A small intelligent multi-modal laboratory bottle washing machine, system and cleaning method

Abstract

The application relates to the technical field of laboratory bottle washing machines, and discloses a small intelligent multi-modal laboratory bottle washing machine, a system and a cleaning method. The method is applied to a bottle washing machine equipped with a multi-modal sensing module, multi-modal cleaning data of a bottle to be cleaned is collected, a cleaning interaction area is obtained, the cleaning interaction area is divided into a main sub-area and an auxiliary sub-area according to a cleaning type, an environmental interaction sub-area is determined, a cleaning weight graph is constructed, a cleaning execution sequence is determined through a maximum spanning tree, and finally, cleaning operations of each sub-area are driven. The application can accurately perceive bottle information, scientifically divide cleaning areas, optimize a cleaning sequence, compensate for cleaning errors, realize efficient, accurate and intelligent cleaning, solve the problems of blindness of traditional bottle washing machines, unreasonable area division, fixed sequence and weak error compensation capacity and the like, and improve laboratory bottle washing efficiency and quality, so the application has wide application value.

Description

Technical Field

[0001] This invention relates to the field of laboratory bottle washing machine technology, specifically a small intelligent multimodal laboratory bottle washing machine, system, and cleaning method. Background Technology

[0002] In the daily operation of modern laboratories, cleaning laboratory glassware is an extremely important and tedious task. With the rapid development of technology, the types of bottles used in laboratories are increasing, and their structures are becoming more complex. Special glass instruments and precision plastic containers vary greatly in terms of materials, shapes, and types of dirt. Traditional laboratory bottle washing machines are gradually showing many shortcomings when faced with these diverse cleaning needs.

[0003] Traditional bottle washing machines mostly use a single cleaning mode and lack the ability to perceive multimodal information about the bottle. For example, they cannot accurately obtain information such as the bottle material, the type of surface stains, the degree of stain adhesion, and the specific structure of the bottle. This makes it difficult to perform targeted cleaning according to the characteristics of different bottles during the cleaning process. Either the cleaning is not thorough, affecting the accuracy of subsequent experiments; or it is over-cleaned, wasting water and detergent, and may also damage the bottle.

[0004] Traditional bottle washing machines often lack scientific and rational division of cleaning zones. They tend to treat the entire cleaning space as a single, uniform area, failing to differentiate between critical and non-critical parts of the bottle. As a result, during the cleaning process, some areas that do not require special attention may be over-cleaned, while the truly critical areas that need thorough cleaning may be under-cleaned, thus affecting the overall cleaning effect.

[0005] Traditional bottle washing machines lack scientific basis and intelligent algorithm support when determining the washing sequence. They usually clean in a fixed order, without considering the correlation between different areas or the efficiency of the cleaning process. This fixed cleaning sequence may lead to excessively long cleaning times, or repeated cleaning or missed cleaning during the process, reducing cleaning efficiency and quality.

[0006] Traditional bottle washing machines are relatively weak in handling errors and interference during the washing process. Due to the diversity of bottles and the complexity of the washing environment, some errors are inevitable during the washing process, such as fluctuations in washing pressure and changes in washing fluid flow. Traditional bottle washing machines cannot compensate for and adjust for these errors in a timely manner, thus affecting the stability of the washing effect.

[0007] With the continuous improvement of laboratory automation and intelligence, higher demands are being placed on the intelligence, precision, and efficiency of bottle washing machines. Existing small-scale laboratory bottle washing machines have significant technical bottlenecks in areas such as multimodal sensing, intelligent area division, cleaning sequence optimization, and error compensation, and cannot meet the needs of modern laboratories for efficient and precise cleaning. Summary of the Invention

[0008] The purpose of this invention is to provide a small, intelligent, multimodal laboratory bottle washing machine, system, and cleaning method to solve the problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides the following technical solution: a cleaning method for a small intelligent multimodal laboratory bottle washer, the method comprising:

[0010] Collect multimodal cleaning data of the bottles to be cleaned currently performing the cleaning task, and obtain the cleaning interaction area between the bottles to be cleaned and the small intelligent multimodal laboratory bottle washer based on the multimodal cleaning data;

[0011] Based on the cleaning type of the bottle to be cleaned, the cleaning sub-area covered by the cleaning interaction area is divided into a cleaning-dominant sub-area and a cleaning-auxiliary sub-area, and an environmental interaction sub-area is determined in the cleaning sub-area not covered by the cleaning interaction area. The cleaning-dominant sub-area, the cleaning-auxiliary sub-area, and the environmental interaction sub-area constitute a set of areas to be processed.

[0012] Determine the edges between any two adjacent sub-regions in the set of regions to be processed, and construct a cleaning weight graph of the set of regions to be processed, with the sub-regions in the set of regions to be processed as vertices and the perceptual fusion discrimination value as the edge connecting the vertices.

[0013] Determine the maximum spanning tree of the cleaning weight graph, and based on the maximum spanning tree, determine the cleaning execution order of each sub-region in the set of regions to be processed;

[0014] The cleaning operation is performed by sequentially driving each sub-region in the set of regions to be processed according to the cleaning execution order.

[0015] Preferably, the multimodal cleaning data acquisition of the cleaning interaction area between the bottle to be cleaned and the small intelligent multimodal laboratory bottle washer includes:

[0016] Based on the multimodal cleaning data, the cleaning interaction area between the bottle to be cleaned and the small intelligent multimodal laboratory bottle washer in historical cleaning tasks is statistically analyzed.

[0017] Based on the state of the key parts of the bottle to be cleaned in each cleaning interaction area, the statistically obtained cleaning interaction areas are classified to form one or more sets of cleaning interaction areas.

[0018] A target set of cleaning interaction areas that matches the current cleaning type of the bottle to be cleaned is determined from the set of cleaning interaction areas, and the area covered by the target set of cleaning interaction areas is taken as the cleaning interaction area between the bottle to be cleaned and the small intelligent multimodal laboratory bottle washing machine.

[0019] Preferably, the state of the key parts of the bottle to be cleaned is categorized based on the statistically obtained cleaning interaction areas, including:

[0020] Obtain the predefined state standard interval in the small intelligent multimodal laboratory bottle washing machine;

[0021] For any cleaning interaction area obtained from statistics, identify the state standard interval to which the state of the key part of the bottle to be cleaned in the cleaning interaction area belongs, and divide the cleaning interaction area into the set of cleaning interaction areas corresponding to the identified state standard interval.

[0022] Preferably, the cleaning type of the bottle to be cleaned, which divides the cleaning sub-area covered by the cleaning interaction area into a primary cleaning sub-area and a secondary cleaning sub-area, includes:

[0023] Determine the primary and secondary cleaning areas represented by the cleaning type of the bottle to be cleaned;

[0024] A first range of motion for the primary cleaning part is determined in the cleaning interaction area, and a second range of motion for the secondary cleaning part is determined in the cleaning interaction area.

[0025] The cleaning sub-area covered by the first movement range within the cleaning range of the small intelligent multimodal laboratory bottle washer is defined as the primary cleaning sub-area, and the cleaning sub-area covered by the second movement range within the cleaning range of the small intelligent multimodal laboratory bottle washer is defined as the auxiliary cleaning sub-area.

