structure mapping, response structured information constructing strategy execution sequence. Introduce artificial labeling "good / bad strategy" samples, use PPO algorithm to train reinforcement learning model, optimize language model response quality and compliance.

[0110] Thirdly, the model verification and iteration phase: the verification dimension is set as disinfection strategy compliance, adaptability to different scenarios, disinfection priority judgment ability, output text logical consistency and execution. The sample strategy is submitted to infection control experts for scoring, combined with real environment data for robot simulation test.

[0111] Finally, the deployment and integration phase: the trained large prophecy model is packaged as a local SDK service, supporting structured environment information input such as JSON format. The output includes: disinfection mode, duration, operation steps, etc. Then embedded into disinfection robot control center (ROS robot operating system), linkage environment perception module input, trigger strategy generation and control execution.

[0112] As a possible implementation manner, the disinfection mode includes: plasma disinfection mode, ultraviolet lamp disinfection mode and aerosol chemical disinfection mode; the plasma disinfection mode adopts atmospheric pressure cold plasma jet method, generates active free radicals through solid-state fast pulse generator driven dielectric barrier discharge, destroys microbial membrane structure, and realizes disinfection.

[0113] As a possible implementation manner, the disinfection task includes daily disinfection task and emergency disinfection task;

[0114] After the disinfection robot receives the daily disinfection task, the following disinfection strategy is generated:

[0115] Determine whether there are people in the target disinfection area. If there are people, select the plasma disinfection mode; if there are no people, select a composite disinfection mode including the ultraviolet lamp disinfection mode, the plasma disinfection mode and the aerosol chemical disinfection mode;

[0116] Calculate the disinfection time based on the pathogenic microorganism content in the target disinfection area;

[0117] Use a single robotic arm to perform surface disinfection operations;

[0118] After receiving the emergency disinfection task, the disinfection robot generates the following disinfection strategy:

[0119] A vision-guided dual-arm collaborative control algorithm is used to drive the collaborative operation of two robotic arms to carry out emergency treatment in the spill area, including laying paper, wiping, and cleaning.

[0120] After the emergency treatment is completed, select the aerosol chemical disinfection mode for disinfection;

[0121] Calculate the disinfection time based on the pathogenic microorganism content in the spill area.

[0122] Generally, the aerosol chemical disinfection mode is suitable for space disinfection. For example, the use of atomized hydrogen peroxide can effectively kill bacteria, viruses, fungi and other microorganisms, and is non-corrosive to instruments and equipment. Alcohol or chlorine-containing disinfectants can be used for emergency treatment of spills, quickly killing pathogenic microorganisms and preventing the spread of contamination. The ultraviolet lamp disinfection mode is suitable for surface disinfection of objects and will not cause damage to the surface of objects, especially sensitive surfaces. The plasma disinfection mode is suitable for air purification and killing pathogenic microorganisms. The plasma disinfection mode uses an atmospheric pressure cold plasma jet method. Through an all-solid-state fast pulse generator, dielectric barrier discharge is driven to produce active free radicals, destroying the microbial membrane structure. It can effectively kill bacteria, viruses and other microorganisms in the air and is harmless to the human body. During daily disinfection, in order to reduce harm to personnel, if there are personnel present, select the plasma disinfection mode. When there are no personnel present, select multiple disinfection modes in a collaborative form for complex disinfection. The disinfection time is determined based on the selected disinfection mode and the actual requirements for the disinfection effect. For example, if a high concentration of air pollution and surface pollution is detected in the environment and no personnel are present, the disinfection mode is as follows: plasma disinfection of air pollution, setting the power to 80%, and disinfection time to 10 minutes; hydrogen peroxide spray disinfection of the entire contaminated area, spray concentration of 6%, and disinfection time to 12 minutes; combined with ultraviolet irradiation, auxiliary disinfection of local areas, with the intensity set to 200μW / cm 2 , disinfection time is 8 minutes.

