Medical record robot closed-loop control method for disinfection-transfer collaborative optimization and robot

Abstract

The invention discloses a disinfection-transfer collaborative optimization medical record robot closed-loop control method and a robot, and belongs to the field of medical record transfer, and the method comprises the following steps: S1, constructing a hospital environment real-time dynamic model; s2, on the basis of the hospital environment real-time dynamic model, an environment model and task requirements are combined, and an optimal multi-machine collaborative transfer path is generated through double-layer reinforcement learning; s3, considering the optimal multi-machine collaborative transfer path, the road surface state, the environment state and the current state of the robot, and dynamically regulating and controlling magneto-rheological suspension damping; meanwhile, the disinfection parameters are dynamically adjusted in combination with the medical record state and the optimal transfer path. According to the disinfection-transfer collaborative optimization medical record robot closed-loop control method and the robot, a dynamic environment perception-path planning-motion control-disinfection collaborative-multi-machine collaborative closed-loop control system is constructed, so that deep collaborative optimization of medical record disinfection and transfer is realized; the limitation of the prior art in the aspects of environmental adaptability, dynamic decision and multi-system collaboration is overcome.

Description

Technical Field

[0001] This invention relates to the field of medical record transportation technology, and in particular to a closed-loop control method and robot for medical record robots that optimizes disinfection and transportation. Background Technology

[0002] Hospital medical record management is a core component of medical quality management, and its efficiency and security directly impact the treatment process and patient privacy protection. With the advancement of hospital information technology, there is an urgent need for automation of the entire process of medical record retrieval, disinfection, and archiving. However, current automated robots used for transporting hospital medical records have the following limitations in areas such as environmental perception, path planning, motion control, disinfection coordination, and multi-robot collaboration:

[0003] 1. Current Status and Defects of Environmental Perception Technology: Existing medical record robots mostly use a single sensor for environmental detection, such as LiDAR for obstacle recognition or RGB-D cameras for scene reconstruction. The principle is to obtain static environmental information (such as walls and fixed equipment) through laser point cloud or image feature matching, which has the following defects: (1) Insufficient dynamic environmental perception: The coupling relationship between pedestrian density and road surface condition is not integrated. For example, only the location of pedestrians is identified by vision, without combining speed prediction (such as medical staff passing quickly and patients moving slowly), which leads to frequent adjustments in path planning during peak hours (such as the outpatient peak period from 8 to 10 am), and the robot stall rate exceeds 30%. (2) Ignoring road surface condition: The road surface adhesion coefficient is not introduced to quantify the risks of water accumulation, icing, etc. Traditional methods only judge slippage by wheel speed difference, without associating the influence of water film thickness (h) and temperature (T). In humid environments, the robot vibration amplitude increases by 40%, and the risk of medical record damage increases.

[0004] 2. Current Status and Defects of Path Planning Technology: Existing technologies mostly use the A algorithm or Rapid Random Tree Exploration (RRT) for path planning. The principle is to find the shortest path through heuristic functions (such as Manhattan distance), but there are limitations in hospital scenarios: (1) Poor dynamic adaptability: The two-layer optimization architecture of "task priority-environmental constraints" has not been constructed. For example, emergency cases and ordinary cases use the same path cost function, resulting in a delay of more than 5 minutes in emergency task response. (2) Lack of multi-robot collaboration: When planning paths for multiple robots, only physical distance obstacle avoidance (such as spacing ≥ 0.5m) is considered, and the priority is not dynamically adjusted through reinforcement learning (such as based on real-time load and task urgency), resulting in a conflict rate of more than 25% at intersections.

[0005] 3. Current Status and Defects of Motion Control Technology: Most existing robot suspension systems adopt fixed damping design or single PID control. The principle is to adjust the damping force through feedback from displacement sensors, but the following problems exist: (1) Disconnection from path planning: Suspension parameters (such as the control force of magnetorheological dampers) are not associated with path characteristics (such as turning radius and start-stop acceleration). For example, the damping coefficient that maintains straight-line driving when turning causes the body tilt angle to exceed 8°, increasing the risk of tipping over. (2) Weak road adaptability: Control parameters are not dynamically corrected based on the adhesion coefficient. For example, when the road surface is icy, the damping strategy of dry road surface is still used, the robot braking distance increases by 60%, and collisions are more likely to occur.

[0006] 4. Current Status and Deficiencies of Disinfection and Transportation Coordination Technology: In the existing system, the disinfection module and the transportation path are controlled independently. The principle is to use a fixed UV irradiation time (e.g., 30s) and hydrogen peroxide concentration (e.g., 0.5mg / L), which has the following deficiencies: (1) Fixed parameters: The disinfection parameters are not dynamically adjusted according to the transportation time. For example, when transporting over long distances (e.g., across floors), the disinfection time is not extended, resulting in the residual bacterial count on the surface of the medical records exceeding 10 CFU / cm. 2 (2) Lack of collaborative logic: Disinfection start / end is not synchronized with path nodes (e.g., pre-processing is not started during elevator waiting time), resulting in an increase of 20% in the overall process time.

