Intelligent navigation system of medical waste cleaning robot

Abstract

The invention discloses an intelligent navigation system of a medical waste cleaning robot, and belongs to the field of intelligent navigation systems, and the system comprises an obstacle state recognition module which is used for recognizing the state of an obstacle object on the advancing track of the robot, and the state of the obstacle object comprises a moving state and a static state; the path planning module is connected with the obstacle state recognition module and used for planning a movement track according to the obstacle object state; when the obstacle object is in a static state, the path planning module generates an arc-shaped bypassing path; when the obstacle object is in a moving state, the path planning module compares the real-time speed of the obstacle object with a preset speed threshold value, if the real-time speed is larger than the threshold value, an emergency state is determined, the emergency obstacle avoidance sub-module is started, and if the real-time speed is smaller than or equal to the threshold value, a non-emergency state is determined. According to the invention, a basic obstacle avoidance decision can be made according to the dynamic attribute of the obstacle, and emergency degree grading can be carried out according to the speed of the moving obstacle when the moving obstacle is encountered, so that the operation efficiency is improved and collision is reduced.

Description

Technical Field

[0001] This invention relates to the field of intelligent navigation systems, specifically to an intelligent navigation system for a medical waste cleaning robot. Background Technology

[0002] With the deep integration of medical informatization and intelligentization, the standardized treatment of medical waste has become a key link in hospital infection control and public health safety. Medical waste disposal robots, as automated equipment replacing manual waste transfer, have their navigation system's safety, efficiency, and reliability directly determining their practical value in the medical environment. The medical environment is characterized by high population density, complex obstacle types, diverse spatial channels, and strong dynamism: hospital corridors, elevator entrances, ward doors, and other areas are constantly filled with various moving and stationary obstacles such as medical staff, patients with mobility impairments, wheelchairs, ambulances, and medical equipment, and these obstacles exhibit significant differences in their status (moving / stationary), movement speed, and avoidance intentions.

[0003] Existing patent 201811333599.5 describes a robot navigation system and method. The robot navigation system includes two or more channel areas, at least one readable tag, and at least one robot. Each robot includes an instruction acquisition unit, a traveling device, a tag reading unit, and a position determination unit. The robot navigation method includes the following steps: an instruction acquisition step, a traveling control step, a tag reading step, and a position determination step.

[0004] Existing technologies suffer from the following problems: traditional systems employ simplistic obstacle avoidance strategies, leading to frequent collisions or low efficiency. Therefore, those skilled in the art have developed an intelligent navigation system for medical waste disposal robots to address the issues raised in the background. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an intelligent navigation system for a medical waste disposal robot, comprising:

[0006] The obstacle state recognition module is used to identify the state of obstacle objects on the robot's forward trajectory. The state of the obstacle objects includes both moving and stationary states.

[0007] The path planning module, connected to the obstacle state recognition module, is used to plan a motion trajectory based on the obstacle object's state. When the obstacle object is stationary, the path planning module generates an arc-shaped detour path. When the obstacle object is moving, the path planning module compares the obstacle object's real-time speed with a preset speed threshold. If the speed is greater than the threshold, it is determined to be an emergency state, and the emergency obstacle avoidance submodule is activated. If the speed is less than or equal to the threshold, it is determined to be a non-emergency state, and the non-emergency obstacle avoidance submodule is activated.

[0008] The speed control module, in conjunction with the path planning module, is used to adjust the robot speed in real time based on the number and speed of surrounding moving obstacles, road slope and flatness, and the amount of garbage loaded by the robot.

[0009] The positioning verification module is used to verify the real-time positioning accuracy, speed consistency, and path deviation through the robot's built-in sensors and distributed cameras. When an anomaly is detected, an anomaly weight value is calculated, and the robot's actions are controlled according to whether the anomaly weight value falls within the preset range [Qmin, Qmax]. If it is less than Qmin, the task continues and repair is performed. If it is greater than Qmax, the task is abandoned immediately and repair is performed. If it is within the range, the path is replanned and the robot is added to the repair point.