[0123] As a possible implementation method, a dual-arm collaborative control algorithm based on vision guidance is used to drive the collaborative operation of two robotic arms, including the following steps:

[0124] S130 real-time collection of RGB images and depth information of the spill area, and converted into point cloud data for three-dimensional reconstruction of the spill area;

[0125] S131. Discretize the point cloud data into a voxel grid. Use a 3D convolution-based UNet 3D network to process the voxel grid and predict the Q value of each voxel. The Q value represents the long-term benefit of choosing a certain action in the current state.

[0126] For example, a voxel grid is a discrete representation of three-dimensional space that can transform continuous environmental information into computer-processable discrete data, providing input for subsequent Q-value prediction. By finding the voxel index with the maximum Q-value, the robot's global target area can be determined, providing guidance for subsequent local voxelization and motion planning.

[0127] S132. Perform local voxelization on the area surrounding the voxel index with the largest Q value, construct a new Q value prediction network, and determine the translation motion of the robotic arm based on the local voxelized maximum Q value index;

[0128] S133. Discretize the rotation of the manipulator into a rotation index every 5°, use a fully connected neural network to predict the rotation and gripper motion of the manipulator, and generate a 7D pose for controlling the motion of the manipulator;

[0129] As an example, during the global and local voxelization process, the parameters of the fully connected neural network are updated and optimized using the Q-learning algorithm. Simultaneously, the masked voxel grid is reconstructed with the Q value of the original voxel output as the target, and the difference in Q value between the reconstructed voxel and the original voxel is evaluated using cosine similarity. The fully connected neural network loss and the reconstruction loss are combined and acted on the fully connected neural network parameter update process to improve the fully connected neural network's perception of three-dimensional space, as follows:

[0130]

[0131] L total =L p +λL r ,λ∈(0,1)

[0132] Among them, L r represents the reconstruction loss, L p is the fully connected neural network loss, L total represents the total loss function, λ is a hyperparameter between 0 and 1, L r represents y and Cosine similarity, y represents the target value, The predicted value S134 is transmitted to the robot controller through the ROS framework.

[0133] As an example, the generated 7D pose is published via a ROS topic for use by the robot arm control node. After obtaining the target pose, the robot arm control node first uses the robot arm's inverse kinematics solver to convert the 7D pose into joint angles. Then, collision-free path planning is performed based on the ROS MoveIt! framework and the open-source motion planning library (OMPL). After motion planning is complete, the robot arm's end effector begins executing the task. The planned joint trajectory is sent to the robot arm control node via the ROS interface, driving the robot arm to the target pose.

[0134] S13. The disinfection robot drives the robotic arm based on the disinfection strategy to perform the disinfection operation, monitors the environmental information of the target disinfection area in real time during the disinfection process, and records the disinfection data of the entire disinfection process;

[0135] S14. After the disinfection is completed, a disinfection result evaluation report for this disinfection task is generated based on the recorded disinfection data;

[0136] As an example, a process log is used to generate a full-process disinfection task report, including key data such as path trajectory, disinfection parameters, and microbial attenuation curve, and saved to an external device for optimization in future tasks. For example, if it is found that the diffusion rate of hydrogen peroxide is slow in a high-humidity environment, the spray time can be increased during the next disinfection. If the angle of ultraviolet radiation is limited in a small space, the movement strategy of the robotic arm can be optimized to improve the disinfection efficiency. In actual application, the concentration of harmful substances in the disinfection area is detected based on the actual disinfection needs to determine whether the disinfection is completed and the disinfection work is terminated. For example, a safety threshold for the concentration of pathogenic microorganisms is preset. When the safety threshold is not exceeded for three consecutive tests, the disinfection work is terminated.

[0137] Traditional pest control technologies typically rely on a single method, such as ultraviolet light, chemical spray, or plasma. In contrast, the present invention integrates three modes: aerosol chemical disinfection, ultraviolet light disinfection, and plasma disinfection. Leveraging a proprietary large language model, it can automatically switch between or simultaneously activate multiple modes based on environmental and task requirements, significantly improving pest control effectiveness and efficiency.

[0138] S15. Return to the standby point and wait to receive the next round of disinfection tasks.