[0007] In summary, existing technologies have significant limitations in terms of the comprehensiveness of environmental perception, the dynamic adaptability of path planning, the linkage between motion control and path, the synergy of disinfection and transportation, and the security of multi-machine evidence storage. Summary of the Invention

[0008] The purpose of this invention is to provide a closed-loop control method and robot for medical record robots that optimize disinfection and transportation, thereby solving the aforementioned technical problems.

[0009] To achieve the above objectives, this invention provides a closed-loop control method for a medical record robot with optimized disinfection and transport functions, comprising the following steps:

[0010] S1. Dynamic Environment Perception and Multimodal Modeling: Construct a real-time dynamic model of the hospital environment that includes pedestrian density field, road surface adhesion coefficient, elevator real-time status, and obstacle trajectory prediction.

[0011] S2. Dynamic path planning based on reinforcement learning: Based on a real-time dynamic model of the hospital environment, combined with the environment model and task requirements, the optimal multi-machine collaborative transport path is generated through two-layer reinforcement learning.

[0012] S3. Dynamic suspension and disinfection control: Considering the optimal multi-machine collaborative transfer path, road surface condition, environmental condition and robot current state, dynamically adjust the magnetorheological suspension damping;

[0013] Simultaneously, disinfection parameters are dynamically adjusted based on the patient's condition and the optimal transport route.

[0014] Therefore, the beneficial effects of the closed-loop control method and robot for medical record robots with the above-mentioned disinfection-transfer collaborative optimization are as follows:

[0015] 1. Enhance the comprehensiveness and dynamism of environmental perception: By fusing multiple sources of sensors, a multimodal environmental model is constructed, which includes pedestrian density field, road surface adhesion coefficient, and obstacle trajectory prediction. This solves the problem that existing single sensors are insufficient for perceiving dynamic pedestrian flow and road surface conditions, reducing the robot's stall rate to below 10% and reducing vibration amplitude by 40%.

[0016] 2. Enhance the dynamic adaptability and multi-machine collaboration of path planning: Based on two-layer reinforcement learning and Hungarian algorithm, a cost model integrating "task priority-path conflict rate" is constructed to achieve priority response for emergency tasks (response delay ≤ 2 minutes) and reduce the conflict rate of multi-machine intersections to below 5%, overcoming the shortcomings of poor dynamic adaptability and lack of multi-machine collaboration in traditional path planning.

[0017] 3. Improve the stability of motion control and road adaptability: Link the magnetorheological suspension system with path characteristics (turning radius, road adhesion coefficient) and use PID combined with road adaptive correction strategy to make the body roll angle ≤3° and reduce the braking distance by 40%, which solves the problems of existing suspension systems being disconnected from the path and having weak road adaptability.

[0018] 4. Achieve efficient coordination between disinfection and transportation: Dynamically adjust disinfection parameters (UV time, hydrogen peroxide concentration) according to transportation time and case status, and start pretreatment synchronously with path nodes (such as elevator waiting time), so that the residual bacterial rate of case records meets 100% and the overall process time is reduced by 25%, breaking through the limitations of fixed disinfection parameters and lack of coordination logic in existing technologies.

[0019] 5. Ensure the security of multi-machine collaboration and the reliability of data traceability: Real-time adjustment of multi-machine load and path conflicts is achieved through dynamic task priority allocation.

[0020] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0021] Figure 1 This is a flowchart of the closed-loop control method for the medical record robot with disinfection-transfer collaborative optimization according to the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the embodiments of the present invention and are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout.

[0023] It should be noted that the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, such as a process, method, system, product, or server that includes a series of steps or units, not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product, or device.

[0024] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0025] like Figure 1 As shown, the closed-loop control method for the medical record robot with disinfection-transfer collaborative optimization includes the following steps:

[0026] S1. Dynamic Environment Perception and Multimodal Modeling: Construct a real-time dynamic model of the hospital environment that includes pedestrian density field, road surface adhesion coefficient, elevator real-time status, and obstacle trajectory prediction.

[0027] Step S1 specifically includes the following steps:

[0028] S11. Synchronous Acquisition of Multi-Source Sensor Data: Synchronously acquires LiDAR point cloud data, RGB-D images, capacitive water film sensors, temperature sensors, and elevator IoT signals. Preprocesses the LiDAR point cloud and RGB-D images, outputting a dataset R = {P} cd ,I de ,d water ,T environment ,L e W e}, where P cd ,L de ,d water ,T encironment ,L e W e These represent the denoised point cloud, the image frame with depth, the water film thickness, the ambient temperature, the real-time elevator floor position, and the load, respectively.