[0026] Preferably, the environmental interaction sub-regions identified in the cleaning sub-regions not covered by the cleaning interaction area include:

[0027] A first extension range is set for the cleaning-dominant sub-region in the cleaning interaction area, and a second extension range is set for the cleaning-assistant sub-region in the cleaning interaction area;

[0028] For cleaning sub-areas not covered by the cleaning interaction area, if they are adjacent to the cleaning dominant sub-area, the cleaning sub-area of ​​the first extended range is selected as the environmental interaction sub-area; if they are adjacent to the cleaning auxiliary sub-area, the cleaning sub-area of ​​the second extended range is selected as the environmental interaction sub-area.

[0029] Preferably, the perceptual fusion discrimination value between any two adjacent sub-regions in the set of regions to be processed includes:

[0030] For the first and second adjacent sub-regions in the set of regions to be processed, the respective region categories of the first and second sub-regions are identified. The region category includes one of the following: cleaning-dominant sub-region, cleaning-assisted sub-region, and environmental interaction sub-region.

[0031] Based on the identified region category, determine the first perceptual fusion parameter of the first sub-region and the second perceptual fusion parameter of the second sub-region;

[0032] Calculate the difference between the first perceptual fusion parameter and the second perceptual fusion parameter, and determine the perceptual fusion distinction value between the first sub-region and the second sub-region based on the difference value.

[0033] Preferably, the difference value used to determine the perceptual fusion distinction between the first sub-region and the second sub-region includes:

[0034] Identify the perceptual fusion parameters of each sub-region in the set of regions to be processed, and determine the average value of the identified perceptual fusion parameters;

[0035] Multiple difference intervals are determined based on the average value, and the target difference interval in which the calculated difference value between the first perception fusion parameter and the second perception fusion parameter is located is determined.

[0036] The preset discrimination value corresponding to the target difference interval is determined as the perceptual fusion discrimination value between the first sub-region and the second sub-region.

[0037] Preferably, the cleaning execution sequence sequentially drives each sub-region in the set of regions to be processed to perform the cleaning operation, including:

[0038] The execution parameters of each execution node in the cleaning execution sequence are identified, and an error parameter sequence is constructed based on the identified execution parameters. The order of each parameter in the error parameter sequence is consistent with the cleaning execution sequence.

[0039] The error parameter sequence is input into the error compensation network to generate cleaning action outputs for each of the execution nodes through the error compensation network;

[0040] The generated cleaning actions output drive corresponding to each sub-region to perform the cleaning operation.

[0041] Preferably, the present invention also includes a cleaning system for a small intelligent multimodal laboratory bottle washer, the system being applied in the small intelligent multimodal laboratory bottle washer, the small intelligent multimodal laboratory bottle washer being pre-configured with a multimodal sensing module, the system comprising:

[0042] The data acquisition unit is used to acquire multimodal cleaning data of the bottle to be cleaned currently performing the cleaning task, and to obtain the cleaning interaction area between the bottle to be cleaned and the small intelligent multimodal laboratory bottle washer based on the multimodal cleaning data.

[0043] The area division unit is used to divide the cleaning sub-area covered by the cleaning interaction area into a cleaning-dominant sub-area and a cleaning-auxiliary sub-area based on the cleaning type of the bottle to be cleaned, and to determine an environmental interaction sub-area in the cleaning sub-area not covered by the cleaning interaction area. The cleaning-dominant sub-area, the cleaning-auxiliary sub-area, and the environmental interaction sub-area constitute a set of areas to be processed.

[0044] The weight graph construction unit is used to determine the perceptual fusion discrimination value between any two adjacent sub-regions in the set of regions to be processed, and to construct a cleaning weight graph of the set of regions to be processed using the sub-regions in the set of regions to be processed as vertices and the perceptual fusion discrimination value as the edge connecting the vertices.

[0045] The sequence determination unit is used to determine the maximum spanning tree of the cleaning weight graph, and to determine the cleaning execution order of each sub-region in the set of regions to be processed based on the maximum spanning tree.

[0046] An operation execution unit is used to sequentially drive each sub-region in the set of regions to be processed according to the cleaning execution order to perform cleaning operations;

[0047] The operation execution unit is specifically used to: identify the execution parameters of each execution node in the cleaning execution sequence, and construct an error parameter sequence based on the identified execution parameters, wherein the order of each parameter in the error parameter sequence is consistent with the cleaning execution sequence; input the error parameter sequence into an error compensation network to generate cleaning action outputs for each execution node through the error compensation network; and drive the corresponding sub-regions according to the generated cleaning action outputs to perform cleaning operations.

[0048] Preferably, the present invention also includes a small intelligent multimodal laboratory bottle washing machine, the bottle washing machine further including a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the computer program to implement the cleaning method of the small intelligent multimodal laboratory bottle washing machine described above.

[0049] Compared with the prior art, the beneficial effects of the present invention are:

[0050] By collecting multimodal cleaning data from the bottles to be cleaned through a multimodal sensing module, information such as bottle material, stain type, adhesion degree, and structure can be accurately obtained, providing comprehensive and accurate data support for subsequent cleaning work. This changes the problem of blind cleaning caused by the lack of multimodal sensing capabilities in traditional bottle washing machines, greatly improving the adaptability to cleaning different types of bottles. It can ensure that all kinds of bottles can be cleaned just right, avoiding incomplete cleaning that affects experiments, and preventing over-cleaning that wastes resources and damages the bottles.

[0051] In terms of cleaning zone division, the sub-areas covered by the cleaning interaction area are scientifically divided into primary cleaning sub-areas and auxiliary cleaning sub-areas based on the cleaning type, and environmental interaction sub-areas are also identified, forming a set of areas to be treated. This division method fully considers the critical and non-critical parts of the bottle, enabling the rational allocation of cleaning resources. For the primary cleaning sub-area, where the primary cleaning areas are located, more cleaning resources can be concentrated for focused cleaning to ensure the cleanliness of critical parts; while the auxiliary sub-areas use relatively gentle cleaning methods to ensure cleaning effectiveness while reducing unnecessary damage to the bottle. Compared with the traditional bottle washing machine's uniform area cleaning method, this precise zone division significantly improves the targeting and effectiveness of cleaning.

[0052] When determining the cleaning execution order, the sequence is planned by constructing a cleaning weight graph and determining its maximum spanning tree. This fully considers the perceptual fusion discrimination values ​​between each sub-region, making the cleaning order more scientific and reasonable. This order determination method based on intelligent algorithms can avoid problems such as repeated cleaning and missed cleaning that occur in traditional fixed-order cleaning, greatly improving cleaning efficiency and shortening cleaning time. At the same time, a reasonable cleaning order can also reduce energy consumption during the cleaning process, achieving energy-saving effects.

[0053] Regarding error compensation, an error parameter sequence is constructed by identifying execution parameters and input into an error compensation network to generate cleaning action output, enabling timely compensation and adjustment for errors occurring during the cleaning process. This intelligent error compensation mechanism allows the bottle washer to maintain stable cleaning results when faced with diverse bottle types and complex cleaning environments, improving the reliability and practicality of the bottle washer.