[0139] As an example, during the disinfection process, sensors continuously collect environmental data, detect the residual microorganisms in the air after disinfection, and evaluate the changes in pollutants before and after disinfection. Based on this feedback data, the disinfection strategy is dynamically adjusted. For example, if it detects that the contamination has not been completely removed, the disinfection time will be automatically extended or the disinfection intensity will be increased; if it detects that the effectiveness of ultraviolet disinfection is limited (for example, the light is blocked by an obstacle), the angle of the robotic arm will be automatically adjusted to ensure more comprehensive coverage.

[0140] As a possible implementation method, the following method is used to automatically construct a 3D semantic map and a 2D grid map of the disinfection space:

[0141] S20. Use LiDAR to obtain environmental point cloud data of the disinfecting space, and use a depth camera to collect color and depth information of the disinfecting space;

[0142] S21. Perform spatiotemporal calibration and registration of color and depth information to obtain registered color and depth information;

[0143] S22. Using the registered color and depth information to color and density-complement the environmental point cloud data to obtain a dense color point cloud;

[0144] S23. Perform semantic segmentation on the dense color point cloud using a lightweight RandLA-Net network to identify polluted areas, fixed obstacles, and traversable areas, thereby obtaining a three-dimensional semantic map.

[0145] S24. Project the fixed obstacles and traversable areas in the three-dimensional semantic map onto a two-dimensional plane to obtain a two-dimensional grid map.

[0146] In the second aspect, the present invention provides a multi-mode adaptive disinfecting robot for executing the multi-mode adaptive disinfecting method provided in the first aspect, see Figure 3 The multi-mode adaptive disinfection robot includes: a mobile unit 1, a robot body 2, an environment perception unit 3, a disinfection unit 4, a robotic arm collaboration unit 5 and a control unit 6;

[0147] The environmental sensing unit 3 collects environmental information in real time and transmits it to the control unit 6. The environmental information includes: spatial structure, contaminated area, pathogenic microorganism content, object type and position;

[0148] The control unit 6 includes a host computer 60 and a slave computer 61; the host computer 60 is used to process environmental information, formulate navigation paths and disinfection strategies, and generate motion control instructions, which include manipulator motion control instructions and mobile unit motion control instructions; the manipulator motion control instructions are used to control the manipulator to perform surface disinfection and emergency treatment of spills, and the mobile unit motion control instructions are used for autonomous navigation and dynamic obstacle avoidance;

[0149] The lower computer 61 receives the motion control instructions issued by the upper computer 60, and drives the mobile unit 1, the disinfection unit 4 and the robotic arm collaboration unit 5 to perform corresponding tasks.

[0150] During implementation, control unit 6 generates a corresponding disinfection strategy and autonomous navigation path based on environmental information, converts it into control instructions, and transmits them to slave computer 61. Slave computer 61 coordinates and controls mobile unit 1, disinfection unit 4, and robotic arm collaboration unit 5 to perform disinfection tasks. Specifically, mobile unit 1 drives robot body 2 to the target disinfection area based on the received mobile unit motion control instructions, and robotic arm collaboration unit 5 drives the dual robotic arms to operate collaboratively based on the received robotic arm motion control instructions. Simultaneously, slave computer 61 collects multi-source data in real time, including encoders, IMUs, and robotic arm joint position and torque, and transmits it back to master computer 60 to support the system's closed-loop control and dynamic response.

[0151] See also Figures 3 to 4 The robotic arm collaboration unit 5 includes two six-degree-of-freedom robotic arms 50, each of which is equipped with a quick-release assembly 500. The quick-release assembly 500 can realize the autonomous replacement of the modular end effector 501. The modular end effector 501 has two replaceable working tools, including: a multi-finger gripper and an adaptive suction cup.

[0152] Traditional disinfection robots are typically equipped with only a single robotic arm, making them inefficient when handling complex tasks. This new invention, however, enables dual arms to work collaboratively, effectively handling spills and leaks, significantly improving the robot's performance and flexibility in complex tasks.