[0029] In step S11, the preprocessing steps for the lidar point cloud are as follows: statistical filtering is used to remove noise points while preserving the obstacle outlines;

[0030] The preprocessing steps for RGB-D images are as follows: use an interpolation algorithm to fill in the missing regions.

[0031] S12. Predict movement trajectory based on pedestrian speed classification;

[0032] S121. Pedestrian Detection: Using the YOLOv8 algorithm to identify depth-bound image frames I de For pedestrian i, output the bounding box and confidence score, and combine the LiDAR point cloud to obtain the 3D coordinates (x, y, y) of pedestrian i. i ,y i ,z i );

[0033] S122, Set the speed classification threshold v th In this embodiment, v th =1.2 m / s, when the average speed v of the pedestrian within the sliding window is... avg Greater than the velocity classification threshold v th Classified as medical staff; otherwise, classified as patients; the average velocity v within the sliding window. avg The calculation formula is as follows:

[0034]

[0035] In the formula, n represents the number of sliding window frames; Δt represents the sampling interval; (x i,t ,y i,t (x) represents the two-dimensional coordinates of the pedestrian at time t; i,t-1 ,y i,t-1 () represents the two-dimensional coordinates of the pedestrian at time t-1;

[0036] S123. A GRU network is used to predict the patient's movement trajectory, and a Kalman filter is used to predict the movement trajectory of medical staff, thus obtaining the predicted pedestrian movement trajectory at time t+k.

[0037] S13, Based on water film thickness d water and ambient temperature T environment Calculate the adhesion coefficient μ(a) water ,T environment ):

[0038]

[0039] S14. Estimate elevator waiting time t elevator :

[0040]

[0041] In the formula, L current Indicates the current floor the robot is on; v elevator Indicates the elevator's operating speed; t open This indicates the elevator door opening time.

[0042] S2. Dynamic path planning based on reinforcement learning: Based on a real-time dynamic model of the hospital environment, combined with the environment model and task requirements, the optimal multi-machine collaborative transport path is generated through two-layer reinforcement learning.

[0043] Step S2 specifically includes the following steps:

[0044] S21. Task Priority Quantification: Dynamically adjust task priority P based on urgency and the distance D between the robot and the target. r :

[0045] P r =ω u ×urgency+ω D ×(1-D norm );

[0046] In the formula, D norm Let D represent the distance between the robot and the target after normalization, and D norm =D / D max D max ω represents the maximum distance; u and ω D Both represent weighting factors, and ω u +ω D =1;

[0047] S22, Integrate distance, pedestrian flow, road surface, and elevator costs, using the total cost function C total Advantages and disadvantages of quantification approaches:

[0048]

[0049] In the formula, L represents the path length; ρ(x,y,t) represents the pedestrian density, and n represents the number of people detected, σ represents the spatial diffusion coefficient, and (x,y) represents the planar coordinates of discrete nodes on the path.

[0050] S23. Consider the total cost function C total Multiple robot positions S = [s1, s2, ..., s m And the predicted pedestrian movement trajectory at time t+k. Where s1, s2, ..., s mLet Γ represent the positions of the 1st, 2nd, ..., mth robots respectively. Using an improved A algorithm combined with time window constraints, a multi-robot cooperative transport path Γ = [Γ1, Γ2, ..., Γm] is generated. m ],Γ1,Γ2,...,Γ m These represent the paths of the 1st, 2nd, ..., mth robots generated, respectively.

[0051] Step S23 specifically includes the following steps:

[0052] S231, Path Search: Based on the improved A algorithm, using the local cost function C part As a heuristic function, search for the initial path;

[0053] Step S231 specifically includes the following steps:

[0054] S2311. Initialize nodes and cost table;

[0055] S23111. Discretize the hospital environment into grid nodes, and each node corresponds to physical coordinates n. o = (x, y, z), and set the starting node n. o,start =S start = (x0, y0, z0), target node n o,target =S target =(x target ,y target ,z target ), where S start and S target Let (x0, y0, z0) represent the robot's starting and stopping positions, respectively. Let (x0, y0, z0) represent the initial three-dimensional coordinates. target ,y target ,z target () represents the three-dimensional coordinates of the target location;

[0056] S23112, Calculate the total cost f(n) o ):

[0057] f(n o )=g(n o )+h(n o );

[0058] In the formula, g(n) o ) indicates starting from the initial node n o,start To the current node n o,current The cumulative cost, which is the actual cost, and the initial cost g(n) o,start h(n) = 0; o ) indicates starting from the current node n o,current to target node n o,target The estimated cost is the heuristic cost;

[0059] S23113, Set the starting node n o,start and its corresponding f(n) o ), g(n o ) and h(n o Store in the OPEN table;

[0060] S2312, Node Expansion and Local Cost Function Calculation;

[0061] S23121. Select f(n) from the OPEN table. o The current minimum node n) o,current,min And transfer it to the CLOSED table, while determining n o,current,min Is it n? o,target If not, proceed to step S23122; if yes, proceed to step S2314.

