Distributed car washing robot intelligent scheduling decision control method and system

Abstract

The invention relates to the technical field of intelligent scheduling, and particularly discloses an intelligent scheduling decision control method and system for a distributed car washing robot, and the method comprises the steps: constructing a weighted graph model which comprises a vehicle position, obstacle distribution and path reachability, and dynamically adjusting the weight of an edge according to the path length, energy consumption and time; optimizing the graph structure by adopting an improved minimum spanning tree algorithm, and calculating a path efficiency index in combination with fuzzy logic and a neural network technology; on the basis of the optimized subgraph, applying a shortest path algorithm to obtain a local optimal path, and introducing a dynamic adaptation coefficient to evaluate the response capability of the path to environment change; fusing the path efficiency index and the dynamic adaptation coefficient into a comprehensive scheduling feature vector, inputting the comprehensive scheduling feature vector into a trained deep learning scheduling model, and predicting an optimal scheduling strategy in combination with a task priority and a robot resource state; and dynamically adjusting the moving path and the operation density of the robot according to a scheduling result, and realizing self-adaptive evolution and cooperative control of the system.

Description

Technical Field

[0001] This invention relates to the field of intelligent scheduling technology, specifically to an intelligent scheduling decision-making and control method and system for distributed car wash robots. Background Technology

[0002] Distributed car wash robot systems are gaining attention as an emerging intelligent service model. This system deploys multiple mobile robots with autonomous navigation and cleaning capabilities, enabling flexible and efficient vehicle cleaning without requiring a fixed washing location.

[0003] The existing technology has the following shortcomings: A key technical challenge in applying Pareto optimization to path planning in distributed car wash robot systems is the excessively large set of non-dominated solutions. This problem stems from the optimization process identifying numerous seemingly "optimal" path choices, each with unique advantages and disadvantages, especially when considering multiple objectives such as path length, time consumption, and energy efficiency. This phenomenon significantly increases decision-making complexity, making it difficult for the system to quickly select the optimal path under tight time constraints, thus affecting the timeliness and service quality of the overall task execution. Specifically, in situations requiring rapid response, such as prioritizing emergency vehicles or avoiding sudden obstacles, a large set of non-dominated solutions can lead to sluggish system response, thereby weakening the user experience and reducing operational efficiency. Therefore, effectively reducing the size of the non-dominated solution set and improving decision-making speed without sacrificing solution quality has become a pressing technical challenge for improving the system's performance. Summary of the Invention

[0004] The purpose of this invention is to provide a distributed car wash robot intelligent scheduling decision control method and system to solve the problems mentioned above.

[0005] The objective of this invention can be achieved through the following technical solutions: A distributed car wash robot intelligent scheduling and decision-making control method includes the following steps: S1: Construct a weighted graph model that includes vehicle location, obstacle distribution, and path reachability, and set the weights of the edges according to path length, energy consumption, and time; wherein, the vehicle location is defined as a node; S2: The minimum spanning tree algorithm is used to perform preliminary optimization of the weighted graph, identify the path structure with the minimum total weight that covers all nodes, and calculate the path efficiency index to characterize the overall scheduling efficiency of each candidate path. S3: Based on the subgraph after the minimum spanning tree selection, the shortest path algorithm is applied to determine the local optimal path between the starting point and the target point, and the dynamic adaptation coefficient is calculated in combination with dynamic environment information to reflect the path's responsiveness to environmental changes. S4: The path efficiency index and dynamic adaptation coefficient are fused to form a comprehensive scheduling feature vector, which is then input into the trained intelligent scheduling model to predict the scheduling strategy in combination with task priority and resource status. S5: Based on the output of the scheduling model, the robot's movement path and work density are dynamically adjusted, and the scheduling parameters are iteratively optimized through a closed-loop feedback mechanism to achieve efficient collaborative control of the entire car wash task.