[0153] See also Figures 3 to 4 As a possible implementation, the mobile unit 1 includes a chassis 10, a drive motor 11, wheels 12, a transmission assembly 13, a suspension assembly 14 and a power supply 15;

[0154] The drive motor 11 is mounted on the front of the chassis 10 and transmits power to the four wheels 12 via the transmission assembly 13. Suspension assemblies 14 are mounted at the four corners of the chassis and connected to the four wheels 12 to adapt to different terrains. The power supply 15 provides stable power to the control unit 6, the mobile unit 1, and the robotic arm coordination unit 5.

[0155] See also Figures 3 to 4 As a possible implementation, the robot body 2 includes a disassembly component 20, a protective cover 21, a main body component 22 and a storage bin 23;

[0156] The disassembly component 20 is used to realize the rapid connection and disassembly between the mobile unit 1 and the robot body 2. The storage bin 23 is arranged in the cavity of the protective cover 21. The storage bin 23 includes an air inlet, an air outlet and a heat dissipation port, which are used to store the plasma device and overall heat dissipation.

[0157] As a possible implementation method, the environmental perception unit 3 includes: a pathogen detection component 30, a depth camera 31, a laser radar 32, a distance sensor 33 and a temperature and humidity sensor 34; the pathogen detection component 30 is set on the rear side of the main component 22 and is used to detect the concentration of pathogens in the disinfection area in real time. The laser radar 32 and the depth camera 31 are both installed on the front side of the protective cover 21. The depth camera 31 is used to collect RGB images and depth information of the disinfection area, and the laser radar 32 is used to perform three-dimensional reconstruction of the disinfection area; the distance sensor 33 is installed on the front side of the main component 22 for obstacle detection and distance measurement; the temperature and humidity sensor 34 is used to collect temperature and humidity data of the environment so that the aerosol chemical disinfection device 42 can dynamically adjust the spray particle size. For example: in a high temperature and low humidity environment, the droplets evaporate quickly, and larger spray particles (>10μm) can be used to extend the action time; in a low temperature and high humidity environment, the droplets are easy to condense, and smaller particles (<10μm) can be used to improve the stability of the aerosol.

[0158] See also Figures 3 to 7 As a possible implementation, the disinfection unit 4 includes an ultraviolet lamp disinfection device 40, a plasma disinfection device 41 and an aerosol chemical disinfection device 42;

[0159] The UV lamp disinfection device 40 and plasma disinfection device 41 are installed in the storage bin 23, and the aerosol chemical disinfection device 42 is installed on the robot body 2. The installation location is not specifically limited. The UV lamp disinfection device 40 is suitable for surface disinfection; the plasma disinfection device 41 is suitable for air purification and killing pathogenic microorganisms; and the aerosol chemical disinfection device 42 is suitable for space disinfection.

[0160] See also Figure 6As an example, the plasma sterilization device 41 includes a housing, a plasma electrode housed in the internal cavity of the housing, a boost circuit board, a transformer, a fan, and conductive glue. The housing is provided with a through hole and a power interface for plasma active substance diffusion. The plasma electrode is composed of two dielectric tubes and a metal electrode tube for generating dielectric barrier discharge plasma. The boost circuit board uses an inverter boost principle to convert low voltage into an alternating current signal. The power interface is used to introduce an input voltage, and the transformer further amplifies the alternating current signal to achieve high voltage output. Conductive glue is added between the dielectric tube and the metal electrode tube, which is made of silver or graphite to enhance the plasma. After the plasma sterilization device 41 is started, the fan sucks air from the air inlet, enters the plasma discharge cavity to excite plasma, generates a large amount of active substances, effectively degrades pollutants in the air and sterilizes microorganisms. The purified air is further purified by activated carbon adsorption net and organic framework metal material fiber loaded, and finally released to the indoor through the air outlet. The whole process is managed by the control circuit and the time relay to ensure that the device operates efficiently and safely, and achieves the best effect of air purification. A fragrance can also be added to the middle cavity of the housing and placed in the atomizer, which atomizes and volatilizes the fragrance through the atomizer to improve the air freshness.