[0062] S23122, Extended n o,current,min From the 8 adjacent nodes, invalid nodes that exceed the map boundary or are less than 0.5m away from static obstacles are filtered out, resulting in n valid adjacent nodes. o,current,neighbor ;

[0063] S23123, Calculate the current minimum node n o,current,min With adjacent valid nodes n o,current,neighbor Local cost C between part :

[0064]

[0065] In the formula, λ1, λ2, and λ3 all represent weighting coefficients, and λ1 = 1.0, λ2 = 0.7, and λ3 = 0.2; λ1·L cn L represents distance cost. cn Represents the current smallest node n o,current,min With adjacent valid nodes n o,current,neighbor Planar distance between them; Let v represent the road surface incentive cost, v represent the robot's planned travel speed on the current road segment, and z represent the road surface incentive cost. q (t) represents the road surface elevation, and f0 represents the medium frequency, f0 = 0.1 Hz; s f represents the standard spatial frequency. s =0.1m -1 G q (f s G represents the road surface roughness coefficient. q (f s ) = 64 × 10 -6 m 3 W(t) is Gaussian white noise; Indicates the cost of speed adaptation, v max v represents the robot's maximum speed. max =1.2m / s;

[0066] S2313, Cost Update and Dynamic Velocity Planning;

[0067] S23131, Updating actual costs and heuristic costs:

[0068] g(n neighbor )=g(n o )+C part ;

[0069]

[0070] In the formula, g(n) neighbor ) and h(n neighbor ) represent the actual cost and heuristic cost after the update, respectively; L g Indicates the number of adjacent valid nodes n o,current,neighbor With target node n o,target The straight-line distance between them; max(z) r ) represents the maximum bump height, and max(z) r ) = 0.1m; v avg Let v represent the robot's average speed. avg =0.8m / s;

[0071] S23132. Determine if there is a transient turbulence height z. r If the value is ≥0.05m, then reduce v to 0.5m / s-0.8m / s and update the local cost function C accordingly. part The possible values ​​of v;

[0072] S2314, Path Generation and Verification;

[0073] When n o,target When a node is added to the CLOSED table, the initial path Γ is generated by tracing back through the parent node pointer. init ;

[0074] Simultaneously verify the following indicators:

[0075] Peak road excitation: maximum bump height (z) r ≤0.1m;

[0076] Speed ​​rationality: Planned speed v for bumpy road sections bumpy ≤0.8m / s, planned speed v on straight road sections straight ≤1.2m / s;

[0077] Actual total cost C actual,total=g(n o,tatget ), g(n o,target ) indicates starting from the initial node n o,start to target node n o,target The actual cost.

[0078] S232, Time Window Collision Avoidance: Assign a time window [t] to each node on the initial path. enter ,t exit This ensures that the distances between multiple robots and between a robot and a pedestrian are both greater than safe distances. If a collision is predicted, reinforcement learning is used to adjust the robot's speed v. adj :

[0079]

[0080] In the formula, v0 represents the robot's initial speed; d represents the distance between multiple robots or the distance between a robot and a pedestrian; σ d The spatial diffusion coefficient representing distance; t enter ,t exit These represent the time when the robot enters the path node and the time when the robot leaves the path node, respectively.

[0081] S233. Multi-machine task allocation: based on the modified task priority P′ r and path conflict rate O IJ The Hungarian algorithm is used to generate multi-machine collaborative transport paths.

[0082] Step S233 specifically includes the following steps:

[0083] S2331. Adjusting task priority based on robot load rate:

[0084] P′ r =P r / N r ×(1-L r );

[0085] In the formula, N r Indicates the number of robots; L r Indicates load rate;

[0086] S2332, Calculate the path conflict rate O IJ :

[0087]

[0088] In the formula, O IJE This represents the overlap rate between the pre-planned path of robot I and robot E when robot I performs task J, and Γ IJ,init Indicates the initial path of robot I when performing task J; ΓEJ′,init This represents the initial path of robot E when it performs task J′;

[0089] S2333, Integrating priority correction and conflict rate, constructing the cost matrix of the Hungarian algorithm. C IJ C represents the cost of robot I performing task J. IJ =ω α ×(1-P′ r,IJ )+ω β ×O IJ ω α and ω β All are weighting coefficients, (1-P′) r,IJ ) represents the task priority cost, P′ r,IJ N represents the corrected task priority when robot I executes task J; J Indicates the number of tasks;

[0090] S2334, Regarding the cost matrix Normalize and cover the cost matrix with the fewest lines. If all zero elements in the equation are equal to N, then the number of lines is equal to N. r If the zero element position corresponds to the optimal allocation, then proceed to step S2336.