[0010] As a further aspect of the present invention: the emergency obstacle avoidance submodule is configured to: scan other obstacle objects within a preset range of the current obstacle object, assign a first weight value to the obstacle object in motion and a second weight value to the obstacle object in stationary state, calculate the sum of the weight values, and if the sum is less than a preset weight threshold, output a motion avoidance command and plan the best avoidance path; if the sum is greater than or equal to the preset weight threshold, output a stationary avoidance command and plan the best obstacle avoidance point, wherein the first weight value is greater than the second weight value;

[0011] The non-emergency obstacle avoidance submodule is configured to: control the robot to issue an early warning alarm, predict the obstacle object's avoidance intention, if avoidance is predicted, the robot will move forward along the original trajectory, if no avoidance is predicted, the robot will output a movement avoidance command; at the same time, identify the obstacle object type, update the weight factor based on historical avoidance behavior, and calculate the total weight factor V. If V is less than a preset type threshold, the avoidance path will be planned in advance.

[0012] As a further aspect of the present invention: the obstacle state recognition module includes a sensor array composed of a lidar, a visual camera and an infrared sensor, as well as a data fusion unit. The data fusion unit uses a Kalman filter algorithm to fuse multi-sensor data and outputs the position, speed and state information of the obstacle. Among them, moving obstacle objects are identified by continuous frame image difference and optical flow method, and stationary obstacle objects are confirmed by point cloud matching and map comparison.

[0013] As a further aspect of the present invention: the arc trajectory generation in the path planning module is based on the Bézier curve algorithm, and the curve control points are dynamically adjusted according to the obstacle boundary and the robot's turning radius to ensure that the avoidance distance is not less than the safety threshold; the optimal avoidance path planning adopts the A* search algorithm with the goal of minimizing task delay.

[0014] As a further aspect of the present invention: the weight value calculation in the emergency obstacle avoidance submodule adopts a mathematical expression:

[0015] ;

[0016] Where S is the sum of weight values, n is the number of other obstacle objects within the preset range, and wi is the weight value of the i-th obstacle object. Among them, wi=5 for obstacle objects in motion and wi=1 for obstacle objects in stationary state. The preset weight threshold is set to 10, and the preset range is set to a radius of 10 meters.

[0017] As a further aspect of the present invention: the avoidance intention prediction in the non-emergency obstacle avoidance submodule uses a convolutional neural network model. The input is the obstacle object posture sequence, and the output is the avoidance probability value. If the probability is greater than 0.7, it is determined that the obstacle will be avoided. The total weight factor V is calculated based on a historical database. The database records the number of avoidances for each type of obstacle. V is the sum of the weight factors of the most recent 100 interactions. If V < -5, it is determined that the obstacle of this type tends not to be avoided.

[0018] As a further aspect of the present invention: the speed control module includes an environmental sensing unit and a fuzzy controller. The environmental sensing unit collects parameters such as the density of moving obstacles, average speed, road slope, flatness, and amount of garbage in real time. The fuzzy controller inputs these parameters and outputs a speed adjustment factor, with rules including: speed reduction when the density of moving obstacles is high and the average speed is fast, speed reduction when the road slope is greater than 5 degrees, and speed increase when the amount of garbage is less than 50 kg.

[0019] As a further aspect of the present invention: the positioning verification module synchronizes robot sensor data and camera data through timestamps, verifies positioning accuracy by comparing GPS / IMU fusion position and visual positioning results, verifies speed consistency by comparing encoder speed measurement and optical flow speed measurement, and calculates path deviation as the Hausdorff distance between the actual trajectory and the preset path; the anomaly detection threshold is set to positioning error > 0.5m, or speed deviation > 0.2m / s, or deviation > 1m.

[0020] As a further aspect of the present invention: the calculation of the abnormal weight value adopts a mathematical expression:

[0021] ;

[0022] Where E is the abnormal weight value, ΔP is the positioning error (unit: meters), ΔV is the speed deviation (unit: meters / second), ΔD is the path deviation (unit: meters), α, β, and γ are weight coefficients, and α+β+γ=1, α=0.5, β=0.3, and γ=0.2; the preset interval [Qmin,Qmax] is set to [1,5].

