A method and system for path planning of a cleaning vehicle

Abstract

This invention relates to a method and system for planning cleaning vehicle paths. The method includes acquiring environmental attribute information of a target cleaning area and determining the cleaning strategy parameters corresponding to each sub-area of ​​the target cleaning area. Based on the degree of difference in cleaning strategy parameters between adjacent sub-areas, the target cleaning area is divided into multiple policy-coordinated zones. The target cleaning strategy parameters for each policy-coordinated zone are determined according to the cleaning strategy parameters corresponding to each sub-area within the policy-coordinated zone. The strategy switching cost between adjacent policy-coordinated zones is determined based on the positional relationship between the policy-coordinated zones and the target cleaning strategy parameters. A cleaning path is planned within the target cleaning area with the objective of minimizing the total cost of the cost model, where the cost term of the cost model includes path length cost and strategy switching cost. This method enables path planning suitable for high-intensity scenarios with variable workloads, such as road cleaning.

Description

Technical Field

[0001] This invention relates to the field of sanitation, and in particular to a method and system for planning the route of cleaning vehicles. Background Technology

[0002] With the advancement of network communication capabilities and intelligent application technologies, intelligent sweepers can be used to achieve unmanned, automated sanitation work. Currently, the mainstream approach for sweeper operations is to borrow the full-coverage algorithm from indoor robotic vacuum cleaners. The core objective of this approach is to achieve full area coverage with the shortest geometric path (highest efficiency). However, sweepers differ significantly from ordinary indoor robotic vacuum cleaners in terms of workload and the difficulty of cleaning environments. This means that traditional full-coverage path algorithms are not well-suited to the working scenarios of sanitation cleaning. Summary of the Invention

[0003] The purpose of this application is to propose a multi-source data fusion method and system based on a smart construction and management cloud platform, addressing the problem that traditional full-coverage path algorithms cannot well adapt to the work scenarios of sanitation cleaning.

[0004] The cleaning vehicle path planning method of this application includes:

[0005] Obtain environmental attribute information of the target cleaning area;

[0006] Based on the environmental attribute information, the cleaning strategy parameters corresponding to each sub-area of ​​the target cleaning area are determined;

[0007] Based on the degree of difference in the cleaning strategy parameters between adjacent sub-regions, the target cleaning area is divided into multiple homogeneous policy zones, wherein the degree of difference in the cleaning strategy parameters between sub-regions within the same homogeneous policy zone is less than a preset threshold.

[0008] Based on the cleaning strategy parameters corresponding to each sub-region within the same policy area, determine the cleaning strategy parameters that satisfy all sub-regions of the same policy area as the target cleaning strategy parameters for the same policy area.

[0009] A directed topology graph representing the connectivity between the policy-similar regions is constructed. In this graph, directed edges from a first policy-similar region to a second policy-similar region are associated with a first policy switching cost, which is related to the actuator adjustment required to switch from the cleaning policy corresponding to the first policy-similar region to the cleaning policy corresponding to the second policy-similar region. Directed edges from the second policy-similar region to the first policy-similar region are associated with a second policy switching cost, which is also related to the actuator adjustment required to switch from the cleaning policy corresponding to the second policy-similar region to the cleaning policy corresponding to the first policy-similar region. Furthermore, the values ​​of the first and second policy switching costs are configured to be determined asymmetrically based on the direction of actuator parameter change required for the corresponding switching direction.

[0010] With the goal of minimizing the total cost of the cost model, a cleaning path is planned within the target cleaning area. The cost component of the cost model includes path length cost and strategy switching cost.

[0011] Optionally, dividing the target cleaning area into multiple homogeneous policy zones based on the degree of difference in cleaning strategy parameters between adjacent sub-regions includes:

[0012] The policy difference between two adjacent sub-regions is determined by normalized weighted Euclidean distance; different sweeping policy parameters are assigned severity weights, which reflect the severity of the impact caused by switching the parameter.

[0013] Optionally, after determining the policy difference between two adjacent sub-regions using normalized weighted Euclidean distance, the method further includes:

[0014] Based on the policy difference degree, the same policy region is divided into segments using a region growing segmentation algorithm.