[0091] S2335. Find the minimum value d among the uncovered elements. Subtract d from the uncovered rows and add d to the covered columns. Repeat step S2334 until the optimal allocation is found, obtaining the optimal allocation matrix A. This is the optimal allocation matrix A when robot I performs task J. IJ The corresponding normalized cost matrix C′ IJ ;

[0092] S2336. Dynamically verify and adjust the allocation results:

[0093] If there exists O in the optimal allocation IJ A value ≥0.1 is considered a high-conflict pair. In this case, other task assignments are fixed, and only the cost matrix of this task and other robots is recalculated. The Hungarian algorithm is used for local optimization until O(n). IJ <0.1;

[0094] Meanwhile, if a high-conflict event involves priority P′ r,IJ If the conflict rate exceeds the set level, it will be forcibly reduced to 0. IJ ≤0.05, allowing for an increase of ≤10% in the cost of other tasks.

[0095] S3. Dynamic suspension and disinfection control: Considering the optimal multi-machine collaborative transfer path, road surface condition, environmental condition and robot current state, dynamically adjust the magnetorheological suspension damping;

[0096] Simultaneously, disinfection parameters are dynamically adjusted based on the patient's condition and the optimal transport route.

[0097] The steps for adjusting the magnetorheological suspension damping described in step S3 are as follows:

[0098] Step 1: Calculate the foundation damping

[0099]

[0100] In the formula, K p K i and K d These represent the proportional, integral, and differential coefficients, respectively; d h and These represent the derivatives of the suspended vertical displacement and the suspended vertical displacement, respectively.

[0101] Step 2: Adjust the damping force based on the adhesion coefficient:

[0102]

[0103] In the formula, F d This indicates the adjusted suspension damping force;

[0104] At the same time, determine whether a turn has occurred; if so, determine the turning radius R. t and speed v t Calculate the roll moment and adjust the left and right damping difference ΔF. d This makes the inner damping F d +ΔF d The outer damping is F d -ΔF d And the damping difference ΔF between the left and right sides d The calculation formula is as follows:

[0105]

[0106] In the formula, M represents the mass of the robot; K t Indicates the proportionality coefficient;

[0107] The steps for dynamically adjusting disinfection parameters are as follows:

[0108] Step 1: Calculate the total disinfection time t d :

[0109] t d =K N N+K H H+K P P d ;

[0110] In the formula, N and H represent the number and thickness of medical records, respectively; P dIndicates the pollution level; K N K H and K P Indicates the corresponding coefficient;

[0111] Step 2: Allocate time:

[0112]

[0113] In the formula, t d,UV and These represent UV irradiation time and hydrogen peroxide fumigation time, respectively.

[0114] Simultaneously, the following disinfection and transportation strategies were customized:

[0115] When the elevator waiting time t wait If the exposure time is ≥30 seconds, pre-disinfection is initiated, with a UV irradiation time of 0.2t. d ;

[0116] When the transit time t trans Update the total disinfection time when >5 minutes have elapsed.

[0117] And add the following constraints:

[0118] t disinfectend =t arrival -10s;

[0119] In the formula, t disinfectend Indicates the end time of disinfection; t arrival Indicates the time taken to reach the target location.

[0120] The robot used for the closed-loop control method of the medical record robot for performing disinfection-transfer collaborative optimization includes a robot body, a state perception and multimodal modeling module, a dynamic path planning module, and a dynamic suspension and disinfection control module set on the robot body.

[0121] in,

[0122] The state perception and multimodal modeling module is used to build a real-time dynamic model of the hospital environment, including pedestrian density field, road surface adhesion coefficient, elevator real-time status, and obstacle trajectory prediction.

[0123] Dynamic path planning is used to generate the optimal multi-machine collaborative transport path based on a real-time dynamic model of the hospital environment, combined with the environment model and task requirements, through two-layer reinforcement learning.

[0124] The dynamic suspension and disinfection control module takes into account the optimal multi-machine collaborative transport path, road surface condition, environmental condition, and the robot's current state, and dynamically adjusts the magnetorheological suspension damping; at the same time, it dynamically adjusts the disinfection parameters based on the patient's condition and the optimal transport path.

[0125] It should be noted that the structural and working principles of the robots described above are common knowledge in the field, and therefore will not be elaborated upon here.