[0023] As a further aspect of the present invention, the system also includes a maintenance point management module, which stores the coordinates and status of maintenance points. When the positioning verification module requests the addition of a maintenance point, it calculates the shortest path based on the Dijkstra algorithm and inserts the maintenance point as a waypoint into the task path.

[0024] The beneficial effects of this invention are reflected in:

[0025] The intelligent navigation system of the medical waste cleaning robot of the present invention accurately distinguishes between moving and stationary obstacles through an obstacle status recognition module. The path planning module intelligently selects arc-shaped detour or dynamic avoidance strategies based on the obstacle status and urgency, and combines a weighted calculation mechanism to realize risk assessment and optimal path decision. The speed control module adaptively adjusts the robot speed in real time according to the number and speed of moving obstacles, road conditions and load, effectively improving motion stability and obstacle avoidance safety. At the same time, the positioning verification module ensures positioning accuracy and path tracking reliability through multi-source data fusion and dual verification mechanism, and intelligently decides the task execution mode based on the abnormal weight value in abnormal situations. Thus, the overall system significantly improves the robot's navigation intelligence, operation efficiency and operational reliability in complex medical environments. Attached Figure Description

[0026] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0027] Figure 1 A structural block diagram of the intelligent navigation system for a medical waste cleaning robot. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] As mentioned in the background section of this application, research has found that existing obstacle avoidance strategies of garbage cleaning robot navigation systems are simplistic, prone to collisions or inefficient, and have certain shortcomings.

[0030] To address the aforementioned shortcomings, this application discloses an intelligent navigation system for a medical waste disposal robot, which can make basic obstacle avoidance decisions based on the dynamic attributes (moving / stationary) of obstacles, and can also classify the urgency level (urgent / non-urgent) based on the speed of moving obstacles, thereby improving operational efficiency and reducing collisions.

[0031] The following will describe in detail, with reference to the accompanying drawings, how the solution of this application solves the above-mentioned technical problems.

[0032] Please see Figure 1 In this embodiment of the invention, the intelligent navigation system of the medical waste cleaning robot includes: an obstacle state recognition module, used to identify the state of obstacles on the robot's forward trajectory, the obstacle state including a moving state and a stationary state; a path planning module, connected to the obstacle state recognition module, used to plan the motion trajectory according to the obstacle state; when the obstacle is stationary, the path planning module generates an arc-shaped detour path; when the obstacle is moving, the path planning module compares the real-time speed of the obstacle with a preset speed threshold, if it is greater than the threshold, it is determined to be an emergency state, and the emergency obstacle avoidance submodule is activated; if it is less than or equal to the threshold, it is determined to be a non-emergency state, and the emergency obstacle avoidance submodule is activated. The system includes a non-emergency obstacle avoidance submodule; a speed control module, which works in conjunction with the path planning module to adjust the robot's speed in real time based on the number and speed of surrounding moving obstacles, road slope and flatness, and the amount of garbage loaded on the robot; and a positioning verification module, which verifies the real-time positioning accuracy, speed consistency, and path deviation using both the robot's built-in sensors and distributed cameras. When an anomaly is detected, an anomaly weight value is calculated, and the robot's action is controlled based on whether the anomaly weight value falls within a preset range [Qmin, Qmax]. If the weight value is less than Qmin, the task continues and repairs are performed; if it is greater than Qmax, the task is immediately abandoned and repairs are performed; if the weight value is within the range, a new path is planned and the robot is added to the repair point. This application achieves full-process intelligent navigation of the robot in a complex hospital environment by constructing a collaborative system framework consisting of four core modules (obstacle state recognition, path planning, speed control, and positioning verification). This system can not only make basic obstacle avoidance decisions based on the dynamic attributes (moving / stationary) of obstacles, but also classify the urgency level (urgent / non-urgent) based on the speed of moving obstacles when they are encountered. Finally, it links the navigation decision with the robot's real-time status (such as speed and positioning accuracy), thereby comprehensively optimizing the efficiency and reliability of path planning and task execution while ensuring safety.