[0015] Optionally, the first strategy switching cost is determined by the transient energy consumption or impact level of the actuator when switching from the cleaning strategy corresponding to the first same strategy zone to the cleaning strategy corresponding to the second same strategy zone; the second strategy switching cost is determined by the transient energy consumption or impact level of the actuator when switching from the cleaning strategy corresponding to the second same strategy zone to the cleaning strategy corresponding to the first same strategy zone; the unit cost coefficient of the actuator acceleration process is different from the unit cost coefficient of the deceleration process.

[0016] Optionally, the total cost of the cost model is a weighted sum of the path length cost and the policy switching cost, wherein the weight of the policy switching cost is an adjustable global parameter.

[0017] Optionally, obtaining the environmental attribute information of the target cleaning area includes:

[0018] Acquire visual inspection images of the target cleaning area;

[0019] The visual inspection images are processed by a pre-trained semantic segmentation model to classify the ground conditions of each sub-region.

[0020] The category with the highest percentage will be used as the environmental attribute label for the sub-region;

[0021] The cleaning strategy parameters corresponding to each sub-region of the target cleaning area include:

[0022] Based on the environmental attribute tags, the corresponding cleaning strategy parameters are matched from the preset configuration.

[0023] Optionally, the ground conditions of each sub-area can be categorized, including road surface category and dirt category.

[0024] Optionally, the cleaning strategy parameters include at least one of the following: driving speed, cleaning force, suction power, and water spray volume.

[0025] Optionally, the strategy switching cost is determined based on the change magnitude of at least one parameter in the cleaning strategy parameters and a unit cost coefficient preset for that parameter.

[0026] On the other hand, this application also provides a cleaning vehicle path planning system, including:

[0027] The strategy planning unit is used to acquire environmental attribute information of the target cleaning area and determine the cleaning strategy parameters corresponding to each sub-area of ​​the target cleaning area based on the environmental attribute information.

[0028] A region division unit is used to divide the target cleaning area into multiple homogeneous policy zones based on the degree of difference in the cleaning strategy parameters between adjacent sub-regions, wherein the degree of difference in the cleaning strategy parameters between sub-regions within the same homogeneous policy zone is less than a preset threshold.

[0029] The strategy merging unit is used to determine the cleaning strategy parameters that satisfy all sub-regions of the same policy area based on the cleaning strategy parameters corresponding to each sub-region within the same policy area, and use them as the target cleaning strategy parameters for the same policy area.

[0030] The cost determination unit constructs a directed topology graph representing the connectivity between the co-policy zones. In this graph, directed edges from the first co-policy zone to the second co-policy zone are associated with a first policy switching cost, which is related to the actuator adjustment required to switch from the cleaning policy corresponding to the first co-policy zone to the cleaning policy corresponding to the second co-policy zone. Directed edges from the second co-policy zone to the first co-policy zone are associated with a second policy switching cost, which is also related to the actuator adjustment required to switch from the cleaning policy corresponding to the second co-policy zone to the cleaning policy corresponding to the first co-policy zone. Furthermore, the values ​​of the first and second policy switching costs are configured to be determined asymmetrically based on the direction of actuator parameter change required for the corresponding switching direction.

[0031] The path generation unit is used to plan a cleaning path within the target cleaning area with the goal of minimizing the total cost of the cost model, wherein the cost term of the cost model includes path length cost and strategy switching cost.

[0032] The cleaning vehicle path planning method provided in this application merges areas with similar cleaning strategies, reducing the frequency of strategy switching and avoiding excessive area division. When planning the cleaning path, a strategy switching cost is introduced, resulting in a shorter path with a lower strategy switching cost, thus balancing efficiency, cost, and maintainability. Compared to ordinary full-coverage algorithms, it is better suited for high-intensity scenarios with variable workloads, such as road cleaning. Attached Figure Description

[0033] Figure 1 A basic flowchart illustrating the cleaning vehicle path planning method provided in this application embodiment;

[0034] Figure 2 This is a schematic diagram of the structure of the sweeping vehicle path planning system provided in the embodiments of this application. Detailed Implementation

[0035] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0036] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0037] Example 1:

[0038] This embodiment provides a method for planning the path of a cleaning vehicle. Please refer to [link / reference]. Figure 1 This method includes, but is not limited to, the following steps:

[0039] S101. Obtain environmental attribute information of the target cleaning area;

[0040] Environmental attribute information reflects the environmental conditions of the target cleaning area, such as the condition of the ground and the level of dirt, providing a reference for subsequent cleaning strategies.