[0006] As a further aspect of the present invention: the construction process of the weighted graph model is as follows: By integrating sensor networks to acquire vehicle location and obstacle distribution information in real time, and using geographic information system technology to analyze path accessibility, the location of all vehicles is accurately identified and located. Each vehicle position, obstacle position, and predetermined task point is defined as a node in the graph; an edge is established between any two directly reachable nodes, and the weight of the edge is calculated according to a preset standard, including the straight-line distance between the two points, the expected travel time, and the estimated energy consumption. A dynamic weight adjustment algorithm is adopted to dynamically adjust the weight value of each edge based on real-time traffic conditions, weather conditions, and the robot's current battery status to reflect changes in the actual operating environment. At the same time, different weight adjustment strategies are set to prioritize time efficiency, and a weighted graph model is constructed.

[0007] As a further aspect of the present invention: the preliminary optimization of the weighted graph using the minimum spanning tree algorithm specifically includes: Real-time monitoring of environmental changes, including obstacle movement and the location of new vehicles, and immediate updating of the weights of relevant nodes and edges when changes are detected; All candidate edges are stored using a priority queue, with priority set according to the edge weight, ensuring that the current best edge is selected in each iteration to reduce computation. For the partially constructed minimum spanning tree, only the newly added nodes and their adjacent nodes are incrementally updated, rather than recalculating the minimum spanning tree of the entire graph, thereby reducing computational complexity and speeding up processing.

[0008] As a further aspect of the present invention: the process for obtaining the path efficiency index is as follows: Define a set of key quantitative indicators that affect path efficiency, including total path length, expected travel time, energy consumption, and obstacle detour frequency. Each indicator is calculated using the attribute values ​​of the corresponding nodes and edges in the weighted graph. Design a fuzzy logic system containing multiple input and output variables. Each input variable is mapped to a fuzzy set and assigned a corresponding membership function to reflect its impact on path efficiency. Based on historical data and real-time feedback, an adaptive mechanism is used to dynamically adjust the weight coefficients of each quantitative indicator to ensure that the actual efficiency of the path can be accurately reflected under different conditions. By using a pre-trained neural network model, the quantified indicators processed by fuzzy logic are integrated into a single path efficiency index.

[0009] As a further aspect of the present invention: the integration of quantified indicators processed by fuzzy logic into a single path efficiency index through a pre-trained neural network model specifically includes: A neural network model for path efficiency evaluation is built and trained based on historical data. The model input is a multi-dimensional quantitative index processed by fuzzy logic, and the output is a single value representing the overall scheduling efficiency of the path. Historical route planning datasets are collected, which include indicators such as total route length, travel time, energy consumption, and detour frequency. The corresponding route efficiency scores are obtained by combining system feedback. The data is then cleaned, missing values ​​are imputed, and normalized to form standardized training samples. A neural network model with a multilayer perceptron structure is constructed. The model includes an input layer, at least two hidden layers, and an output layer. The input layer receives a preprocessed feature vector, the hidden layers use the ReLU activation function for nonlinear mapping, and the output layer outputs the path efficiency index. The mean squared error loss function is used as the optimization objective, the model parameters are adjusted through the backpropagation algorithm, and a K-fold cross-validation mechanism is introduced during the training process to improve the model's generalization ability. By applying a model ensemble strategy, the prediction results of multiple neural network models are fused using the Bagging method, and the learning rate, batch size, and number of hidden layer nodes are optimized through hyperparameter tuning techniques to improve path efficiency and prediction accuracy. The trained neural network is deployed into the scheduling and control system. During task execution, environmental status and task requirements information are input in real time to dynamically generate the path efficiency index. Execution performance data is continuously collected through a closed-loop feedback mechanism for online model updates and iterative optimization.

[0010] As a further aspect of the present invention: based on the subgraph after the minimum spanning tree selection, the shortest path algorithm is applied to determine the locally optimal path between the starting point and the target point, specifically including: Based on the subgraph structure optimized by the minimum spanning tree, the starting node and target node of the current task are identified, and the path search range is set. Dijkstra's algorithm is used to calculate the local optimal path between two points while ensuring path connectivity. The path selection criteria include at least one of path length, estimated travel time and energy consumption. A set of candidate paths is generated by combining the path planning results, and a topology analysis is performed on each path to mark key intermediate nodes and potential conflict areas, providing basic data support for subsequent dynamic adaptation coefficient calculation; By analyzing historical execution records and real-time environmental feedback, the stability and reachability of each candidate path are evaluated, and several feasible paths suitable for the current state are preliminarily selected.