[0033] In this embodiment, the emergency obstacle avoidance submodule is configured to: scan other obstacle objects within a preset range of the current obstacle object, assign a first weight value to obstacle objects in motion and a second weight value to obstacle objects in stationary state, calculate the sum of the weight values, and if the sum is less than a preset weight threshold, output a motion avoidance command and plan the best avoidance path; if the sum is greater than or equal to the preset weight threshold, output a stationary avoidance command and plan the best obstacle avoidance point (the obstacle avoidance point is selected from a safe area based on the environmental map), wherein the first weight value is greater than the second weight value; the non-emergency obstacle avoidance submodule is configured to: control the robot to issue a warning alarm, predict the obstacle object's avoidance intention, if avoidance is predicted, the robot moves forward along the original trajectory, if no avoidance is predicted, output a motion avoidance command; at the same time, identify the obstacle object type (such as personnel, equipment vehicles), update the weight factor based on historical avoidance behavior (if avoidance is achieved this time, the weight factor is +1; if no avoidance is achieved, the weight factor is -1), and calculate the total weight factor V. If V is less than a preset type threshold, an avoidance path is planned in advance. This setup introduces an intelligent decision-making mechanism based on quantitative assessment of environmental risks: In an emergency, by scanning surrounding obstacles and assigning them different weights, the abstract "environmental congestion" is transformed into a quantifiable value (weighted sum), thus objectively deciding whether the robot should "move to avoid" or "remain stationary and wait"; In a non-emergency situation, by combining prediction, learning (recording weight factors), and early warning, human-machine friendly interaction and long-term obstacle avoidance strategies are optimized, demonstrating the intelligence and adaptability of the system's decision-making.

[0034] In this embodiment, the obstacle status recognition module includes a sensor array consisting of a lidar, a visual camera, and an infrared sensor, as well as a data fusion unit. The data fusion unit uses a Kalman filter algorithm to fuse multi-sensor data and outputs the position, velocity, and status information of the obstacle. Moving obstacles are identified using consecutive frame image difference and optical flow (two computer vision techniques used to detect and track moving objects from video captured by a camera), while stationary obstacles are confirmed through point cloud matching and map comparison. Through multi-sensor information complementarity and fusion processing, the system can accurately and stably output key navigation parameters (position, velocity, and status) of obstacles, providing a solid and reliable data foundation for subsequent path planning and effectively overcoming the shortcomings of single sensors being susceptible to interference and providing incomplete perception information.

[0035] In this embodiment, the arc trajectory generation in the path planning module is based on the Bézier curve algorithm. The curve control points are dynamically adjusted according to the obstacle boundary and the robot's turning radius to ensure that the avoidance distance is not less than the safety threshold. The optimal avoidance path planning adopts the A* search algorithm (A* search algorithm, commonly known as the A-star algorithm, is a heuristic search algorithm that finds the path with the lowest passage cost on a graphical plane with multiple nodes) with the goal of minimizing task delay. By using the Bézier curve algorithm to generate an arc path, the smoothness of the robot's motion when bypassing stationary obstacles is ensured, reducing mechanical wear. By using the A* search algorithm to plan the optimal avoidance path and minimizing task delay, the overall work efficiency of the robot when performing obstacle avoidance actions is guaranteed, achieving a balance between path safety, smoothness, and efficiency.

[0036] In this embodiment, the weight value calculation in the emergency obstacle avoidance submodule uses a mathematical expression: Where S is the sum of weight values, n is the number of other obstacles within the preset range, and wi is the weight value of the i-th obstacle, where wi=5 for moving obstacles and wi=1 for stationary obstacles; the preset weight threshold is set to 10, and the preset range is set to a radius of 10 meters. The preset weight threshold is set to 10 through experimental calibration, representing the environmental risk level. When S < 10, the environmental risk is low, and moving avoidance is allowed; when S ≥ 10, the environmental risk is high, and stationary avoidance is required. By assigning clear weight values ​​(5 for moving, 1 for stationary) to obstacles in different states and setting specific judgment thresholds (sum of 10) and scanning ranges (radius of 10 meters), the abstract decision-making logic is transformed into clear and unambiguous algorithm steps, enhancing clarity and feasibility, and limiting the protection range of the emergency obstacle avoidance function.