[0041] Environmental attribute information can be obtained through methods such as acquiring patrol data (e.g., drone patrol images), real-time photography, and real-time sensing.

[0042] S102. Based on the environmental attribute information, determine the cleaning strategy parameters corresponding to each sub-area of ​​the target cleaning area;

[0043] In practical applications, sweepers may need to adopt different cleaning methods depending on different environmental conditions, such as adjusting their cleaning intensity and speed. These cleaning methods can be switched by adjusting parameters such as vehicle speed and the working power / mode of the cleaning unit. In practical applications, a single sub-area is cleaned using the same cleaning strategy parameters. Based on this, the target cleaning area can be divided into several different sub-areas, each using different cleaning strategy parameters. For example, during planning, the area can be divided according to a predetermined grid size as the smallest unit, with each sub-area including at least one grid.

[0044] S103. Based on the degree of difference in cleaning strategy parameters between adjacent sub-regions, the target cleaning area is divided into multiple areas with the same strategy.

[0045] Within the same policy zone, the difference in cleaning strategy parameters between sub-regions is less than a preset threshold. In this embodiment, sub-regions with small differences in cleaning strategy parameters are merged into one policy zone. This reduces the number of times the sweeper switches its operating strategy during operation, allowing its operating modules to work continuously in a relatively stable state. This avoids frequent changes in the operating power of units such as motors, reducing their wear and tear. During high-intensity outdoor operations, this makes the sweeper's operation more stable, reduces maintenance frequency, and increases the lifespan of its components. In some implementations, area considerations can also be introduced, dynamically setting preset thresholds based on the area of ​​sub-regions or policy zones. For example, when the average area of ​​sub-regions is large, a smaller preset threshold is set to reduce the number of sub-regions within the policy zone, achieving more precise cleaning; when the average area of ​​sub-regions is small, a larger preset threshold is set to expand the area of ​​the policy zone, reducing the frequency of cleaning strategy switching.

[0046] S104. Based on the cleaning strategy parameters corresponding to each sub-area within the same policy area, determine the cleaning strategy parameters that satisfy all sub-areas within the same policy area as the target cleaning strategy parameters for the same policy area.

[0047] Generally speaking, among the cleaning strategy parameters corresponding to each sub-area within the same policy zone, the cleaning strategy parameter with the highest cleaning ability can be selected as the target cleaning strategy parameter for the same policy zone.

[0048] S105. Construct a directed topological graph representing the connectivity between regions with the same policy;

[0049] In the directed topology graph, the directed edge from the first co-policy region to the second co-policy region is associated with a first policy switching cost, which is related to the actuator adjustment range required to switch from the cleaning policy corresponding to the first co-policy region to the cleaning policy corresponding to the second co-policy region. Similarly, the directed edge from the second co-policy region to the first co-policy region is associated with a second policy switching cost, which is also related to the actuator adjustment range required to switch from the cleaning policy corresponding to the second co-policy region to the cleaning policy corresponding to the first co-policy region. Furthermore, the values ​​of the first and second policy switching costs are configured to be determined asymmetrically based on the direction of change in the actuator parameters required for the corresponding switching direction. In this embodiment, the actuator refers to the physical unit or device that actually performs the corresponding action when the corresponding cleaning policy changes, such as a motor or fan, and its adjustment range can be expressed as speed, power percentage, etc.

[0050] It is understandable that after a sweeper enters another designated cleaning zone to perform its work, its cleaning strategy parameters need to be adjusted accordingly. In practical applications, the process of adjusting the sweeper's cleaning strategy is not lossless. It involves not only the efficiency impact caused by waiting for various actuators to switch to the new working conditions, but also the mechanical shock losses caused by braking and motor acceleration / deceleration. In this embodiment, these impacts are transformed into a quantitative description of the strategy switching cost. In practical applications, cost strength coefficients for different actuators when various strategy parameters are changed can be set according to the actual degree of impact, so that the strategy switching cost can effectively reflect the actual strength of the impact.