[0011] As a further aspect of the present invention: the process of obtaining the dynamic adaptation coefficient is as follows: Collect dynamic information about the current environment, including but not limited to changes in traffic flow, updates to the location of temporary obstacles, weather influencing factors, and the robot's own state parameters; The current environmental dynamic information is mapped to the key nodes and edges of the candidate path to construct a dynamic weight adjustment model, which is used to simulate the performance of the path under different environmental disturbances. Define a function for calculating the dynamic fitness coefficient. This function takes the stability, recoverability, and replanning cost of the path in a dynamic environment as input variables and outputs a comprehensive score that reflects the robustness of the path. Candidate paths are ranked based on dynamic adaptation coefficients and combined with path efficiency indices to form a multi-dimensional evaluation system, providing decision support for subsequent path optimization and scheduling strategy adjustments.

[0012] As a further aspect of the present invention: the scheduling strategy prediction based on task priority and resource status specifically includes: The path efficiency index and dynamic adaptation coefficient are standardized to have the same dimensions and range, and then merged into a comprehensive scheduling feature vector. A deep learning-based intelligent scheduling model is constructed. The model includes an input layer, multiple hidden layers, and an output layer. The input layer receives the comprehensive scheduling feature vector as input. The hidden layers use the ReLU activation function to enhance the nonlinear expression capability. The output layer outputs the predicted scheduling strategy. The intelligent scheduling model is pre-trained using a large amount of historical scheduling data. The mean squared error loss function is used to measure the difference between the model's predicted value and the actual value. The model parameters are continuously adjusted until convergence is achieved through the backpropagation algorithm. Define task priority evaluation criteria, including but not limited to task urgency, customer satisfaction influencing factors, and economic benefit evaluation indicators, and quantify and integrate these factors into the comprehensive scheduling feature vector; The system monitors the resource status of each robot in real time, including power level, workload and maintenance requirements, and incorporates these as additional input features into the intelligent scheduling model to ensure that the scheduling strategy fully considers resource availability and allocation rationality. Based on comprehensive scheduling feature vectors and real-time resource status information, the intelligent scheduling model outputs specific scheduling strategies for each task, including robot selection, path planning, and execution order arrangement. Establish a closed-loop feedback mechanism to continuously monitor the execution effect of the scheduling strategy, and dynamically adjust the model parameters and scheduling strategy based on the feedback data to ensure the efficient operation and continuous optimization of the system.

[0013] As a further aspect of the present invention: S5 further includes: The scheduling strategy output by the scheduling model is analyzed, and the robot movement path, priority order, and task density information corresponding to each task are extracted. Based on the task density information, a detailed execution plan is generated to ensure that each robot operates according to the optimal path and workload. Real-time monitoring of the operating status and environmental changes of each robot, including but not limited to the robot's current location, battery level, obstacle location, and traffic flow; This data is collected using sensor networks and communication modules and then transmitted to the central control system for analysis and processing. Establish a closed-loop feedback mechanism to compare and analyze the actual operation data with the expected scheduling results. For deviations from expectations, including path congestion and robot malfunctions, the system automatically generates adjustment instructions to dynamically modify the robot's movement path and work density to minimize the impact on the overall task. Reinforcement learning algorithms are used to continuously optimize scheduling parameters. Based on historical data and real-time feedback information, a reinforcement learning model is trained to identify the best scheduling strategy. By continuously iterating and updating the model parameters, the adaptability and response speed of the system are improved, ensuring efficient and coordinated control of the entire car wash task.