[0126] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A closed-loop control method for a medical record robot with optimized disinfection and transport functions, characterized by: Includes the following steps: S1. Dynamic Environment Perception and Multimodal Modeling: Construct a real-time dynamic model of the hospital environment that includes pedestrian density field, road surface adhesion coefficient, elevator real-time status, and obstacle trajectory prediction. S2. Dynamic path planning based on reinforcement learning: Based on a real-time dynamic model of the hospital environment, combined with the environment model and task requirements, the optimal multi-machine collaborative transport path is generated through two-layer reinforcement learning. S3. Dynamic suspension and disinfection control: Considering the optimal multi-machine collaborative transfer path, road surface condition, environmental condition and robot current state, dynamically adjust the magnetorheological suspension damping; Simultaneously, disinfection parameters are dynamically adjusted based on the patient's condition and the optimal transport route.

2. The closed-loop control method for the medical record robot with disinfection-transfer collaborative optimization according to claim 1, characterized in that: Step S1 specifically includes the following steps: S11. Synchronous Acquisition of Multi-Source Sensor Data: Synchronously acquires LiDAR point cloud data, RGB-D images, capacitive water film sensors, temperature sensors, and elevator IoT signals. Preprocesses the LiDAR point cloud and RGB-D images, outputting a dataset R = {P} cd ,I de ,d water ,T environment ,L e W e }, where P cd ,I de ,d water ,T environment ,L e W e These represent the denoised point cloud, the image frame with depth, the water film thickness, the ambient temperature, the real-time elevator floor position, and the load, respectively. S12. Predict movement trajectory based on pedestrian speed classification; S121. Pedestrian Detection: Using the YOLOv8 algorithm to identify depth-bound image frames I de For pedestrian i, output the bounding box and confidence score, and combine the LiDAR point cloud to obtain the 3D coordinates (x, y, y) of pedestrian i. i ,y i ,z i ); S122, Set the speed classification threshold v th When the average speed v of the pedestrians inside the sliding window avg Greater than the velocity classification threshold v th Classified as medical staff; otherwise, classified as patients; the average velocity v within the sliding window. avg The calculation formula is as follows: In the formula, n represents the number of sliding window frames; Δt represents the sampling interval; (x i,t ,y i,t (x) represents the two-dimensional coordinates of the pedestrian at time y; i,t-1 ,y i,t-1 () represents the two-dimensional coordinates of the pedestrian at time t-1; S123. A GRU network is used to predict the patient's movement trajectory, and a Kalman filter is used to predict the movement trajectory of medical staff, thus obtaining the predicted pedestrian movement trajectory at time t+k. S13, Based on water film thickness d water and ambient temperature T enviroment Calculate the adhesion coefficient μ(d) water ,T environment ): S14. Estimate elevator waiting time t elevator : In the formula, L current Indicates the current floor the robot is on; v elevator Indicates the elevator's operating speed; t open This indicates the elevator door opening time.

3. The closed-loop control method for the medical record robot with disinfection-transfer collaborative optimization according to claim 2, characterized in that: In step S11, the preprocessing steps for the lidar point cloud are as follows: statistical filtering is used to remove noise points while preserving the obstacle outlines; The preprocessing steps for RGB-D images are as follows: use an interpolation algorithm to fill in the missing regions.

4. The closed-loop control method for the medical record robot with disinfection-transfer collaborative optimization according to claim 2, characterized in that: Step S2 specifically includes the following steps: S21. Task Priority Quantification: Dynamically adjust task priority P based on urgency and the distance D between the robot and the target. r : P r =ω u ×urgency+ω D ×(1-D norm ); In the formula, D norm Let D represent the distance between the robot and the target after normalization, and D norm =D / D max D max ω represents the maximum distance; u and ω D Both represent weighting factors, and ω u +ω D =1; S22, Integrate distance, pedestrian flow, road surface, and elevator costs, using the total cost function C total Advantages and disadvantages of quantification approaches: In the formula, L represents the path length; ρ(x,y,t) represents the pedestrian density, and n represents the number of people detected, σ represents the spatial diffusion coefficient, and (x,y) represents the planar coordinates of discrete nodes on the path. S23. Consider the total cost function C total Multiple robot positions S = [s1, s2, ..., s m And the predicted pedestrian movement trajectory at time t+k. Where s1, s2, ..., s m Let Γ represent the positions of the 1st, 2nd, ..., mth robots respectively. Using an improved A algorithm combined with time window constraints, a multi-robot cooperative transport path Γ = [Γ1, Γ2, ..., Γm] is generated. m ],Γ1,Γ2,...,Γ m These represent the paths of the 1st, 2nd, ..., mth robots generated, respectively.