[0051] For example, when entering a gravel spill area from a standard road surface, the side brush motor needs to suddenly accelerate, and the vacuum fan needs to quickly establish a high negative pressure. This process is accompanied by a large surge current and mechanical impact. When entering a standard road surface from a gravel spill area, the side brush motor slows down, and the fan depressurizes, resulting in much less impact and transient energy consumption. In traditional path planning methods, not only is the cost of area switching usually ignored, but the physical asymmetry between different areas is also completely disregarded. In this embodiment, the "asymmetry" of the cost of switching cleaning strategies is introduced. The cost of switching cleaning strategies is determined asymmetrically based on the area switching method, and a cost model reflecting the ignored physical asymmetry is built, enabling refined job scheduling.

[0052] The cost model of a weighted directed topology graph can be effectively adapted to the path planning method of this embodiment. In practical applications, other costs can also be introduced. In this case, the corresponding cost term can be added to the edges connecting the nodes. This gives the path planning method of this embodiment a certain potential for expansion, allowing more factors to be considered to form a more complete global path planning method.

[0053] S106. With the goal of minimizing the total cost of the cost model, plan the cleaning path within the target cleaning area;

[0054] The cost model includes path length cost and strategy switching cost. It is understood that the cleaning path planned based on the steps described above in this embodiment has a shorter path length and a lower strategy switching cost, thus balancing efficiency, cost, and maintainability. Compared to the full-coverage algorithm of ordinary household robot vacuums, the method in this embodiment is better suited for high-intensity scenarios with variable workloads, such as road cleaning.

[0055] In some implementations, the target cleaning area is divided into multiple homogeneous policy zones based on the degree of difference in cleaning strategy parameters between adjacent sub-regions, including:

[0056] The policy difference between two adjacent sub-regions is determined by normalized weighted Euclidean distance; different sweeping policy parameters are assigned severity weights, which reflect the severity of the impact caused by switching parameters.

[0057] It is understandable that different cleaning strategy parameters may have different impacts when switching. For example, the wear and tear on the wheel motor, brush motor, and vacuum fan differs during start-up and shutdown. Generally, units that are more susceptible to mechanical shock and wear when adjusting their operating status, or units with shorter lifespans, will be assigned a higher severity weight to minimize frequent start-up, shutdown, or power adjustments. This embodiment defines the differences between strategies by the severity of the impact caused by parameter switching, rather than simply setting a fixed parameter change threshold. This allows the differences between strategies to actually reflect the potential impact of switching between different strategies, truly merging areas with similar cleaning needs and ensuring cleaning efficiency within the same strategy zone.

[0058] In some implementations, after determining the policy dissimilarity between two adjacent sub-regions using normalized weighted Euclidean distance, the method further includes:

[0059] Based on policy dissimilarity, a region-growing-based segmentation algorithm divides regions into areas with similar policies. Of course, in other examples, other algorithms can also be used to divide these areas, as long as they can effectively merge sub-regions that meet the policy dissimilarity requirements.

[0060] In some implementations, the first strategy switching cost is determined by the transient energy consumption or impact level of the actuator when switching from the cleaning strategy corresponding to the first same strategy zone to the cleaning strategy corresponding to the second same strategy zone; the second strategy switching cost is determined by the transient energy consumption or impact level of the actuator when switching from the cleaning strategy corresponding to the second same strategy zone to the cleaning strategy corresponding to the first same strategy zone; the unit cost coefficient of the actuator acceleration process is different from the unit cost coefficient of the deceleration process.

[0061] In some implementations, the total cost of the cost model is a weighted sum of path length cost and policy switching cost, wherein the weight of the policy switching cost is an adjustable global parameter.

[0062] The adjustable weighting of strategy switching costs allows sweeper users to balance path length and strategy switching based on actual conditions.

[0063] In some implementations, obtaining environmental attribute information of the target cleaning area includes:

[0064] Acquire visual inspection images of the target cleaning area;

[0065] The visual inspection images are processed by a pre-trained semantic segmentation model to classify the ground conditions of each sub-region.

[0066] The category with the highest percentage will be used as the environmental attribute label for the sub-region;

[0067] The cleaning strategy parameters corresponding to each sub-area of ​​the target cleaning area include:

[0068] Based on the environmental attribute tags, the corresponding cleaning strategy parameters are matched from the preset configuration.

[0069] In some implementations, the ground conditions of each sub-area are categorized, including road surface type and dirt type. For example, road surface types include asphalt, cement, and asphalt-gravel roads; dirt types include standing water, oil stains, and dust. In some examples, the degree of dirtiness may also be included to adjust the cleaning intensity according to different levels. In practical applications, special areas that are impassable or impossible to clean can be marked.