[0054] This invention integrates data acquisition, area division, weighted graph construction, sequence determination, and operation execution into a single system design, achieving fully automated and intelligent control of the cleaning process. It not only reduces the tediousness and errors of manual operation but also improves laboratory efficiency, providing strong support for the automation and intelligent development of laboratories. Furthermore, the compact structure of this bottle washing machine makes it suitable for small laboratories and offers broad application prospects. Attached Figure Description

[0055] Figure 1 This is a schematic diagram illustrating the working principle of the cleaning method of the small intelligent multimodal laboratory bottle washer described in this invention.

[0056] Figure 2 Design diagram for determining the method of cleaning interaction area;

[0057] Figure 3 Design diagram for the method of dividing the cleaning sub-region;

[0058] Figure 4 Design diagram for a method to determine the distinguishing value of perceptual fusion;

[0059] Figure 5 This is a design drawing of the cleaning operation execution system. Detailed Implementation

[0060] 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.

[0061] Please see Figures 1-5 This invention relates to a cleaning method for a small, intelligent, multimodal laboratory bottle washer, which is applied to a bottle washer equipped with a multimodal sensing module. The specific implementation steps are as follows:

[0062] The system collects multimodal cleaning data of the bottles currently being cleaned. This multimodal cleaning data may include, but is not limited to, bottle appearance images acquired by image sensors, pressure data of the bottle's placement position acquired by pressure sensors, and bottle temperature data acquired by infrared sensors. Based on this multimodal cleaning data, the cleaning interaction area between the bottle and the bottle washer is determined.

[0063] Based on the cleaning type of the bottle to be cleaned, the cleaning sub-area covered by the cleaning interaction area is divided into a primary cleaning sub-area and a secondary cleaning sub-area. Simultaneously, an environmental interaction sub-area is determined within the cleaning sub-areas not covered by the primary cleaning interaction area. The primary cleaning sub-area, secondary cleaning sub-area, and environmental interaction sub-area together constitute the set of areas to be processed.

[0064] Determine the perceptual fusion discrimination value between any two adjacent sub-regions in the set of regions to be processed, and construct a cleaning weight graph of the set of regions to be processed with sub-regions as vertices and perceptual fusion discrimination values ​​as edges connecting the vertices.

[0065] Determine the maximum spanning tree of the cleaning weight graph, and determine the cleaning execution order of each sub-region in the set of regions to be processed based on the maximum spanning tree.

[0066] The cleaning operation is performed sequentially on each sub-region in the set of regions to be processed, following the cleaning execution order.

[0067] Example 1:

[0068] This embodiment describes in detail the specific process of obtaining the interaction area between the bottle to be cleaned and the small intelligent multimodal laboratory bottle washer based on multimodal cleaning data.

[0069] The system collects multimodal cleaning data of the bottles currently undergoing cleaning through a multimodal sensing module. This data includes image data such as the bottle's outline, color, and stain distribution acquired by an image sensor; pressure distribution data of the bottle's position in the bottle washer acquired by a pressure sensor; and surface temperature data of the bottle acquired by an infrared sensor. After data collection, the system accesses historical cleaning task data stored in its memory to analyze the cleaning interaction areas between the bottles and the bottle washer in previous cleaning tasks. This historical cleaning task data is generated by recording and storing the areas where the bottles interacted with the bottle washer after each previous cleaning task. Each record includes information such as the bottle's position, orientation, and the range of interaction with various components of the bottle washer during that cleaning session.

[0070] After statistically analyzing the historical cleaning interaction areas, the system needs to categorize these areas based on the state of key parts of the bottles to be cleaned within each area. It obtains predefined state standard intervals from the bottle washer, which are divided by technicians based on different state characteristics of key bottle parts. For example, based on the degree of stain adhesion on key bottle parts, it can be divided into lightly soiled, moderately soiled, and heavily soiled state standard intervals; based on the bottle's placement angle, it can be divided into different intervals based on the deviation angle from the standard station of the bottle washer. For any statistically analyzed cleaning interaction area, the system identifies the state of the key parts of the bottle to be cleaned. Taking stain adhesion as an example, the system uses image data collected by image sensors and image processing algorithms to analyze the pixel distribution and color depth of stains on key parts, thereby determining the state standard interval to which the stain adhesion level of that key part belongs. Similarly, for placement angle, the system can comprehensively determine the bottle's tilt angle using pressure sensor data and image data to determine its corresponding interval. After identification is completed, the cleaning interaction area is divided into the cleaning interaction area set corresponding to the corresponding state standard interval, thus forming one or more sets of cleaning interaction areas classified according to different state standard intervals.

[0071] The system needs to identify a target set of cleaning interaction areas from these sets that matches the current cleaning type of the bottle to be cleaned. The current cleaning type can be determined by the user input through the interface or automatically by the system based on the bottle type and preliminary stain detection results. Different cleaning types correspond to different cleaning needs. For example, "normal cleaning" is mainly for light stains, "stubborn stain cleaning" is for heavily adhered stains, and "precision instrument cleaning" requires a higher level of gentleness and cleanliness. The system will match the historical cleaning types corresponding to each set of cleaning interaction areas with the current cleaning type. The matching process can use methods such as keyword matching and feature similarity calculation. Taking "stubborn stain cleaning" as an example, the system will search all sets of cleaning interaction areas where the historical cleaning type is marked as "stubborn stain cleaning" or whose features have a high similarity to the current stubborn stain features, and identify these as the target set of cleaning interaction areas.

[0072] The area covered by the target cleaning interaction area set is defined as the cleaning interaction area between the bottle to be cleaned and the small intelligent multimodal laboratory bottle washer. Here, the covered area refers to the union of all historical cleaning interaction areas in the target cleaning interaction area set within the bottle washer space. This area reflects the main range of effective cleaning interaction between the bottle and the bottle washer in historical cleaning tasks of the same or similar cleaning type as the current one. The cleaning interaction area determined in this way, based on historical data and multimodal perception information, can more accurately locate the areas requiring focus in the current cleaning task, providing an accurate basis for the effective range of subsequent cleaning operations. Throughout the process, the system dynamically acquires the cleaning interaction area through multimodal data collection, statistical analysis of historical data, identification and classification of state intervals, and matching of cleaning types. This ensures that the area can adapt to the needs of different bottles and different cleaning types, laying the foundation for subsequent area division and determination of the cleaning execution sequence.

[0073] Example 2:

[0074] This embodiment details the process of dividing the cleaning sub-area covered by the cleaning interaction area into a primary cleaning sub-area and a secondary cleaning sub-area based on the cleaning type of the bottle to be cleaned.