[0161] The specific implementation of the multi-mode adaptive sterilization robot is as follows: the multi-mode adaptive sterilization robot uses a Velodyne VLP-16 laser radar with a scanning frequency of 20Hz and a ranging accuracy of ±2cm, an Intel RealSense D435i depth camera with a resolution of 1920x1080 and a frame rate of 30fps, an LORDDOM LGCB1000-18GM-D5-V1 ultrasonic range finder with a detection range of 1-6m, and a pathogenic microorganism detection instrument with a sensitivity of 10 3 CFU / m 3 . Start the robot, the control unit controls the battery to power the whole robot, and checks each component. The robot completes remote control through 5G / WiFi dual-mode communication, ensures that the communication delay is less than 200ms, and ensures the stability and efficiency of real-time control. The robot combines multi-modal perception data, decomposes the sterilization task through a large language model, and uses a deep camera to collect environment images on the host computer (Intel NUC), identifies sterilization prohibited areas through a U-Net network, and semantically labels key areas such as experimental tables and biological safety cabinets based on a RandLA-Net network, so as to accurately define the task range. The robot can choose plasma sterilization, aerosol chemical sterilization, and ultraviolet lamp sterilization, or use a multi-mode fusion strategy to improve sterilization efficiency.

[0162] In this embodiment, the plasma sterilization device uses a full solid-state fast pulse generator (operating frequency 10-50 kHz) to drive the dielectric barrier discharge structure (electrode spacing 0.5-2 mm), which efficiently generates low-temperature plasma, and the active oxygen species concentration can reach 107-109 CFU / m 3 , realizing air purification and pathogenic microorganism sterilization. The aerosol chemical sterilization device integrates a two-fluid atomizing nozzle (atomized particle size 5-50 μm) and a centrifugal separator (rotational speed 3000-8000 rpm), and intelligently switches between hydrogen peroxide or chlorine-containing sterilizing agents according to pathogen detection data in the air, adapting to different pollution environments. The ultraviolet lamp sterilization device is composed of a honeycomb array of short-wave ultraviolet light (UV-C) and medium-wave ultraviolet light (UV-B) beads, with adjustable radiation angle (30°-60°). The radiation dose (50-500 μW / cm 2 ) is adjusted in real time by an ultraviolet irradiance meter to ensure efficient sterilization. A trained large language model is called to generate a sterilization strategy.

[0163] When performing sterilization operations, the robot determines whether the pollution in the sterilization area is a biological leak or a spill based on environmental information. If it is a biological leak, a single robot arm is used to perform sterilization operations; if it is a spill, two robot arms are used to perform sterilization operations cooperatively. After sterilization is completed, the robot performs a reset operation, including robot arm reset, ultraviolet lamp cooling, aerosol chemical sterilization device cleaning, etc., and returns to the robot standby point.

[0164] The multi-mode adaptive sterilization robot provided in this embodiment adopts a multi-task adaptive design, can simultaneously perform multiple tasks such as sterilization, detection, path planning, etc., and has autonomous decision-making and task scheduling capabilities, significantly improving the working efficiency of the robot in complex environments.

[0165] Although the present application is described herein in conjunction with various embodiments, other variations of the disclosed embodiments can be understood and implemented by those skilled in the art by referring to the drawings, the disclosure, and the drawings description, etc. In the specification, the word "comprising" does not exclude other components or steps, and "one" or "a" does not exclude multiple. A single processor or other unit can implement several functions listed in the specification. Some measures are described in different embodiments, but this does not mean that these measures cannot be combined to produce good results.

[0166] Although the present invention has been described with reference to specific features and embodiments thereof, it will be apparent that various modifications and combinations thereof may be made without departing from the spirit and scope of the invention. Accordingly, this specification and drawings are merely illustrative of the present invention and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the invention. It will be apparent that various modifications and variations of the present invention may be made by those skilled in the art without departing from the spirit and scope of the invention. Thus, the present invention is intended to include such modifications and variations as fall within the scope of the invention and its equivalents.