[0037] In this embodiment, the obstacle avoidance intention prediction in the non-emergency obstacle avoidance submodule uses a convolutional neural network model (an artificial intelligence model adept at processing grid-like data (pixels arranged in a grid) such as images and videos). The input is a sequence of obstacle object postures, and the output is an obstacle avoidance probability value. A probability greater than 0.7 indicates that the obstacle will be avoided. The total weight factor V is calculated based on a historical database, which records the number of times each obstacle type (such as doctors, nurses, and trolleys) has been avoided. V is the sum of the weight factors from the most recent 100 interactions. If V < -5, the obstacle type is determined to be non-avoidable. This setting limits the advanced artificial intelligence model (convolutional neural network) used for obstacle avoidance intention prediction and clarifies the judgment criteria for the prediction result (probability > 0.7). At the same time, it concretizes the implementation method of the "learning" function, that is, by statistically analyzing the sum of the weight factors (V) from the most recent 100 interactions to determine the behavioral tendency of a certain type of obstacle (V < -5 indicates a tendency not to avoid), enabling the system to continuously optimize its obstacle avoidance strategy for specific types of obstacles (such as medical personnel) based on historical experience data, reflecting the system's long-term learning and adaptability.

[0038] In this embodiment, the speed control module includes an environmental perception unit and a fuzzy controller (an intelligent control method specifically designed to handle imprecise and fuzzy information). The environmental perception unit collects parameters in real time, including the density of moving obstacles, average speed, road slope, smoothness, and amount of garbage. The fuzzy controller takes these parameters as input and outputs a speed adjustment factor, with rules including: speed reduction when the density of moving obstacles is high and the average speed is fast; speed reduction when the road slope is greater than 5 degrees; and speed increase when the amount of garbage is less than 50 kg. This setting reveals that speed adjustment is not a simple linear operation, but a complex, nonlinear intelligent decision-making process based on multi-source environmental information (density and speed of moving obstacles, road conditions, and its own load). By employing a fuzzy controller to process these interrelated and potentially uncertain parameters, the system can simulate the decision-making process of a human driver, achieving more flexible, reasonable, and adaptive speed control that adapts to complex environmental changes, thereby improving operational stability and energy efficiency while ensuring safety.

[0039] In this embodiment, the positioning verification module synchronizes robot sensor data and camera data using timestamps. Positioning accuracy is verified by comparing GPS / IMU fusion location with visual positioning results. Speed ​​consistency is verified by comparing encoder velocimetry with optical flow velocimetry. Path deviation is calculated as the Hausdorff distance between the actual trajectory and the preset path. Hausdorff distance is a mathematical method for measuring the distance between two sets of points; the larger the distance, the greater the difference in the overall shape of the two paths. Anomaly detection thresholds are set as follows: positioning error > 0.5m, speed deviation > 0.2m / s, or deviation > 1m. This setting clarifies the specific implementation of "dual verification," namely, comparing data from different sources (robot's built-in sensors and external cameras) through timestamp synchronization. It specifically defines three key verification indicators (positioning accuracy, speed consistency, and path deviation) and their respective calculation or comparison methods, and sets specific numerical thresholds to trigger anomaly judgment. This ensures that the positioning verification process is objective, repeatable, and efficient, providing accurate input for subsequent anomaly weight calculation.

[0040] In this embodiment, the abnormal weight value is calculated using a mathematical expression: Where E is the anomaly weight value, ΔP is the positioning error (unit: meters), ΔV is the speed deviation (unit: meters / second), ΔD is the path deviation (unit: meters), α, β, and γ are weighting coefficients, and α+β+γ=1, α=0.5, β=0.3, and γ=0.2; the preset interval [Qmin, Qmax] is set to [1, 5]. This setting provides a comprehensive mathematical model ( The model integrates anomaly parameters (ΔP, ΔV, ΔD) from different dimensions into a single, quantifiable anomaly weight value (E) through weighted summation. When E < 1, the anomaly is minor; when E > 5, the anomaly is severe; and when 1 ≤ E ≤ 5, the anomaly is moderate. This model (where α + β + γ = 1 ensures the standardization of the calculation) enables the system to comprehensively evaluate the overall health status of the navigation system and compare this comprehensive evaluation value (E) with a preset decision interval ([1, 5]) to make different levels of management decisions (continue the mission, perform mid-journey maintenance, or perform immediate maintenance), thus achieving refined and intelligent fault management.