[0014] A distributed intelligent scheduling and decision-making control system for car wash robots includes: A weighted graph construction module is used to construct a weighted graph model that includes vehicle location, obstacle distribution, and path reachability, and to set the weights of the edges based on path length, energy consumption, and time; wherein, the vehicle location is defined as a node; The preliminary optimization module uses the minimum spanning tree algorithm to perform preliminary optimization on the weighted graph, identifies the minimum total weight path structure covering all nodes, and calculates the path efficiency index to characterize the overall scheduling efficiency of each candidate path. The local optimal evaluation module, based on the subgraph after minimum spanning tree screening, applies the shortest path algorithm to determine the local optimal path between the starting point and the target point, and calculates the dynamic adaptation coefficient in combination with dynamic environment information to reflect the path's responsiveness to environmental changes. The intelligent scheduling model application module integrates the path efficiency index and the dynamic adaptation coefficient to form a comprehensive scheduling feature vector, which is then input into the trained intelligent scheduling model to predict scheduling strategies based on task priority and resource status. The dynamic adjustment and closed-loop feedback optimization module dynamically adjusts the robot's movement path and work density based on the output results of the scheduling model, and iteratively optimizes the scheduling parameters through a closed-loop feedback mechanism to achieve efficient collaborative control of the entire car wash task.

[0015] The beneficial effects of this invention are: (1) This invention provides a structured and quantifiable environmental representation basis for path planning and scheduling decisions of a car wash robot system by constructing a weighted graph model that includes vehicle location, obstacle distribution, and path reachability. Based on this, an improved minimum spanning tree (MST) algorithm is introduced to initially optimize the global path structure, identifying the minimum total weight path skeleton covering all task nodes, thereby effectively reducing the space complexity of subsequent path searches. Subsequently, a shortest path algorithm (such as Dijkstra's algorithm) is further applied to this optimized subgraph, combining task start and target point information to accurately calculate the locally optimal path that satisfies multi-objective constraints (such as time, energy consumption, path length, etc.). Furthermore, this invention innovatively introduces a dynamic weight adjustment mechanism, mapping real-time environmental factors (such as traffic flow changes, weather effects, temporary obstacle appearances, robot state fluctuations, etc.) to nodes and edges in the graph model, realizing online updating and adaptive adjustment of path weights. This path planning strategy, which integrates static modeling and dynamic perception, not only accurately reflects the current environmental state but also predicts the impact of potential disturbances on path execution, significantly improving the robustness and foresight of path planning. The combined application of the aforementioned technologies enables car wash robots to quickly respond to complex and ever-changing task scenarios and select an optimal path that balances efficiency, energy consumption, and safety, significantly improving the overall scheduling efficiency, resource utilization, and service response speed of the system. Furthermore, by incorporating multi-dimensional evaluation indicators such as path efficiency index and dynamic adaptation coefficient into a unified scheduling decision framework, the system's intelligent judgment and autonomous decision-making capabilities in uncertain environments are further enhanced, providing solid technical support for achieving efficient, stable, and sustainable distributed collaborative operation of car wash robots.

[0016] (2) This invention innovatively introduces a deep learning-based intelligent scheduling model and reinforcement learning algorithm to construct a distributed car wash robot collaborative scheduling system with high adaptability and continuous evolution capability. This system integrates multi-dimensional evaluation indicators such as path efficiency index and dynamic adaptation coefficient into a unified comprehensive scheduling feature vector, which serves as the core input feature of the intelligent scheduling model. This enables comprehensive perception and quantitative modeling of path performance, resource status, and environmental changes in complex task scenarios. In its specific implementation, the deep learning scheduling model adopts a multi-layer perceptron structure, combined with the ReLU nonlinear activation function and mean squared error loss function. This effectively captures the complex nonlinear relationship between task priority, path attributes, and robot resource status, and continuously optimizes model parameters through backpropagation algorithm to improve prediction accuracy. Simultaneously, the system introduces a K-fold cross-validation mechanism and model ensemble strategy (such as Bagging) to further enhance the model's generalization ability and robustness, ensuring stable and efficient scheduling performance under different task densities and environmental disturbances. Based on this, this invention further integrates reinforcement learning mechanisms to establish a multi-objective optimization framework with task completion efficiency, energy consumption control, and user satisfaction as objective functions. By defining a multi-dimensional state space, action space, and reward function that includes environmental conditions, resource distribution, and task requirements, the system can autonomously explore and iteratively update the optimal scheduling strategy, driven by both historical data and real-time feedback. This combination of "online learning + offline training" enables the system not only to make rapid responses and optimization decisions based on the current environmental conditions, but also to continuously accumulate knowledge from long-term operational experience, gradually forming a forward-looking scheduling logic, significantly improving the system's intelligence level and autonomous decision-making capabilities. Attached Figure Description