[0075] The system needs to determine the primary and secondary cleaning areas of the bottle to be cleaned, as indicated by the cleaning type. The cleaning type can be determined manually by the user through the bottle washer's interface, such as by setting options like "normal cleaning," "stubborn stain cleaning," "inner wall cleaning," and "bottle neck cleaning," allowing the user to select based on the bottle's condition. Alternatively, the system can automatically identify the type, for example, by using an image sensor to capture the bottle's appearance features and combining this with a pre-set bottle type database to determine if the bottle is a conical flask, beaker, or volumetric flask, and then automatically matching the appropriate cleaning type based on the bottle type and the initial stain detection.

[0076] Different cleaning types correspond to different primary and secondary cleaning areas. For example, when the cleaning type is "conical flask inner wall cleaning," the primary cleaning area is the inner wall of the conical flask, as this is the area where the main stains need to be removed. The secondary cleaning area is the flask neck, which, although not the primary area for stain adhesion, still requires some cleaning to ensure overall cleanliness. When the cleaning type is "beaker outer wall cleaning," the primary cleaning area is the outer wall of the beaker, and the secondary cleaning area is the bottom of the beaker. When the cleaning type is "volume flask neck cleaning," the primary cleaning area is the neck of the volumetric flask, and the secondary cleaning area is the body of the flask. These correspondences between primary and secondary cleaning areas are pre-set in the system and stored in the bottle washer's memory, forming a mapping table between cleaning types and cleaning areas.

[0077] After determining the primary and secondary cleaning areas, it is necessary to define the first range of motion for the primary cleaning area and the second range of motion for the secondary cleaning area within the cleaning interaction zone. The cleaning interaction zone is obtained from previous steps based on multimodal cleaning data; this zone defines the main spatial range where the bottle interacts with the bottle washer. Determining the first and second ranges of motion requires the comprehensive use of multimodal sensing data. Taking an image sensor as an example, the system acquires high-resolution images of the bottle to be cleaned using an image sensor, processes the images using computer vision algorithms, and identifies information such as the bottle's outline, shape, and size, thereby determining the specific location and range of the primary and secondary cleaning areas in space.

[0078] Taking the cleaning of the inner wall of a conical flask as an example, the contour of the inner wall of the flask is determined through image recognition. Combined with the mechanical structural parameters of the bottle washer, such as the movement range of the nozzle, the length and rotation range of the brush, the area that the nozzle or brush can cover when cleaning the inner wall of the flask is determined. This area is the first movement range of the primary cleaning part. For the auxiliary cleaning part, the bottle mouth, the position and size of the bottle mouth are also obtained based on image recognition. Combined with the movement capabilities of the bottle washer components, the movement range of the bottle washer components when cleaning the bottle mouth is determined, i.e., the second movement range. In determining the movement range, the actual placement position of the bottle in the bottle washer also needs to be considered. This can be determined using data from pressure sensors. For example, pressure sensors are distributed on the base of the bottle washer. The distribution of pressure values ​​from each sensor can determine the center of gravity position and placement posture of the bottle, thereby correcting the movement range and ensuring its accuracy.

[0079] After determining the first and second motion ranges, these motion ranges need to be mapped to the cleaning range of the bottle washer, thereby determining the primary and secondary cleaning sub-areas. The cleaning range of the bottle washer refers to the spatial range that each actuator of the bottle washer, such as nozzles and brushes, can reach and perform cleaning operations during normal operation. This range is determined by the mechanical design of the bottle washer, such as the up, down, left, and right movement stroke of the nozzles and the rotation radius of the brushes.

[0080] The cleaning sub-area covered by the first movement range within the bottle washing machine's cleaning range is defined as the primary cleaning sub-area. Here, the cleaning sub-area is defined as the spatial division of the bottle washing machine's cleaning range into multiple small regional units, each with a specific coordinate range. The system determines whether each point within the first movement range falls within the bottle washing machine's cleaning range through coordinate transformation and spatial calculation, and marks the cleaning sub-area corresponding to the point falling within the cleaning range as the primary cleaning sub-area. Similarly, the cleaning sub-area covered by the second movement range within the bottle washing machine's cleaning range is defined as the auxiliary cleaning sub-area.

[0081] For example, the cleaning range of a bottle washer can be represented as a cube in three-dimensional space, which can be divided into multiple smaller cubes with specific side lengths as cleaning sub-regions. When the spatial region corresponding to the first movement range intersects with the cube of the bottle washer's cleaning range, the smaller cubes contained in the intersection are the dominant cleaning sub-regions; similarly, the smaller cubes contained in the intersection of the second movement range and the bottle washer's cleaning range are the auxiliary cleaning sub-regions. In this way, the movement range of the bottle's cleaning parts is transformed into a division of actually operable cleaning sub-regions for the bottle washer, providing clear target areas for subsequent cleaning operations.

[0082] Throughout the process, the system accurately divides the primary and secondary cleaning sub-areas by determining the cleaning type, identifying the cleaning areas, calculating the range of motion, and mapping it to the cleaning range of the bottle washer. This division method can dynamically adjust the primary and secondary cleaning areas according to different cleaning types and bottle characteristics, improving cleaning efficiency and effectiveness while rationally allocating cleaning resources and avoiding unnecessary energy and time consumption.

[0083] Example 3:

[0084] This embodiment describes in detail the specific process of determining the environmental interaction sub-region in the cleaning sub-region that is not covered by the cleaning interaction region.

[0085] The system needs to set extension ranges for the primary cleaning sub-area and the auxiliary cleaning sub-area within the cleaning interaction area, namely, a first extension range and a second extension range. The setting of the extension ranges is based on cleaning process requirements and actual cleaning experience, aiming to include potentially affected areas adjacent to the core cleaning area within the treatment range to ensure that environmental factors do not interfere with the cleaning effect during the cleaning process.

[0086] For cleaning the primary sub-area, the setting of the first extension range needs to consider the potential impact of this area on the surrounding environment during the cleaning process, as well as the potential effect of the surrounding environment on its cleaning effect. For example, when the primary sub-area is the inner wall of the bottle, its first extension range can be set as extending a certain spatial distance outward from this sub-area. This distance is usually determined based on the spray range of the bottle washer nozzle or the oscillation amplitude of the brush, ensuring that when cleaning the inner wall, the cleaning fluid sprayed by the nozzle or the movement of the brush will not be incompletely cleaned due to boundary restrictions, and that dirt in the surrounding area will not flow back to the inner wall due to factors such as water flow. Specifically, the system can obtain the maximum spray distance of the nozzle or the maximum rotation radius of the brush by reading the bottle washer's mechanical structure parameter database, using this as a reference for the first extension range, and then adjust it in conjunction with the optimal extension distance for cleaning similar bottles in historical cleaning data.

[0087] For the cleaning auxiliary sub-area, the setting principle for the second extension range is similar to that of the first extension range, but it will differ depending on the function and importance of the auxiliary sub-area. For example, when the cleaning auxiliary sub-area is the bottle mouth, its second extension range can be set to extend a small distance from the bottle mouth towards the bottle body and external space, because cleaning the bottle mouth is less important than the inner wall, but it is still necessary to ensure the cleanliness of its surrounding area to prevent stains from spreading to other areas during the cleaning process. The specific value of the second extension range is also stored in the system's parameter library and can be dynamically adjusted according to different bottle types and cleaning types.