[0041] In this embodiment, the system also includes a maintenance point management module, which stores the coordinates and status of maintenance points. When the positioning verification module requests the addition of a maintenance point, it calculates the shortest path based on Dijkstra's algorithm (a classic pathfinding algorithm used to find the shortest path between two points in a graph, such as a network path on a map) and inserts the maintenance point as a waypoint into the task path. When the system determines that maintenance is needed, it can dynamically and efficiently integrate the maintenance task into the existing workflow. By storing information through the maintenance point management module and using Dijkstra's algorithm to calculate the optimal path, the system can autonomously decide and navigate to the nearest suitable maintenance point, using it as a waypoint in the task path. This achieves seamless integration of maintenance work and cleaning tasks, minimizing task downtime caused by maintenance and improving the robot's autonomy and attendance efficiency.

[0042] The intelligent navigation system of the medical waste cleaning robot of the present invention accurately distinguishes between moving and stationary obstacles through an obstacle status recognition module. The path planning module intelligently selects arc-shaped detour or dynamic avoidance strategies based on the obstacle status and urgency, and combines a weighted calculation mechanism to realize risk assessment and optimal path decision. The speed control module adaptively adjusts the robot speed in real time according to the number and speed of moving obstacles, road conditions and load, effectively improving motion stability and obstacle avoidance safety. At the same time, the positioning verification module ensures positioning accuracy and path tracking reliability through multi-source data fusion and dual verification mechanism, and intelligently decides the task execution mode based on the abnormal weight value in abnormal situations. Thus, the overall system significantly improves the robot's navigation intelligence, operation efficiency and operational reliability in complex medical environments.

[0043] 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 the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. An intelligent navigation system for a medical waste disposal robot, characterized in that, include: The obstacle state recognition module is used to identify the state of obstacle objects on the robot's forward trajectory. The state of the obstacle objects includes both moving and stationary states. The path planning module, connected to the obstacle state recognition module, is used to plan a motion trajectory based on the state of the obstacle object. When the obstacle object is stationary, the path planning module generates an arc-shaped detour path. When the obstacle object is moving, the path planning module compares the real-time speed of the obstacle object with a preset speed threshold. If the speed is greater than the threshold, it is determined to be an emergency state, and the emergency obstacle avoidance submodule is activated. If the speed is less than or equal to the threshold, it is determined to be a non-emergency state, and the non-emergency obstacle avoidance submodule is activated. The speed control module, in conjunction with the path planning module, is used to adjust the robot speed in real time based on the number and speed of surrounding moving obstacles, road slope and flatness, and the amount of garbage loaded by the robot. The positioning verification module is used to verify the real-time positioning accuracy, speed consistency, and path deviation through the robot's built-in sensors and distributed cameras. When an anomaly is detected, an anomaly weight value is calculated, and the robot's actions are controlled according to whether the anomaly weight value falls within the preset range [Qmin, Qmax]. If it is less than Qmin, the task continues and repair is performed. If it is greater than Qmax, the task is abandoned immediately and repair is performed. If it is within the range, the path is replanned and the robot is added to the repair point.

2. The intelligent navigation system of the medical waste cleaning robot according to claim 1, characterized in that, The emergency obstacle avoidance submodule is configured to: scan other obstacle objects within a preset range of the current obstacle object, assign a first weight value to the obstacle object in motion and a second weight value to the obstacle object in stationary state, calculate the sum of the weight values, and if the sum is less than a preset weight threshold, output a motion avoidance command and plan the best avoidance path; if the sum is greater than or equal to the preset weight threshold, output a stationary avoidance command and plan the best obstacle avoidance point, wherein the first weight value is greater than the second weight value. The non-emergency obstacle avoidance submodule is configured to: control the robot to issue an early warning alarm, predict the obstacle object's avoidance intention, if avoidance is predicted, the robot will move forward along the original trajectory, if no avoidance is predicted, the robot will output a movement avoidance command; at the same time, identify the obstacle object type, update the weight factor based on historical avoidance behavior, and calculate the total weight factor V. If V is less than a preset type threshold, the avoidance path will be planned in advance.