5. The closed-loop control method for the medical record robot with disinfection-transfer collaborative optimization according to claim 4, characterized in that: Step S23 specifically includes the following steps: S231, Path Search: Based on the improved A algorithm, using the local cost function C part As a heuristic function, search for the initial path; S232, Time Window Collision Avoidance: Assign a time window [t] to each node on the initial path. enter ,t exit This ensures that the distances between multiple robots and between a robot and a pedestrian are both greater than safe distances. If a collision is predicted, reinforcement learning is used to adjust the robot's speed v. adj : In the formula, v0 represents the robot's initial speed; d represents the distance between multiple robots or the distance between a robot and a pedestrian; σ d The spatial diffusion coefficient representing distance; t enter ,t exit These represent the time when the robot enters the path node and the time when the robot leaves the path node, respectively. S233. Multi-machine task allocation: based on the modified task priority P′ r and path conflict rate O IJ The Hungarian algorithm is used to generate multi-machine collaborative transport paths.

6. The closed-loop control method for the medical record robot with disinfection-transfer collaborative optimization according to claim 4, characterized in that: Step S231 specifically includes the following steps: S2311. Initialize nodes and cost table; S23111. Discretize the hospital environment into grid nodes, and each node corresponds to physical coordinates n. o = (x, y, z), and set the starting node n. o,start =S start = (x0, y0, z0), target node n o,tatget =S target =(x target ,y target ,z target ), where S start and S target Let (x0, y0, z0) represent the robot's starting and stopping positions, respectively. Let (x0, y0, z0) represent the initial three-dimensional coordinates. target ,y target ,z target () represents the three-dimensional coordinates of the target location; S23112, Calculate the total cost f(n) o ): f(n o )=g(n o )+h(n o ); In the formula, g(n) o ) indicates starting from the initial node n o,start To the current node n o,current The cumulative cost, which is the actual cost, and the initial cost g(n) o,start h(n) = 0; o ) indicates starting from the current node n o,current to target node n o,target The estimated cost is the heuristic cost; S23113, Set the starting node n o,start and its corresponding f(n) o ), g(n o ) and h(n o Store in the OPEN table; S2312, Node Expansion and Local Cost Function Calculation; S23121. Select f(n) from the OPEN table. o The current minimum node n) o,current,min And transfer it to the CLOSED table, while determining n o,current,min Is it n? o,target If not, proceed to step S23122; if yes, proceed to step S2314. S23122, Extended n o,current,min From the 8 adjacent nodes, invalid nodes that exceed the map boundary or are less than 0.5m away from static obstacles are filtered out, resulting in n valid adjacent nodes. o,current,neighbor ; S23123, Calculate the current minimum node n o,current,min With adjacent valid nodes n o,current,neighbor Local cost C between part : In the formula, λ1, λ2, and λ3 all represent weighting coefficients, and λ1 = 1.0, λ2 = 0.7, and λ3 = 0.2; λ1·L cn L represents distance cost. cn Represents the current smallest node n o,current,min With adjacent valid nodes n o,current,neighbor Planar distance between them; Let v represent the road surface incentive cost, v represent the robot's planned travel speed on the current road segment, and z represent the road surface incentive cost. q (t) represents the road surface elevation, and f0 represents the medium frequency, f0 = 0.1 Hz; s f represents the standard spatial frequency. s =0.1m -1 G q (f s G represents the road surface roughness coefficient. q (f s ) = 64 × 10 -6 m 3 W(t) is Gaussian white noise; Indicates the cost of speed adaptation, v max v represents the robot's maximum speed. max =1.2m / s; S2313, Cost Update and Dynamic Velocity Planning; S23131, Updating actual costs and heuristic costs: n(n neighbor )=g(n o )+C part ; In the formula, g(n) neighbor ) and h(n neighbor ) represent the actual cost and heuristic cost after the update, respectively; L g Indicates the number of adjacent valid nodes n o,current,neighbor With target node n o,target The straight-line distance between them; max(z) r ) represents the maximum bump height, and max(z) r ) = 0.1m; v avg Let v represent the robot's average speed. avg =0.8m / s; S23132. Determine if there is a transient turbulence height z. r If the value is ≥0.05m, then reduce v to 0.5m / s-0.8m / s and update the local cost function C accordingly. part The possible values ​​of v; S2314, Path Generation and Verification; When n o,target When a node is added to the CLOSED table, the initial path Γ is generated by tracing back through the parent node pointer. init ; Simultaneously verify the following indicators: Peak road excitation: maximum bump height (z) r ≤0.1m; Speed ​​rationality: Planned speed v for bumpy road sections bumpy ≤0.8m / s, planned speed v on straight road sections straight ≤1.2m / s; Actual total cost C actual,total =g(n o,target ), g(n o,target ) indicates starting from the initial node n o,start to target node n o,target The actual cost.