[0017] The invention will now be further described with reference to the accompanying drawings.

[0018] Figure 1 This is a flowchart of a distributed car wash robot intelligent scheduling decision control method according to the present invention; Figure 2 This is a flowchart of a distributed intelligent scheduling and decision-making control system for car wash robots according to the present invention. Detailed Implementation

[0019] 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.

[0020] Please see Figure 1As shown, this invention is an intelligent scheduling and decision-making control method for distributed car wash robots, comprising the following steps: S1: Construct a weighted graph model that includes vehicle location, obstacle distribution, and path reachability, and set the weights of the edges according to path length, energy consumption, and time; wherein, the vehicle location is defined as a node; S2: The minimum spanning tree algorithm is used to perform preliminary optimization of the weighted graph, identify the path structure with the minimum total weight that covers all nodes, and calculate the path efficiency index to characterize the overall scheduling efficiency of each candidate path. S3: Based on the subgraph after the minimum spanning tree selection, the shortest path algorithm is applied to determine the local optimal path between the starting point and the target point, and the dynamic adaptation coefficient is calculated in combination with dynamic environment information to reflect the path's responsiveness to environmental changes. S4: The path efficiency index and dynamic adaptation coefficient are fused to form a comprehensive scheduling feature vector, which is then input into the trained intelligent scheduling model to predict the scheduling strategy in combination with task priority and resource status. S5: Based on the output of the scheduling model, the robot's movement path and work density are dynamically adjusted, and the scheduling parameters are iteratively optimized through a closed-loop feedback mechanism to achieve efficient collaborative control of the entire car wash task.

[0021] In S1, a weighted graph model is constructed that includes vehicle location, obstacle distribution, and path reachability, and the weights of the edges are set according to path length, energy consumption, and time; wherein, the vehicle location is defined as a node, specifically including: Real-time vehicle location and obstacle distribution information is acquired by integrating an advanced sensor network. This process utilizes high-precision positioning technologies (such as GPS / BeiDou) combined with a Geographic Information System (GIS) to accurately identify and locate all vehicles. This data not only provides static environmental information but also reflects dynamic changes, such as the positional shifts of other moving objects.

[0022] Each vehicle location, obstacle location, and predetermined task point is defined as a node in the graph. Edges are established between any two directly reachable nodes, representing possible travel paths from one node to another. Edge weights are calculated based on preset criteria, including but not limited to the straight-line distance between the two points, expected travel time, and energy consumption estimates. To accurately reflect changes in the actual operating environment, a dynamic weight adjustment algorithm is employed. This algorithm considers factors such as real-time traffic conditions, weather conditions, and the robot's current battery status, dynamically adjusting the weight value of each edge to ensure the model can adapt to constantly changing real-world conditions.

[0023] Taking time efficiency as a priority, we designed a specific weighting strategy. When faced with high-density or urgent tasks, the system automatically increases the proportion of time-related weights, making the optimized route more inclined to reduce travel time rather than energy consumption or other factors. This strategy helps improve the overall system's operational efficiency and service quality while meeting user needs.

[0024] To ensure the weighted graph model is fully disclosed without revealing key technical details, we have maintained a certain degree of abstraction in describing the specific parameter settings and algorithm implementation. Specific parameter configurations and algorithm details will be detailed in the patent application documents and protected through appropriate confidentiality measures to ensure that the technical advantages of this invention cannot be easily copied or imitated.