[0088] After setting the extended range, the system needs to traverse the cleaning sub-areas not covered by the cleaning interaction area and determine the adjacency relationship between these sub-areas and the main or auxiliary cleaning sub-areas. Cleaning sub-areas not covered by the cleaning interaction area refer to the sub-areas outside the cleaning range of the bottle washer, excluding the portion covered by the cleaning interaction area. These sub-areas may be located in the space surrounding the bottle or in other parts of the bottle washer.

[0089] The adjacency determination is based on the spatial relationship between sub-regions. In three-dimensional space, each cleaning sub-region has its specific coordinate range. The system determines whether two sub-regions are adjacent by calculating the spatial distance and positional relationship between them. For example, if two sub-regions share a common face, edge, or vertex in three-dimensional coordinates, they are considered adjacent; or, if the distance between two sub-regions is less than a preset adjacency threshold, they are also considered adjacent. The preset adjacency threshold can be set according to the precision of the bottle washing machine and the cleaning requirements, such as setting it to the side length of a sub-region, to ensure the accuracy of the adjacency determination.

[0090] For each cleaning sub-area not covered by the cleaning interaction area, the system first determines whether it is adjacent to the dominant cleaning sub-area. If an adjacency exists, the cleaning sub-area within the first extended range is selected as the environmental interaction sub-area. For example, suppose the dominant cleaning sub-area is a sub-area corresponding to the inner wall of the bottle, and its first extended range extends a certain distance outward from the bottle. Among the sub-areas not covered by the cleaning interaction area, there is a sub-area located on the outer side of the bottle, spatially adjacent to the inner wall sub-area and within the first extended range; then this sub-area will be identified as the environmental interaction sub-area.

[0091] If a cleaning sub-area not covered by the cleaning interaction area is adjacent to a cleaning auxiliary sub-area, then the cleaning sub-area within the second extended range is selected as the environmental interaction sub-area. For example, the cleaning auxiliary sub-area is the sub-area corresponding to the bottle opening, and its second extended range extends towards the bottle body. When an uncovered sub-area is located on the bottle body and adjacent to the bottle opening sub-area, and is also within the second extended range, that sub-area will be identified as the environmental interaction sub-area.

[0092] In determining the environmental interaction sub-region, the system needs to call the bottle position and posture data obtained by the multimodal perception module in real time to ensure the accuracy of the expansion range and adjacency relationship judgment. For example, the image sensor confirms the specific placement of the bottle in the bottle washing machine to avoid deviations in the expansion range calculation due to bottle tilt; the pressure sensor data determines the contact between the bottle and the bottle washing machine base to correct the spatial coordinates of the sub-region.

[0093] In addition, the system will also refer to the methods used to determine the environmental interaction sub-regions during the cleaning of similar bottles in historical cleaning data to optimize the current judgment process. For example, in previous conical flask cleaning tasks, when the dominant cleaning sub-region is the inner wall, a specific sub-region outside its first extended range is often identified as the environmental interaction sub-region, and the cleaning of this region has a significant impact on the overall effect. In this case, the system will give higher weight to this region in the current judgment and prioritize its inclusion in the environmental interaction sub-region.

[0094] The purpose of identifying environmental interaction sub-regions is to simultaneously treat the environmental areas adjacent to the core cleaning area during the cleaning process, preventing stains, residual cleaning fluid, or other impurities in these areas from negatively impacting the cleaning effect of the core area. For example, when cleaning the inner wall of a bottle, if the environmental interaction sub-region adjacent to the outer side of the bottle is not treated, stains on the outer side may spread to the inner wall under the action of the cleaning fluid, resulting in incomplete cleaning. Treating the environmental interaction sub-regions can effectively prevent this from happening, ensuring the overall cleaning effect.

[0095] Throughout the determination process, the system achieves dynamic identification and determination of environmental interaction sub-regions through reasonable setting of the extended range, accurate judgment of adjacency relationships, and comprehensive application of multimodal data. This ensures that the set of regions to be processed can fully cover all regions that need attention during the cleaning process, providing a complete regional foundation for the subsequent construction of the cleaning weight map and determination of the cleaning execution order, making the bottle washing machine's cleaning operation more comprehensive and efficient.

[0096] Example 4:

[0097] This implementation details the process of determining the perceptual fusion distinction value between any two adjacent sub-regions in the set of regions to be processed. For the first and second adjacent sub-regions in the set of regions to be processed, the system needs to identify their respective region categories. These region categories include cleaning-dominant sub-regions, cleaning-auxiliary sub-regions, and environmental interaction sub-regions. Taking the cleaning of conical flasks by a small intelligent multimodal laboratory bottle washer as an example, assume that the set of regions to be processed contains a sub-region A located on the inner wall of the flask and a sub-region B located in the adjacent space on the outer side of the flask. Through preliminary segmentation, the system determines that sub-region A, because it falls within the movement range covered by the dominant cleaning part (inner wall) in the cleaning interaction region, is classified as a cleaning-dominant sub-region; sub-region B, because it is located in an area not covered by the cleaning interaction region and is adjacent to the cleaning-dominant sub-region, is classified as an environmental interaction sub-region.

[0098] When identifying region categories, the system retrieves previously stored region segmentation results. Each sub-region is tagged and stored in the database upon segmentation. When determining the category of adjacent sub-regions, the corresponding tags are retrieved directly from the database. For example, in a beaker cleaning scenario, if the set of regions to be processed contains a sub-region C located on the outer wall of the beaker and a sub-region D located at the bottom of the beaker, based on the previously segmented logic, sub-region C belongs to the cleaning-dominant sub-region (the outer wall is the primary cleaning area), and sub-region D belongs to the cleaning-auxiliary sub-region (the bottom is the secondary cleaning area). In this case, the adjacent sub-regions C and D are classified as the cleaning-dominant sub-region and the cleaning-auxiliary sub-region, respectively.

[0099] After identifying the region categories, the system needs to determine the first perception fusion parameter for the first sub-region and the second perception fusion parameter for the second sub-region. The determination of the perception fusion parameters is related to the characteristics of the region category. For a cleaning-dominant sub-region, its perception fusion parameters may include stain detection data, cleaning fluid flow rate data, and nozzle pressure data for that region. For example, in cleaning-dominant sub-region A, which cleans the inner wall of a conical flask, the system acquires stain pixel distribution data for that region using an image sensor, acquires the current cleaning fluid flow rate data sprayed into that region using a flow sensor, and acquires the pressure data of the nozzle spraying into that region using a pressure sensor. These data are then integrated as the first perception fusion parameter. For an environmental interaction sub-region B, its perception fusion parameters may include ambient temperature data, residual cleaning fluid concentration data, and air velocity data for that region. The system acquires the ambient temperature of sub-region B using a temperature sensor, detects the residual cleaning fluid concentration in that region using a concentration sensor, and acquires the air velocity using a wind speed sensor. These data are then integrated as the second perception fusion parameter.