Claims

1. A multi-mode adaptive disinfecting method, characterized in that: A multi-mode adaptive disinfection robot is used to execute the multi-mode adaptive disinfection method. When the disinfection robot enters the disinfection space for the first time, it automatically constructs a three-dimensional semantic map and a two-dimensional grid map of the disinfection space. After the construction is completed, it docks at a standby point and waits to receive disinfection tasks. The disinfection robot performs the following steps after receiving each disinfection task: S10 based on the two-dimensional grid map planning path, and according to the planned path to move to the target disinfecting area; S11 collect environmental information of the target area of ​​​​elimination, and update the three-dimensional semantic map in real time; The environmental information includes: spatial structure, contaminated area, pathogenic microorganism content, object type and position; S12 based on the received disinfecting task and the updated three-dimensional semantic map, the use of private large language model to generate disinfecting strategy, the disinfecting strategy includes: selecting disinfecting mode, determining the disinfecting order and disinfecting duration; S13. The robot drives the robotic arm on board based on the disinfection strategy to perform disinfection operations, monitors the environmental information of the target disinfection area in real time during disinfection, and records disinfection data throughout the disinfection process; S14. After the disinfection is completed, an evaluation report of the disinfection results of the disinfection task is generated based on the recorded disinfection data; S15. Return to the standby point and wait for the next round of disinfection tasks; The following method is used to automatically construct a 3D semantic map and a 2D grid map of the disinfection space: S20. Use LiDAR to obtain environmental point cloud data of the disinfecting space, and use a depth camera to collect color and depth information of the disinfecting space; S21 performs spatiotemporal calibration and registration of the color and depth information to obtain the registered color and depth information; S22. Using the registered color and depth information to color and density-complement the environmental point cloud data to obtain a dense color point cloud; S23. Perform semantic segmentation on the dense color point cloud using a lightweight RandLA-Net network to identify contaminated areas, fixed obstacles, and traversable areas, thereby obtaining a 3D semantic map. S24. Projecting the fixed obstacles and traversable areas in the three-dimensional semantic map onto a two-dimensional plane to obtain a two-dimensional grid map.

2. The multi-mode adaptive disinfecting method according to claim 1, characterized in that: An improved hybrid A* algorithm is used to plan a path based on the two-dimensional grid map. The hybrid A* algorithm is: Where, represents the global path planning cost function, represents the path length cost, represents the obstacle avoidance cost, Indicates the cost of the elimination task, represents the kinematic constraint cost, is the path length weight, is the obstacle avoidance weight, The weight of the elimination task, is the kinematic constraint weight; Improved disinfection mission cost Expressed as: in, express The pollution probability of the location is obtained based on the 3D semantic map and has a value range of (0, 1); represents the indicator function, where the undisinfected area is 1 and the disinfected area is 0); Priority of controlling pollution areas, Balance global coverage, with the optimization goal of prioritizing high-pollution areas while maximizing disinfection coverage; During the movement of the disinfection robot, near-field dynamic obstacles are detected in real time to generate local distance field data; ultrasonic ranging data and lidar data are fused into multi-sensor data, and the occupancy probability of the two-dimensional grid map is dynamically corrected using Bayesian updating; when a sudden obstacle is detected, local path replanning is triggered, and a local real-time obstacle avoidance path is generated based on the TEB algorithm.

3. The multi-mode adaptive disinfecting method according to claim 1, characterized in that: The disinfection modes include: plasma disinfection mode, ultraviolet lamp disinfection mode and aerosol chemical disinfection mode; the plasma disinfection mode adopts an atmospheric pressure cold plasma jet method, and uses an all-solid-state fast pulse generator to drive dielectric barrier discharge to generate active oxygen particles, destroy the microbial membrane structure, and achieve disinfection.

4. The multi-mode adaptive disinfecting method according to claim 1, characterized in that: The following method is used to train the large language model: S120. Construct a corpus of disinfection tasks, clean and label the corpus content, and form training samples for supervised learning; S121. Select the general large language model as the base model and load the pre-trained weights. S122. Based on the provided training samples, use supervised fine-tuning methods to fine-tune the selected base model to the target domain, enabling it to generate compliant disinfection strategies based on the input structured / unstructured data. S123. Introducing human feedback reinforcement learning mechanism to optimize model output.