[0025] The weighted graph model construction method provided by this invention not only achieves accurate modeling of vehicle position, obstacle distribution, and path accessibility, but also enhances the system's ability to cope with complex environmental changes through a dynamic weight adjustment mechanism, thus providing an efficient and reliable navigation foundation for unmanned collaborative car wash robots. This innovative solution is expected to significantly improve several shortcomings of traditional car wash models and promote the intelligent upgrading and sustainable development of the car wash industry.

[0026] In S2, the minimum spanning tree algorithm is used to initially optimize the weighted graph, identify the path structure with the minimum total weight covering all nodes, and calculate the path efficiency index to characterize the overall scheduling efficiency of each candidate path. Specifically, this includes: First, an improved Minimum Spanning Tree (MST) algorithm is used to initially optimize the weighted graph. This algorithm monitors environmental changes in real time, including but not limited to dynamic information such as obstacle movement and the location of newly added vehicles. When the system detects a change in the state of a node or edge, it immediately triggers a weight update mechanism to ensure that the graph model accurately reflects the current environmental state. Subsequently, the system uses a priority queue structure to store all candidate edges, where the priority is set according to the weight of each edge, to ensure that the current optimal edge is selected in each iteration, thereby effectively reducing unnecessary redundant calculations and improving overall processing efficiency.

[0027] To reduce computational complexity and accelerate response time, the system employs an incremental update strategy for the partially constructed minimum spanning tree structure. This means that only newly added nodes and their adjacent nodes are locally reconstructed, rather than rerunning the MST algorithm on the entire graph. This approach significantly reduces computational resource consumption, improves the system's real-time response capability in dynamic environments, and is suitable for rapid path optimization requirements in multi-robot collaborative operation scenarios.

[0028] After identifying the minimum spanning tree structure, the system proceeds to the path efficiency index calculation stage. This index quantifies the overall scheduling efficiency of different path schemes. To this end, the system defines a set of key quantitative indicators affecting path efficiency, including total path length, expected travel time, energy consumption, and detour frequency. Each indicator is calculated based on the attribute values ​​of the corresponding nodes and edges in the weighted graph, forming a multi-dimensional feature input.

[0029] To more accurately integrate these metrics, a fuzzy logic evaluation system was designed. This system includes multiple input variables (such as path length, time, and energy consumption) and one output variable (path efficiency index). Each input variable is mapped to a corresponding fuzzy set and assigned a membership function to characterize its impact on overall path efficiency. Furthermore, the system introduces an adaptive weight adjustment mechanism, dynamically adjusting the weight coefficients of each metric based on historical task data and real-time feedback information, thereby ensuring accurate efficiency evaluation results under different task conditions and environmental states.

[0030] The system further introduces a pre-trained neural network model to fuse multi-dimensional quantitative indicators processed by fuzzy logic into a single path efficiency index. This model is trained on a large amount of historical route planning data, including key indicators such as total route length, travel time, energy consumption, and detour frequency, and uses the path efficiency score obtained from system feedback as the target label. In the data preprocessing stage, the system cleans, imputes missing values, and normalizes the raw data to generate standardized training samples.

[0031] The neural network model employs a multilayer perceptron (MLP) structure, including an input layer, at least two hidden layers, and an output layer. The input layer receives feature vectors processed by fuzzy logic, the hidden layers use the ReLU activation function to enhance nonlinear expressiveness, and the output layer outputs the path efficiency index. During model training, a mean squared error loss function is used, and the model parameters are continuously optimized through backpropagation. Simultaneously, a K-fold cross-validation mechanism is introduced to ensure the model has good generalization ability and prediction stability.

[0032] To further improve prediction accuracy, the system also employs a model ensemble strategy, using the Bagging method to fuse the prediction results of multiple trained neural network models. Furthermore, hyperparameter tuning techniques are used to optimize key parameters such as learning rate, batch size, and the number of hidden layer nodes to achieve optimal performance.