[0100] In the scenario of cleaning a beaker, the sensing fusion parameters of the main cleaning sub-region C may include the spectral data of the stains on the outer wall, the brush rotation speed data, and the cleaning fluid temperature data. The system analyzes the composition and concentration of the stains on the outer wall through a spectral sensor, obtains the brush rotation speed through a motor encoder, and obtains the cleaning fluid temperature through a temperature sensor. The sensing fusion parameters of the auxiliary cleaning sub-region D may include the pressure distribution data at the bottom of the beaker and the cleaning fluid residence time data. The system obtains the pressure distribution at the bottom of the beaker through a pressure sensor array and records the residence time of the cleaning fluid in this area through a timer. These data constitute the first sensing fusion parameter and the second sensing fusion parameter, respectively.

[0101] The system calculates the difference between the first and second perception fusion parameters. This difference can be calculated using a combination of data processing methods. For example, for numerical data such as cleaning fluid flow rate (liters / minute) and nozzle pressure (Pascals), the absolute difference or Euclidean distance can be directly calculated. For non-numerical data, such as stain pixel distribution, the similarity distance between vectors can be calculated after feature vector extraction. Assuming the first perception fusion parameter is a set of data including stain concentration (0.8 mg / cm²), cleaning fluid flow rate (5 L / min), and nozzle pressure (200 Pa), and the second perception fusion parameter is stain concentration (0.2 mg / cm²), ambient temperature (25℃), and residual cleaning fluid concentration (0.1%), the system will standardize the data of different types, converting them into values ​​with uniform dimensions, then calculate the difference between each dimension, and finally obtain the comprehensive difference value through a weighted summation.

[0102] After calculating the difference values, the system needs to identify the sensor fusion parameters of each sub-region in the set of areas to be processed and determine the average value of these parameters. For example, if the set of areas to be processed contains 10 sub-regions, and each sub-region has 3 sensor fusion parameters, the system will calculate the average value of each parameter, such as the average value of stain concentration, the average value of cleaning fluid flow rate, and the average value of nozzle pressure. These average values ​​are calculated based on the sensor fusion parameter data of all sub-regions, obtained by summing them and dividing by the number of sub-regions.

[0103] Based on the average value, the system determines multiple difference intervals. The division of these intervals is pre-defined; for example, the difference values ​​are divided into five intervals: "extremely low," "low," "medium," "high," and "extremely high," each corresponding to a different numerical range. Assuming that after calculating the average value of the perceptual fusion parameters for all sub-regions, the difference values ​​are determined as follows: 0-10 is the "extremely low" interval, 11-20 is the "low" interval, 21-30 is the "medium" interval, 31-40 is the "high" interval, and above 41 is the "extremely high" interval. The system compares the calculated difference values ​​of the first and second sub-regions with these intervals to determine their target difference interval.

[0104] For example, in the conical flask cleaning scenario, the calculated difference value of the perception fusion parameters between sub-region A (the cleaning-dominant sub-region) and sub-region B (the environmental interaction sub-region) is 25. This value is in the "medium" range of 21-30, so the target difference range is the "medium" range. In another scenario, if the difference value between two adjacent sub-regions is 35, then it is in the "high" range of 31-40.

[0105] The system determines the preset discrimination value corresponding to the target difference interval as the perceptual fusion discrimination value between the first and second sub-regions. The preset discrimination values ​​are pre-stored in the system and correspond one-to-one with each difference interval. For example, the "extremely low" interval corresponds to a discrimination value of 5, the "low" interval to 10, the "medium" interval to 15, the "high" interval to 20, and the "extremely high" interval to 25. When the target difference interval is determined to be the "medium" interval, the perceptual fusion discrimination value is 15; if the target difference interval is the "high" interval, the discrimination value is 20.

[0106] In practical applications, the perceptual fusion discrimination value is used to construct the cleaning weight map. This value reflects the degree of difference between adjacent sub-regions; the greater the difference, the higher the discrimination value, and the greater the weight of the corresponding edge in the weight map. For example, the perceptual fusion discrimination value between the cleaning-dominant sub-region and the environmental interaction sub-region may be higher than the discrimination value between the cleaning-dominant sub-region and the cleaning-auxiliary sub-region, because the former has more significant differences in regional functions and characteristics. In this way, the system can quantify the degree of correlation between adjacent sub-regions based on region categories and perceptual data, providing data support for subsequent construction of the cleaning weight map and determination of the cleaning execution order. This ensures that cleaning operations can be arranged in a reasonable order according to the differences in regional characteristics, improving cleaning efficiency and effectiveness.

[0107] Example 5:

[0108] This implementation details the specific process of sequentially driving each sub-region in the set of regions to be processed to perform cleaning operations according to the cleaning execution order. Taking the cleaning of a conical flask as an example, it is assumed that the set of regions to be processed has been determined through previous steps to include the dominant cleaning sub-region located on the inner wall of the flask, the auxiliary cleaning sub-region at the flask opening, and the adjacent environmental interaction sub-region on the outer side of the flask. By constructing a cleaning weight graph and a maximum spanning tree, the determined cleaning execution order is: first clean the dominant cleaning sub-region on the inner wall, then clean the auxiliary cleaning sub-region at the flask opening, and finally clean the environmental interaction sub-region on the outer side of the flask.

[0109] The system needs to identify the execution parameters of each execution node in the cleaning execution sequence. Each execution node corresponds to a sub-region in the set of areas to be processed. The execution parameters are the key control parameters that drive the cleaning operation in that sub-region, and their specific content is determined according to the category and cleaning requirements of the sub-region. For the execution node of the dominant sub-region for cleaning the inner wall, the execution parameters may include the concentration of the cleaning solution, the spray pressure of the nozzle, the spray angle, the cleaning time, and the speed of the brush. The system obtains these parameters by accessing the preset parameter table corresponding to the sub-region. The preset parameter table stores the standard execution parameters for different types of sub-regions under different cleaning types. For example, for the dominant sub-region for cleaning the inner wall of the "stubborn stain cleaning" type, the preset parameter table specifies that the cleaning solution concentration is 5%, the nozzle spray pressure is 0.3MPa, the spray angle is 45 degrees, the cleaning time is 3 minutes, and the brush speed is 150 rpm.

[0110] For the cleaning assistance sub-area execution node at the bottle opening, the execution parameters may include the cleaning fluid flow rate, the nozzle moving speed, and the cleaning time. The preset parameter table may specify a cleaning fluid flow rate of 2L / min, a nozzle moving speed of 10cm / s, and a cleaning time of 1.5 minutes. For the environmental interaction sub-area execution node on the outside of the bottle, the execution parameters may include the cleaning fluid temperature, the nozzle oscillation amplitude, and the cleaning time. The preset parameter table may specify a cleaning fluid temperature of 40℃, a nozzle oscillation amplitude of 60 degrees, and a cleaning time of 1 minute.

[0111] When identifying execution parameters, the system also adjusts preset parameters based on real-time data acquired by the multimodal perception module. For example, if the image sensor detects that the stains on the inner wall are more stubborn than the preset conditions, the system will automatically adjust the cleaning fluid concentration to 6% and the spray pressure to 0.35MPa; if the temperature sensor detects that the current ambient temperature is low, the system will adjust the cleaning fluid temperature in the environmental interaction sub-area to 45℃ to ensure the cleaning effect.