[0070] In this embodiment, the sweeping strategy parameters include at least one of the following: driving speed, sweeping force, suction power, and water spray volume. In practical applications, the specific items of the sweeping strategy parameters will be determined based on the sweeper's functions and sweeping requirements.

[0071] To better understand the method of this embodiment, further explanation is provided below with more specific examples.

[0072] In this example, data acquisition can be accomplished through images obtained during drone patrols and multimodal sensors mounted on the sweeper vehicle. During drone patrols, radar and / or cameras acquire structural and visual information about the target cleaning area. The sweeper vehicle can be equipped with radar, cameras, vehicle attitude sensors, vibration sensors, and other sensors to acquire information including, but not limited to, the structural and visual information of the target cleaning area, the vehicle's own pose, and road surface material. Combining this information, the environmental attributes of the target cleaning area can be determined.

[0073] In this example, existing positioning and mapping algorithms can be used to fuse various data to form a 3D point cloud map of the target cleaning area. For instance, this map may include static geometric information such as road boundaries, curbs, and traffic signs.

[0074] A 3D point cloud map can be projected onto a 2D plane and divided into several square grids. For example, the grid can be divided into 2-meter by 2-meter grids, and each grid can be assigned unique spatial coordinates.

[0075] In this example, the acquired sensor information is processed using a semantic segmentation model based on a deep convolutional neural network. The semantic segmentation model is pre-trained and can take road surface images or other sensor data as input, identify their corresponding environmental attributes, and output the results. Specifically, this example can identify road surface type, dirt level, etc. As an example, the training set for the semantic segmentation model can consist of urban road surface images covering different lighting conditions, weather, and seasons. Elements in the images can be manually labeled, such as road surface type, dirt type, and dirt level, and the semantic segmentation model can be trained based on this labeling.

[0076] In this example, the structure of the semantic segmentation model can be mapped onto a grid. For each grid, the semantic categories of all pixels can be counted, and the category with the highest percentage can be used as its environmental attribute label.

[0077] To facilitate matching cleaning strategy parameters, cleaning strategy parameters corresponding to various environmental attribute labels can be pre-stored through lookup tables or other methods.

[0078] As an example, assuming a sweeper includes side brushes, a vacuum fan, and a water sweeping device, the sweeping strategy parameters might be represented as a four-dimensional vector (v, ω_brush, P_vac, Q_water), where v is the sweeper's speed, ω_brush is the side brush rotation speed, P_vac is the vacuum fan power percentage, and Q_water is the water spray volume. For example, the sweeping strategy parameter set mapped to a cement road with ordinary dust pollution is P_std = (10, 500, 60, 1.0); the sweeping strategy parameter set corresponding to a gravel road with ordinary dust pollution is P_sand = (3, 800, 100, 2.5); and the sweeping strategy parameter set mapped to a cement road with heavy oil pollution is P_oil = (2, 900, 100, 0). In practical applications, the mapping relationship between sweeping strategy parameters and environmental attribute labels can be set or modified by sanitation workers according to actual sweeping needs.

[0079] After the semantic segmentation model completes its identification, each grid can be matched with a specific set of cleaning strategy parameters.

[0080] Of course, this example will also cluster each grid to generate a region with the same strategy. When defining the strategy difference, this example represents the strategy difference between grids A and B as D(A,B). As an example, the strategy difference can be calculated using a normalized weighted Euclidean distance formula as shown in Equation 1, where P_{max,k} is the maximum allowable value of the k-th parameter, used for normalization; the weight w_k reflects the severity of the impact / loss caused by switching different parameters. For example, the impact of switching the suction fan power P_vac is the greatest, and its weight w_3 can be set to 2.0; while the impact of changes in driving speed is smaller, and the driving speed weight w_1 can be set to 1.0. By setting the Euclidean distance calculation weight based on the impact of parameter switching, Equation 1 quantifies the degree of influence of strategy switching, thereby reflecting the difference between the two strategies.