3. The intelligent navigation system of the medical waste cleaning robot according to claim 2, characterized in that, The obstacle status recognition module includes a sensor array consisting of a lidar, a visual camera, and an infrared sensor, as well as a data fusion unit. The data fusion unit uses a Kalman filter algorithm to fuse multi-sensor data and outputs the position, velocity, and status information of the obstacle. Among them, moving obstacles are identified by continuous frame image difference and optical flow method, while stationary obstacles are confirmed by point cloud matching and map comparison.

4. The intelligent navigation system of the medical waste cleaning robot according to claim 3, characterized in that, The arc trajectory generation in the path planning module is based on the Bézier curve algorithm. The curve control points are dynamically adjusted according to the obstacle boundary and the robot's turning radius to ensure that the avoidance distance is not less than the safety threshold. The optimal avoidance path planning adopts the A* search algorithm with the goal of minimizing task delay.

5. The intelligent navigation system of the medical waste cleaning robot according to claim 4, characterized in that, The weight values ​​in the emergency obstacle avoidance submodule are calculated using a mathematical expression: ； Where S is the sum of weight values, n is the number of other obstacle objects within the preset range, and wi is the weight value of the i-th obstacle object, where wi=5 for moving obstacle objects and wi=1 for stationary obstacle objects.

6. The intelligent navigation system of the medical waste cleaning robot according to claim 5, characterized in that, The avoidance intention prediction in the non-emergency obstacle avoidance submodule uses a convolutional neural network model. The input is the obstacle object posture sequence, and the output is the avoidance probability value. If the probability is greater than 0.7, it is determined that the obstacle will be avoided. The total weight factor V is calculated based on the historical database. The database records the number of avoidances for each type of obstacle. V is the sum of the weight factors of the most recent 100 interactions. If V < -5, it is determined that the obstacle of this type is not likely to be avoided.

7. The intelligent navigation system of the medical waste cleaning robot according to claim 6, characterized in that, The speed control module includes an environmental sensing unit and a fuzzy controller. The environmental sensing unit collects parameters such as the density of moving obstacles, average speed, road slope, smoothness, and amount of garbage in real time. The fuzzy controller takes these parameters as input and outputs a speed adjustment factor. The rules include: speed is reduced when the density of moving obstacles is high and the average speed is fast; speed is reduced when the road slope is greater than 5 degrees; and speed is increased when the amount of garbage is less than 50 kg.

8. The intelligent navigation system of the medical waste cleaning robot according to claim 7, characterized in that, The positioning verification module synchronizes robot sensor data and camera data through timestamps, verifies positioning accuracy by comparing GPS / IMU fusion position with visual positioning results, verifies speed consistency by comparing encoder speed measurement and optical flow speed measurement, and calculates path deviation as the Hausdorff distance between the actual trajectory and the preset path; the anomaly detection threshold is set to positioning error > 0.5m, or speed deviation > 0.2m / s, or deviation > 1m.

9. The intelligent navigation system of the medical waste cleaning robot according to claim 8, characterized in that, The abnormal weight value is calculated using a mathematical expression: ； Where E is the anomaly weight value, ΔP is the positioning error, ΔV is the speed deviation, ΔD is the path deviation, α, β, and γ are weight coefficients, and α+β+γ=1.

10. The intelligent navigation system of the medical waste cleaning robot according to claim 9, characterized in that, The system also includes a maintenance point management module, which stores the coordinates and status of maintenance points. When the positioning verification module requests the addition of a maintenance point, it calculates the shortest path based on the Dijkstra algorithm and inserts the maintenance point as a waypoint into the task path.

Images