[0112] After identifying the execution parameters of each execution node, the system constructs an error parameter sequence based on these parameters. The order of the parameters in the error parameter sequence is consistent with the cleaning execution order, and each parameter represents the deviation between the actual parameter and the ideal parameter of the corresponding execution node. The ideal parameter is the standard parameter in the preset parameter table, while the actual parameter is the parameter corrected based on real-time data. The error parameter is calculated by subtracting the ideal parameter from the actual parameter. For example, if the ideal parameter for the cleaning fluid concentration in the main area of ​​the inner wall cleaning is 5%, and the actual parameter is corrected to 6%, then the error value of this parameter is +1%; if the ideal parameter for the nozzle spray pressure is 0.3MPa, and the actual parameter is 0.35MPa, the error value is +0.05MPa.

[0113] Constructing the error parameter sequence requires arranging all error parameters of each execution node sequentially according to the execution order. Assuming the inner wall cleaning main sub-region has 5 execution parameters, the bottle mouth cleaning auxiliary sub-region has 3 execution parameters, and the bottle exterior environmental interaction sub-region has 3 execution parameters, then the error parameter sequence will contain 5+3+3=11 parameters, namely: inner wall cleaning fluid concentration error, inner wall nozzle spray pressure error, inner wall nozzle spray angle error, inner wall cleaning time error, inner wall brush rotation speed error, bottle mouth cleaning fluid flow rate error, bottle mouth nozzle moving speed error, bottle mouth cleaning time error, environmental interaction sub-region cleaning fluid temperature error, environmental interaction sub-region nozzle oscillation amplitude error, and environmental interaction sub-region cleaning time error.

[0114] After constructing the error parameter sequence, the system inputs it into the error compensation network. The error compensation network is a pre-trained neural network model, whose structure can be a multilayer perceptron or a recurrent neural network, capable of generating cleaning action outputs for each execution node based on the input error parameter sequence. The network's training data comes from a large number of historical cleaning tasks and the corresponding optimal cleaning action adjustment schemes. Through machine learning algorithms, the network learns the mapping relationship between error parameters and cleaning actions.

[0115] Taking an error of +1% in the concentration of the inner wall cleaning fluid in the input error parameter sequence as an example, the error compensation network, after calculation and processing, may generate an action output to adjust the concentration of the cleaning fluid, that is, to control the metering pump of the bottle washing machine to increase the amount of detergent injected into the cleaning fluid, so as to maintain the deviation between the actual concentration and the ideal concentration within a reasonable range; for a nozzle spray pressure error of +0.05MPa, the network may generate an action output to adjust the opening of the nozzle pressure valve, so that the spray pressure returns to near the ideal value.

[0116] The cleaning action output generated by the error compensation network is a series of specific control commands that can be recognized and executed by the actuators of the bottle washing machine. For example, a command to adjust the cleaning solution concentration will control the motor speed of the metering pump; a command to adjust the nozzle pressure will control the servo motor of the pressure valve; and a command to adjust the brush speed will control the frequency converter of the motor, etc.

[0117] After generating cleaning action outputs, the system sequentially drives the corresponding sub-areas to perform cleaning operations according to these outputs. During the driving process, the actuators of the bottle washer operate according to the received control commands. For example, when cleaning the main cleaning sub-area of ​​the inner wall, the system controls the metering pump to adjust the detergent injection volume according to the commands generated by the error compensation network, so that the cleaning solution concentration reaches the corrected value; it controls the nozzle to spray the cleaning solution onto the inner wall at a pressure of 0.35 MPa and a spray angle of 45 degrees, while simultaneously controlling the brush to rotate at a speed of 150 rpm for 3 minutes of continuous cleaning.

[0118] When cleaning the auxiliary cleaning area at the bottle opening, the system controls the nozzle to spray cleaning fluid at a flow rate of 2L / min and move around the bottle opening at a speed of 10cm / s for 1.5 minutes. When cleaning the environmental interaction area on the outside of the bottle, the system controls the heating device to maintain the cleaning fluid temperature at 45℃ and controls the nozzle to spray cleaning fluid with a swing amplitude of 60 degrees for 1 minute.

[0119] During the process of driving the cleaning operation in each sub-region, the system monitors the cleaning status in real time through the multimodal perception module, obtains the actual execution parameters, and compares them with the expected parameters in the error parameter sequence. If a new deviation is found, the error parameter sequence will be updated in time and re-input into the error compensation network to generate a new cleaning action output, thereby realizing dynamic error compensation and ensuring the accuracy and stability of the cleaning operation.

[0120] For example, during the cleaning of the inner wall, the pressure sensor detects in real time that the nozzle spray pressure is gradually decreasing, creating a new deviation from the expected value in the error parameter sequence. The system will immediately add this deviation value to the error parameter sequence, re-enter it into the error compensation network, generate a new instruction to adjust the nozzle pressure, and control the pressure valve to increase the pressure in time to maintain the stability of the spray pressure.

[0121] The system achieves precise execution of cleaning operations in each sub-area according to the cleaning sequence by identifying and correcting execution parameters, constructing error parameter sequences, processing error compensation networks, and driving and monitoring cleaning actions in real time. This process, combined with a specific bottle cleaning example, utilizes multimodal sensing data and intelligent algorithms to dynamically adjust cleaning parameters, compensate for execution errors, and ensure that the bottle washer can complete cleaning operations efficiently and accurately, improving cleaning quality and reliability. It also adapts to the needs of different cleaning scenarios and bottle types, demonstrating the intelligent and automated characteristics of a small-scale intelligent multimodal laboratory bottle washer.

[0122] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0123] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A cleaning method for a small, intelligent, multimodal laboratory bottle washer, characterized in that, The cleaning method is applied in a small intelligent multimodal laboratory bottle washer, which is pre-configured with a multimodal sensing module. The cleaning method includes: Collect multimodal cleaning data of the bottles to be cleaned currently performing the cleaning task, and obtain the cleaning interaction area between the bottles to be cleaned and the small intelligent multimodal laboratory bottle washer based on the multimodal cleaning data; Based on the cleaning type of the bottle to be cleaned, the cleaning sub-area covered by the cleaning interaction area is divided into a cleaning-dominant sub-area and a cleaning-auxiliary sub-area, and an environmental interaction sub-area is determined in the cleaning sub-area not covered by the cleaning interaction area. The cleaning-dominant sub-area, the cleaning-auxiliary sub-area, and the environmental interaction sub-area constitute a set of areas to be processed. Determine the perceptual fusion discrimination value between any two adjacent sub-regions in the set of regions to be processed, and construct a cleaning weight graph of the set of regions to be processed, using the sub-regions in the set of regions to be processed as vertices and the perceptual fusion discrimination value as the edge connecting the vertices. Determine the maximum spanning tree of the cleaning weight graph, and based on the maximum spanning tree, determine the cleaning execution order of each sub-region in the set of regions to be processed; The cleaning operation is performed by sequentially driving each sub-region in the set of regions to be processed according to the cleaning execution order.