[0081] Equation 1: D(A,B) = sqrt( Σ_{k=1}^4 w_k * ( (P_{a,k} - P_{b,k}) / P_{max,k})^2 )

[0082] This example uses a region growing segmentation algorithm to generate homogeneous regions. First, each grid cell is marked as "unvisited." A randomly selected "unvisited" grid cell is used as the seed for a new region, added to the growth queue, and marked as "visited." A grid cell is removed from the queue, and its neighboring grid cells are checked. For each "unvisited" neighboring grid cell, its policy difference with the current seed grid is calculated. If the policy difference is less than a preset threshold, it is considered to be of the same type, added to the queue, and assigned to the seed grid's region. This region growing process is repeated until the growth queue is empty, at which point a region is formed. A new "unvisited" grid cell is selected as the seed for a new region until the entire map is segmented into multiple non-overlapping homogeneous regions.

[0083] To enable path planning algorithms to handle co-policy regions, this example constructs a weighted directed topology graph for the optimization algorithm to search. For example, a directed graph G = (V, E) is constructed, where V is the set of nodes, each node corresponding to a co-policy region. In practical applications, a starting node is also specified, representing the initial position of the sweeper. E is the set of edges; if two co-policy regions are spatially adjacent, a set of directed edges connecting the nodes of these two regions is created. In the directed graph G, each edge is assigned a cost, which is not a simple geometric distance but a composite cost formed by path length and policy switching penalty. Furthermore, a directed edge from the first co-policy region to the second co-policy region and a directed edge from the second co-policy region to the first co-policy region are associated with different policy switching costs. The composite cost C can be expressed as C = C_dis + λ * C_swi, where C_dis is the path length cost, C_swi is the policy switching cost, and λ is the adjustment weight of the policy switching cost. The strategy switching cost reflects the negative impact of switching cleaning strategy parameters when a sweeper moves from one policy zone to another. As an example, the strategy switching cost can be defined as Equation 2, where β_k is the switching cost coefficient for a unit change in the k-th parameter. In practical applications, the switching cost coefficient for each parameter's unit change is set according to the actual impact of the change. For example, for every 1% change in suction fan power, β_vac can be set to 0.8 cost units. It can be understood that as λ increases, the path results tend to reduce the frequency of cleaning strategy switching; as λ increases, the path results tend to have shorter path lengths. Therefore, λ can be flexibly adjusted by the sweeper's operator based on factors such as the vehicle's age / wear and tear or maintenance strategies.

[0084] Formula 2: C_swi=Σ_{k=1}^4 β_k * |P_{q,k} - P_{p,k}|

[0085] In this example, path planning can be modeled as a minimum-cost traversal model with region constraints, determining the path covering the regions represented by all nodes and finding the result with the minimum composite cost. Exemplarily, this can be solved using an improved heuristic algorithm, specifically a tabu search algorithm based on the cost model. An initial solution is constructed using a greedy strategy of "minimum composite cost first." Starting from the initial node, in each unvisited node in the same policy region, the node with the minimum composite cost associated with the directed edge originating from the current node is selected as the next node to be visited, until all nodes are visited. A neighborhood solution is generated by performing one of the following: a swap move, an insertion move, or a 2-Opt move on the current path. For each neighborhood solution, its total composite cost is calculated as the evaluation function. A tabu list is used to record recently visited solutions to avoid loops. In each iteration, the neighborhood solution with the minimum total cost that is not in the tabu list is selected as the current solution. If the current optimal solution is not improved for several consecutive iterations, a perturbation mechanism can be triggered, randomly performing several directed insertion moves to escape local optima. Continue iterating until the maximum number of iterations (e.g., 500) is reached or the number of consecutive iterations with no improvement (e.g., 50) is reached, to obtain the final result.

[0086] Understandably, once the global access order is determined, a fully covered path (such as a bow-shaped path or other fully covered path) is generated within each co-playing zone. The starting point of this path is the grid containing the entry edge, and the ending point is the grid containing the exit edge pointing to the next co-playing zone. Within the co-playing zone, vehicles can maintain stable parameters and operate at a uniform speed.

[0087] If a traditional path planning algorithm is used to execute the corresponding cleaning strategy, the final result might be that the sweeper simply traverses all areas according to the shortest path, switching to the corresponding cleaning strategy parameters when entering different environments. To ensure the shortest travel path, the sweeper might repeatedly enter and exit different environments, frequently switching cleaning strategy parameters. However, when performing cleaning operations using the path planning method in this example, the sweeper might first complete the cleaning of a large, continuous area with the same cleaning strategy parameters without needing to switch parameters, and then perform a cleaning strategy switch once to enter the next area with the same strategy to complete the work. This avoids the frequent switching of cleaning strategies caused by frequently entering and exiting different working environments to shorten the path, improving the overall energy efficiency of the cleaning operation, reducing maintenance costs, and also taking path efficiency into consideration.