[0033] The trained neural network model is deployed in the scheduling and control system. During actual task execution, the system inputs real-time environmental status and task requirement information into the model to dynamically generate a path efficiency index. Simultaneously, the system establishes a closed-loop feedback mechanism to continuously collect path execution performance data for online model updates and iterative optimization, thereby continuously improving the accuracy of path evaluation and the system's intelligence level.

[0034] This step, by introducing an improved minimum spanning tree algorithm and a path efficiency evaluation mechanism that integrates fuzzy logic and neural networks, achieves efficient modeling and dynamic evaluation of path scheduling efficiency in complex environments. This provides a solid data foundation and technical support for subsequent path optimization and scheduling strategy generation. This technical solution possesses high novelty and practicality, fully meeting the needs of distributed car wash robot systems for intelligent path scheduling decision-making in dynamic environments.

[0035] In S3, based on the subgraph after minimum spanning tree filtering, the shortest path algorithm is applied to determine the locally optimal path between the starting point and the target point. Furthermore, dynamic environmental information is used to calculate the dynamic fitness coefficient, reflecting the path's responsiveness to environmental changes. Specifically, this includes: Based on the subgraph obtained after minimum spanning tree selection, the system identifies the starting and target nodes according to the current task requirements and sets a reasonable path search range. This process considers the feasibility and efficiency in practical operation, avoiding unnecessary large-scale searches. Next, Dijkstra's algorithm is used for path search to calculate the locally optimal path between two points. When selecting a path, in addition to the basic path length, factors such as estimated travel time and energy consumption are also considered to ensure that the path is both economical and efficient. For each possible candidate path, the system performs topology analysis, marking key intermediate nodes and potential conflict areas. This analysis provides the necessary basic data support for the subsequent calculation of the dynamic fitness coefficient.

[0036] To evaluate the stability and reachability of each candidate route, the system analyzes historical execution records and real-time environmental feedback. This information includes, but is not limited to, past route usage and real-time traffic flow changes, used to initially filter out several feasible routes suitable for the current conditions. In this process, the system not only focuses on the static attributes of the routes but also emphasizes their performance under different environmental conditions, ensuring that the selected routes can adapt to potential future changes.

[0037] For each selected candidate path, the system begins calculating dynamic adaptation coefficients. This step first involves collecting dynamic information about the current environment, such as changes in traffic flow, updates to the locations of temporary obstacles, weather influences, and the robot's own state parameters. Subsequently, this environmental dynamic information is mapped to the key nodes and edges of the candidate path, constructing a dynamic weight adjustment model. This model simulates changes in path performance under various environmental disturbances, thereby helping to predict the actual operational performance of the path.

[0038] The dynamic adaptation coefficient is defined based on three dimensions: stability, recoverability, and replanning cost of a path in a dynamic environment. Specifically, the system designs a calculation function whose input variables include relevant indicators from these three dimensions, and whose output is a comprehensive score reflecting the path's robustness. This score helps quantify the path's performance in the face of unforeseen circumstances, providing a scientific basis for decision-making.

[0039] All candidate paths are ranked based on a dynamic fitness coefficient, and this ranking is combined with a path efficiency index to form a multi-dimensional evaluation system. This system considers not only the basic performance of the paths but also their adaptability to constantly changing environments. In this way, the system can more comprehensively evaluate each path, providing solid decision support for the final path selection and scheduling strategy adjustment.

[0040] This step introduces the concepts of shortest path algorithm and dynamic adaptation coefficient to achieve effective evaluation and optimization of paths in complex and ever-changing environments. This method fully considers various factors in practical applications, improving the intelligence level and ability to respond to emergencies in distributed car wash robot systems. Simultaneously, it provides valuable reference for other application scenarios requiring path planning and dynamic scheduling.

[0041] In S4, the path efficiency index and dynamic adaptation coefficient are fused to form a comprehensive scheduling feature vector, which is then input into the trained intelligent scheduling model. This vector, combined with task priority and resource status, is used to predict scheduling strategies, specifically including: After obtaining the path efficiency index and dynamic adaptation coefficient, the system first standardizes them to give them uniform dimensions and numerical ranges. This process uses Z-score or Min-Max normalization methods to eliminate dimensional differences between different indicators and improve the accuracy of subsequent fusion calculations. Subsequently, the two standardized indicators are integrated into a comprehensive scheduling feature vector through weighted summation or other feature fusion techniques, which serves as the core input feature of the intelligent scheduling model.