5. The multi-mode adaptive disinfecting method according to claim 3, characterized in that: The disinfection tasks include daily disinfection tasks and emergency disinfection tasks; After receiving daily disinfection tasks, the disinfection robot generates the following disinfection strategy: Determine whether there are people in the target disinfection area. If there are people, select the plasma disinfection mode; if there are no people, select a composite disinfection mode including the ultraviolet lamp disinfection mode, the plasma disinfection mode and the aerosol chemical disinfection mode; Calculate the disinfection time based on the pathogenic microorganism content in the target disinfection area; Use a single robotic arm to perform surface disinfection operations; After receiving the emergency disinfection task, the disinfection robot generates the following disinfection strategy: A vision-guided dual-arm collaborative control algorithm is used to drive the collaborative operation of two robotic arms to perform emergency treatment in the spill area, including laying paper, wiping, and cleaning. After the emergency treatment is completed, select the aerosol chemical disinfection mode for disinfection; Calculate the disinfection time based on the pathogenic microorganism content in the spill area.

6. The multi-mode adaptive disinfecting method according to claim 5, characterized in that: The dual-arm collaborative control algorithm based on vision guidance is used to drive the collaborative operation of the dual robotic arms, including the following steps: S130 real-time collection of RGB images and depth information of the spill area, and converted into point cloud data for three-dimensional reconstruction of the spill area; S131. Discretize the point cloud data into a voxel grid, process the voxel grid using a 3D convolution-based UNet 3D network, and predict a Q-value for each voxel; the Q-value represents the long-term benefit of selecting an action in the current state. S132. Perform local voxelization on the area surrounding the voxel index with the largest Q value, construct a new Q value prediction network, and determine the translational motion of the robotic arm based on the local voxelized maximum Q value index; S133. Discretize the manipulator's rotation into rotational indices every 5°. Use a fully connected neural network to predict the manipulator's rotation and gripper motion, generating a 7D pose for controlling the manipulator's motion. S134. Transmit the 7D pose command to the robotic arm controller through the ROS framework.

7. A multi-mode adaptive disinfection robot, characterized in that: Used to execute the multi-mode adaptive disinfecting method according to any one of claims 1 to 6, the multi-mode adaptive disinfecting robot comprises: a mobile unit, a robot body, an environment perception unit, a disinfecting unit, a robotic arm collaboration unit and a control unit; The environmental perception unit collects environmental information in real time and transmits it to the control unit. The environmental information includes: spatial structure, contaminated area, pathogenic microorganism content, object type and posture; The control unit includes an upper computer and a lower computer; the upper computer generates a corresponding disinfection strategy and autonomous navigation path based on environmental information, and converts it into a control instruction and sends it to the lower computer. The lower computer coordinates and controls the mobile unit, the disinfection unit and the robotic arm collaboration unit to perform the disinfection task based on the control instruction; The robotic arm collaboration unit includes two six-degree-of-freedom robotic arms, each of which is equipped with a quick-release assembly to quickly replace a modular end effector. The modular end effector has two replaceable working tools, including a multi-finger gripper and an adaptive suction cup.

8. The multi-mode adaptive disinfection robot according to claim 7, characterized in that: The environmental perception unit includes: a pathogen detection component, a depth camera, a laser radar, a distance sensor, and a temperature and humidity sensor; The laser radar is used to perform three-dimensional reconstruction of the target disinfection area, the depth camera is used to collect RGB images and depth information of the target disinfection area, the distance sensor is used for obstacle detection and distance measurement, the pathogen detection component is used to detect the concentration of pathogens in the target disinfection area in real time, and the temperature and humidity sensor is used to collect temperature and humidity data of the target disinfection area.

9. The multi-mode adaptive disinfection robot according to claim 7, characterized in that: The host computer is used to process environmental information, formulate navigation paths and disinfection strategies, and generate motion control instructions, which include robotic arm motion control instructions and mobile unit motion control instructions; the robotic arm motion control instructions are used to control the robotic arm to perform surface disinfection and emergency treatment of spills, and the mobile unit motion control instructions are used for autonomous navigation and dynamic obstacle avoidance; The lower computer receives motion control instructions issued by the upper computer, drives the mobile unit, disinfection unit and robotic arm collaboration unit to perform corresponding tasks, and collects multi-source data in real time and transmits it back to the upper computer to support the closed-loop control and dynamic response of the system.

Images