[0088] This embodiment also provides a cleaning vehicle path planning system 100, see [link / reference] Figure 2 As shown, it includes, but is not limited to: strategy planning unit 101, region division unit 102, strategy merging unit 103, cost determination unit 104, and path generation unit 105.

[0089] The strategy planning unit 101 is used to obtain the environmental attribute information of the target cleaning area and determine the cleaning strategy parameters corresponding to each sub-area of ​​the target cleaning area based on the environmental attribute information.

[0090] The region division unit 102 is used to divide the target cleaning area into multiple homogeneous policy zones based on the degree of difference in cleaning strategy parameters between adjacent sub-regions. Among them, the degree of difference in cleaning strategy parameters between sub-regions within the same homogeneous policy zone is less than a preset threshold.

[0091] The strategy merging unit 103 is used to determine the cleaning strategy parameters that satisfy all sub-regions of the same policy area based on the cleaning strategy parameters corresponding to each sub-region within the same policy area, and use them as the target cleaning strategy parameters for the same policy area.

[0092] The cost determination unit 104 is used to construct a directed topology graph representing the connectivity between policy-similar regions. In the directed topology graph, the directed edge from the first policy-similar region to the second policy-similar region is associated with a first policy switching cost, which is related to the actuator adjustment range required when switching from the cleaning policy corresponding to the first policy-similar region to the cleaning policy corresponding to the second policy-similar region. The directed edge from the second policy-similar region to the first policy-similar region is associated with a second policy switching cost, which is related to the actuator adjustment range required when switching from the cleaning policy corresponding to the second policy-similar region to the cleaning policy corresponding to the first policy-similar region. The values ​​of the first policy switching cost and the second policy switching cost are configured to be determined asymmetrically according to the direction of change of actuator parameters required for the corresponding switching direction.

[0093] The path generation unit 105 is used to plan a cleaning path within the target cleaning area with the goal of minimizing the total cost of the cost model. The cost items of the cost model include path length cost and strategy switching cost.

[0094] The specific execution process of each unit in the cleaning vehicle path planning system 100 can also refer to the steps of the cleaning vehicle path planning method provided above in this embodiment, and will not be repeated in this embodiment.

[0095] Furthermore, although exemplary embodiments have been described herein, their scope includes any and all embodiments based on this disclosure that have equivalents, modifications, omissions, combinations (e.g., schemes involving intersections of various embodiments), adaptations, or changes. They are not limited to the examples described in this specification or during the implementation of this application, and such examples are to be interpreted as non-exclusive.

[0096] The above description is intended to be illustrative and not restrictive. For example, the above examples (or one or more of them) can be used in combination with each other. Other embodiments can be used by those skilled in the art when reading the above description.

[0097] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, various modifications and variations can be made to the embodiments of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method of path planning for a cleaning vehicle, characterized by, include: Obtain environmental attribute information of the target cleaning area; Based on the environmental attribute information, the cleaning strategy parameters corresponding to each sub-area of ​​the target cleaning area are determined; Based on the degree of difference in the cleaning strategy parameters between adjacent sub-regions, the target cleaning area is divided into multiple homogeneous policy zones, wherein the degree of difference in the cleaning strategy parameters between sub-regions within the same homogeneous policy zone is less than a preset threshold. Based on the cleaning strategy parameters corresponding to each sub-region within the same policy area, determine the cleaning strategy parameters that satisfy all sub-regions of the same policy area as the target cleaning strategy parameters for the same policy area. A directed topology graph representing the connectivity between the policy-similar regions is constructed. In this graph, directed edges from a first policy-similar region to a second policy-similar region are associated with a first policy switching cost, which is related to the actuator adjustment required to switch from the cleaning policy corresponding to the first policy-similar region to the cleaning policy corresponding to the second policy-similar region. Directed edges from the second policy-similar region to the first policy-similar region are associated with a second policy switching cost, which is also related to the actuator adjustment required to switch from the cleaning policy corresponding to the second policy-similar region to the cleaning policy corresponding to the first policy-similar region. Furthermore, the values ​​of the first and second policy switching costs are configured to be determined asymmetrically based on the direction of actuator parameter change required for the corresponding switching direction. With the goal of minimizing the total cost of the cost model, a cleaning path is planned within the target cleaning area. The cost component of the cost model includes path length cost and strategy switching cost.