2. The cleaning method of a small intelligent multimodal laboratory bottle washer according to claim 1, characterized in that, The cleaning interaction area between the bottle to be cleaned and the small intelligent multimodal laboratory bottle washer, obtained based on the multimodal cleaning data, includes: Based on the multimodal cleaning data, the cleaning interaction area between the bottle to be cleaned and the small intelligent multimodal laboratory bottle washer in historical cleaning tasks is statistically analyzed. Based on the state of the key parts of the bottle to be cleaned in each cleaning interaction area, the statistically obtained cleaning interaction areas are classified to form one or more sets of cleaning interaction areas. A target set of cleaning interaction areas that matches the current cleaning type of the bottle to be cleaned is determined from the set of cleaning interaction areas, and the area covered by the target set of cleaning interaction areas is taken as the cleaning interaction area between the bottle to be cleaned and the small intelligent multimodal laboratory bottle washing machine.

3. The cleaning method of a small intelligent multimodal laboratory bottle washer according to claim 2, characterized in that, Based on the state of the key parts of the bottle to be cleaned described in each cleaning interaction area, the statistically obtained cleaning interaction areas are categorized as follows: Obtain the predefined state standard range in the small intelligent multimodal laboratory bottle washer; For any cleaning interaction area obtained from statistics, identify the state standard interval to which the state of the key part of the bottle to be cleaned in the cleaning interaction area belongs, and divide the cleaning interaction area into the set of cleaning interaction areas corresponding to the identified state standard interval.

4. The cleaning method of a small intelligent multimodal laboratory bottle washer according to claim 1, characterized in that, Based on the cleaning type of the bottle to be cleaned, the cleaning sub-area covered by the cleaning interaction area is divided into a primary cleaning sub-area and a secondary cleaning sub-area, including: Determine the primary and secondary cleaning areas represented by the cleaning type of the bottle to be cleaned; A first range of motion for the primary cleaning part is determined in the cleaning interaction area, and a second range of motion for the secondary cleaning part is determined in the cleaning interaction area. The cleaning sub-area covered by the first movement range within the cleaning range of the small intelligent multimodal laboratory bottle washer is defined as the primary cleaning sub-area, and the cleaning sub-area covered by the second movement range within the cleaning range of the small intelligent multimodal laboratory bottle washer is defined as the auxiliary cleaning sub-area.

5. The cleaning method of a small intelligent multimodal laboratory bottle washer according to claim 1, characterized in that, Determining environmental interaction sub-regions within cleaning sub-regions not covered by the cleaning interaction area includes: A first extension range is set for the cleaning-dominant sub-region in the cleaning interaction area, and a second extension range is set for the cleaning-assistant sub-region in the cleaning interaction area; For cleaning sub-areas not covered by the cleaning interaction area, if they are adjacent to the cleaning dominant sub-area, the cleaning sub-area of ​​the first extended range is selected as the environmental interaction sub-area; if they are adjacent to the cleaning auxiliary sub-area, the cleaning sub-area of ​​the second extended range is selected as the environmental interaction sub-area.

6. The cleaning method of a small intelligent multimodal laboratory bottle washer according to claim 1, characterized in that, Determining the perceptual fusion discrimination value between any two adjacent sub-regions in the set of regions to be processed includes: For the first and second adjacent sub-regions in the set of regions to be processed, the respective region categories of the first and second sub-regions are identified. The region category includes one of the following: cleaning-dominant sub-region, cleaning-assisted sub-region, and environmental interaction sub-region. Based on the identified region category, determine the first perceptual fusion parameter of the first sub-region and the second perceptual fusion parameter of the second sub-region; Calculate the difference between the first perceptual fusion parameter and the second perceptual fusion parameter, and determine the perceptual fusion distinction value between the first sub-region and the second sub-region based on the difference value.

7. The cleaning method of a small intelligent multimodal laboratory bottle washer according to claim 6, characterized in that, Determining the perceptual fusion distinction value between the first sub-region and the second sub-region based on the difference value includes: Identify the perceptual fusion parameters of each sub-region in the set of regions to be processed, and determine the average value of the identified perceptual fusion parameters; Multiple difference intervals are determined based on the average value, and the target difference interval in which the calculated difference value between the first perception fusion parameter and the second perception fusion parameter is located is determined. The preset discrimination value corresponding to the target difference interval is determined as the perceptual fusion discrimination value between the first sub-region and the second sub-region.

8. The cleaning method of a small intelligent multimodal laboratory bottle washer according to claim 1, characterized in that, The cleaning operation is performed by sequentially driving each sub-region in the set of regions to be processed according to the cleaning execution order, including: The execution parameters of each execution node in the cleaning execution sequence are identified, and an error parameter sequence is constructed based on the identified execution parameters. The order of each parameter in the error parameter sequence is consistent with the cleaning execution sequence. The error parameter sequence is input into the error compensation network to generate cleaning action outputs for each of the execution nodes through the error compensation network; The generated cleaning actions output drive corresponding sub-regions to perform cleaning operations.

9. A cleaning system for a small, intelligent, multimodal laboratory bottle washer, characterized in that, The system is applied in a small intelligent multimodal laboratory bottle washer, which is pre-configured with a multimodal sensing module. The system includes: The data acquisition unit is used to acquire multimodal cleaning data of the bottle to be cleaned currently performing the cleaning task, and to obtain the cleaning interaction area between the bottle to be cleaned and the small intelligent multimodal laboratory bottle washer based on the multimodal cleaning data. The area division unit is used to divide the cleaning sub-area covered by the cleaning interaction area into a cleaning-dominant sub-area and a cleaning-auxiliary sub-area based on the cleaning type of the bottle to be cleaned, and to determine an environmental interaction sub-area in the cleaning sub-area not covered by the cleaning interaction area. The cleaning-dominant sub-area, the cleaning-auxiliary sub-area, and the environmental interaction sub-area constitute a set of areas to be processed. The weight graph construction unit is used to determine the perceptual fusion discrimination value between any two adjacent sub-regions in the set of regions to be processed, and to construct a cleaning weight graph of the set of regions to be processed using the sub-regions in the set of regions to be processed as vertices and the perceptual fusion discrimination value as the edge connecting the vertices. The sequence determination unit is used to determine the maximum spanning tree of the cleaning weight graph, and to determine the cleaning execution order of each sub-region in the set of regions to be processed based on the maximum spanning tree. An operation execution unit is used to sequentially drive each sub-region in the set of regions to be processed according to the cleaning execution order to perform cleaning operations; The operation execution unit is specifically used to: identify the execution parameters of each execution node in the cleaning execution sequence, and construct an error parameter sequence based on the identified execution parameters, wherein the order of each parameter in the error parameter sequence is consistent with the cleaning execution sequence; input the error parameter sequence into an error compensation network to generate cleaning action outputs for each execution node through the error compensation network; and drive the corresponding sub-regions according to the generated cleaning action outputs to perform cleaning operations.

10. A small, intelligent, multimodal laboratory bottle washing machine, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the computer program to implement the cleaning method of the small intelligent multimodal laboratory bottle washer according to any one of claims 1-8.

Images