7. The closed-loop control method for the medical record robot with disinfection-transfer collaborative optimization according to claim 6, characterized in that: Step S233 specifically includes the following steps: S2331. Adjusting task priority based on robot load rate: P′ r =P r / N r ×(1-L r ); In the formula, N r Indicates the number of robots; L r Indicates load rate; S2332, Calculate the path conflict rate O IJ : In the formula, O IJE This represents the overlap rate between the pre-planned path of robot I and robot E when robot I performs task J, and Γ IJ,init Indicates the initial path of robot I when performing task J; Γ EJ′,init This represents the initial path of robot E when it performs task J′; S2333, Integrating priority correction and conflict rate, constructing the cost matrix of the Hungarian algorithm. C IJ C represents the cost of robot I performing task J. IJ =ω α ×(1-P′ r,IJ )+ω β ×O IJ ω α and ω β All are weighting coefficients, (1-P′) r,IJ ) represents the task priority cost, P′ r,IJ N represents the corrected task priority when robot I executes task J; J Indicates the number of tasks; S2334, Regarding the cost matrix Normalize and cover the cost matrix with the fewest lines. If all zero elements in the equation are equal to N, then the number of lines is equal to N. r If the zero element position corresponds to the optimal allocation, then proceed to step S2336. S2335. Find the minimum value d among the uncovered elements. Subtract d from the uncovered rows and add d to the covered columns. Repeat step S2334 until the optimal allocation is found, obtaining the optimal allocation matrix A. This is the optimal allocation matrix A when robot I performs task J. IJ The corresponding normalized cost matrix C′ IJ ; S2336. Dynamically verify and adjust the allocation results: If there exists O in the optimal allocation IJ A value ≥0.1 is considered a high-conflict pair. In this case, other task assignments are fixed, and only the cost matrix of this task and other robots is recalculated. The Hungarian algorithm is used for local optimization until O(n). IJ <0.1; Meanwhile, if a high-conflict event involves priority P′ r,IJ If the conflict rate exceeds the set level, it will be forcibly reduced to 0. IJ ≤0.05, allowing for an increase of ≤10% in the cost of other tasks.

8. The closed-loop control method for the medical record robot with disinfection-transfer collaborative optimization according to claim 7, characterized in that: The steps for adjusting the magnetorheological suspension damping described in step S3 are as follows: Step 1: Calculate the foundation damping In the formula, K p K i and K d These represent the proportional, integral, and differential coefficients, respectively; d h and These represent the derivatives of the suspended vertical displacement and the suspended vertical displacement, respectively. Step 2: Adjust the damping force based on the adhesion coefficient: In the formula, F d This indicates the adjusted suspension damping force; At the same time, determine whether a turn has occurred; if so, determine the turning radius R. t and speed v t Calculate the roll moment and adjust the left and right damping difference ΔF. d This makes the inner damping F d +ΔF d The outer damping is F d -ΔF d And the damping difference ΔF between the left and right sides d The calculation formula is as follows: In the formula, M represents the mass of the robot; K t Indicates the proportionality coefficient; The steps for dynamically adjusting disinfection parameters are as follows: Step 1: Calculate the total disinfection time t d : t d =K N N+K H H+K P P d ; In the formula, N and H represent the number and thickness of medical records, respectively; P d Indicates the pollution level; K N K H and K P Indicates the corresponding coefficient; Step 2: Allocate time: In the formula, t d,UV and These represent UV irradiation time and hydrogen peroxide fumigation time, respectively. Simultaneously, the following disinfection and transportation strategies were customized: When the elevator waiting time t wait If the exposure time is ≥30 seconds, pre-disinfection is initiated, with a UV irradiation time of 0.2t. d ; When the transit time t trans Update the total disinfection time when >5 minutes have elapsed. And add the following constraints: t disinfectend =t arrival -10s; In the formula, t disinfectend Indicates the end time of disinfection; t arrival Indicates the time taken to reach the target location.

9. A robot for executing the closed-loop control method for disinfection-transfer collaborative optimization of medical record robots according to any one of claims 1-8, characterized in that: It includes the robot body, a state perception and multimodal modeling module set on the robot body, a dynamic path planning module, and a dynamic suspension and disinfection control module; in, The state perception and multimodal modeling module is used to build a real-time dynamic model of the hospital environment, including pedestrian density field, road surface adhesion coefficient, elevator real-time status, and obstacle trajectory prediction. Dynamic path planning is used to generate the optimal multi-machine collaborative transport path based on a real-time dynamic model of the hospital environment, combined with the environment model and task requirements, through two-layer reinforcement learning. The dynamic suspension and disinfection control module takes into account the optimal multi-machine collaborative transport path, road surface condition, environmental condition, and the robot's current state, and dynamically adjusts the magnetorheological suspension damping; at the same time, it dynamically adjusts the disinfection parameters based on the patient's condition and the optimal transport path.

Images