2. The cleaning vehicle path planning method of claim 1, wherein, The process of dividing the target cleaning area into multiple homogeneous policy zones based on the degree of difference in cleaning strategy parameters between adjacent sub-regions includes: The policy difference between two adjacent sub-regions is determined by normalized weighted Euclidean distance; different sweeping policy parameters are assigned severity weights, which reflect the severity of the impact caused by switching the parameter.

3. The cleaning vehicle path planning method of claim 2, wherein, After determining the policy dissimilarity between two adjacent sub-regions using normalized weighted Euclidean distance, the process also includes: Based on the policy difference degree, the same policy region is divided into segments using a region growing segmentation algorithm.

4. The cleaning vehicle path planning method of claim 1, wherein, The first strategy switching cost is determined by the transient energy consumption or impact level of the actuator when switching from the cleaning strategy corresponding to the first same strategy zone to the cleaning strategy corresponding to the second same strategy zone; the second strategy switching cost is determined by the transient energy consumption or impact level of the actuator when switching from the cleaning strategy corresponding to the second same strategy zone to the cleaning strategy corresponding to the first same strategy zone; the unit cost coefficient of the actuator acceleration process is different from the unit cost coefficient of the deceleration process.

5. The cleaning vehicle path planning method of claim 1, wherein, The total cost of the cost model is a weighted sum of path length cost and policy switching cost, wherein the weight of the policy switching cost is an adjustable global parameter.

6. The cleaning vehicle path planning method of claim 1, wherein, The acquisition of environmental attribute information of the target cleaning area includes: Acquire visual inspection images of the target cleaning area; The visual inspection images are processed by a pre-trained semantic segmentation model to classify the ground conditions of each sub-region. The category with the highest percentage will be used as the environmental attribute label for the sub-region; The cleaning strategy parameters corresponding to each sub-region of the target cleaning area include: Based on the environmental attribute tags, the corresponding cleaning strategy parameters are matched from the preset configuration.

7. The cleaning vehicle path planning method of claim 6, wherein, The ground conditions in each sub-area are categorized, including road surface type and dirt type.

8. The cleaning vehicle path planning method of any one of claims 1-7, wherein, The cleaning strategy parameters include at least one of the following: driving speed, cleaning force, suction power, and water spray volume.

9. The cleaning vehicle path planning method as described in claim 8, characterized in that, The cost of switching strategies is determined based on the magnitude of change of at least one parameter in the cleaning strategy parameters and a unit cost coefficient preset for that parameter.

10. A cleaning vehicle path planning system, characterized in that, include: The strategy planning unit is used to acquire environmental attribute information of the target cleaning area and determine the cleaning strategy parameters corresponding to each sub-area of ​​the target cleaning area based on the environmental attribute information. A region division unit is used to divide the target cleaning area into multiple homogeneous policy zones based on the degree of difference in the cleaning strategy parameters between adjacent sub-regions, wherein the degree of difference in the cleaning strategy parameters between sub-regions within the same homogeneous policy zone is less than a preset threshold. The strategy merging unit is used to determine the cleaning strategy parameters that satisfy all sub-regions of the same policy area based on the cleaning strategy parameters corresponding to each sub-region within the same policy area, and use them as the target cleaning strategy parameters for the same policy area. The cost determination unit constructs a directed topology graph representing the connectivity between the co-policy zones. In this graph, directed edges from the first co-policy zone to the second co-policy zone are associated with a first policy switching cost, which is related to the actuator adjustment required to switch from the cleaning policy corresponding to the first co-policy zone to the cleaning policy corresponding to the second co-policy zone. Directed edges from the second co-policy zone to the first co-policy zone are associated with a second policy switching cost, which is also related to the actuator adjustment required to switch from the cleaning policy corresponding to the second co-policy zone to the cleaning policy corresponding to the first co-policy zone. Furthermore, the values ​​of the first and second policy switching costs are configured to be determined asymmetrically based on the direction of actuator parameter change required for the corresponding switching direction. The path generation unit is used to plan a cleaning path within the target cleaning area with the goal of minimizing the total cost of the cost model, wherein the cost term of the cost model includes path length cost and strategy switching cost.

Images