[0042] The system constructs a deep learning-based intelligent scheduling model. This model employs a multilayer perceptron (MLP) structure, comprising an input layer, multiple hidden layers, and an output layer. The input layer receives the comprehensive scheduling feature vectors mentioned above; the hidden layers use the ReLU activation function for nonlinear mapping, enhancing the model's ability to learn complex scheduling problems; and the output layer outputs specific scheduling strategy suggestions, including key decision information such as robot selection, path allocation, and execution order. During model training, the mean squared error (MSE) loss function is used to measure the deviation between predicted values ​​and historical actual scheduling results, and the network parameters are continuously optimized through backpropagation until the model converges. Simultaneously, a K-fold cross-validation mechanism is introduced to improve the model's generalization performance and stability.

[0043] To better align the scheduling strategy with actual business needs, the system defines a task priority evaluation standard. This standard comprehensively considers multiple dimensions, including task urgency, customer satisfaction influencing factors, and economic benefit evaluation indicators. All factors are quantified and integrated into the comprehensive scheduling feature vector. Furthermore, the system monitors the resource status of each car wash robot in real time, including remaining battery power, current workload, and whether it is in a maintenance cycle. This resource status data is added as additional input features to the scheduling model, ensuring that the generated scheduling strategy not only achieves path optimization but also fully considers resource availability and allocation rationality.

[0044] The intelligent scheduling model outputs a specific scheduling strategy for each task based on comprehensive scheduling feature vectors and real-time resource status information. This strategy covers robot assignment, path planning, and task execution sequence arrangement, effectively coordinating operational conflicts among multiple robots and improving the overall system's collaborative efficiency.

[0045] To achieve continuous optimization, the system also establishes a closed-loop feedback mechanism. This mechanism collects key data in real time during the execution of the scheduling strategy, such as task completion time, actual energy consumption of the path, and path congestion, through sensors and communication modules, and uses this feedback data for online model updates and iterative optimization. By periodically retraining the scheduling model, the system can continuously adapt to new task modes and environmental changes, significantly improving its long-term robustness and adaptability.

[0046] This step introduces a comprehensive scheduling feature vector that integrates path efficiency and dynamic adaptation characteristics, and combines it with a deep learning-based intelligent scheduling model to achieve precise scheduling and dynamic response for a distributed car wash robot system. This solution improves task scheduling efficiency while enhancing the system's adaptability to complex task scenarios and resource constraints, demonstrating significant technological advancement and engineering practicality.

[0047] In S5, the robot's movement path and task density are dynamically adjusted based on the output of the scheduling model, and the scheduling parameters are iteratively optimized through a closed-loop feedback mechanism to achieve efficient collaborative control of the entire car wash task. Specifically, this includes: Upon receiving the scheduling strategy output by the intelligent scheduling model, the system first analyzes it, extracting key data such as robot allocation information, movement path planning, task priority order, and work density suggestions for each task. The work density information reflects the washing area or operation frequency that the robot should complete per unit time, guiding the robot to adopt different cleaning intensities in different areas. Based on this information, the system generates a detailed execution plan, including the start time of each robot, travel route, work mode switching nodes, and task termination conditions, ensuring that each robot executes its task according to the optimal path and reasonable work intensity.

[0048] To ensure the stability and safety of task execution, the system monitors the operational status of all robots and changes in the surrounding environment in real time. The collected data includes, but is not limited to: the robot's current location, remaining battery power, operating temperature, obstacle location updates, traffic flow changes, and unexpected events within the car wash area (such as pedestrian intrusion or temporary construction). This information is uploaded in real time to the central control system for unified analysis and processing via a multimodal sensor network (such as LiDAR, cameras, and GPS positioning modules) and communication modules (such as 5G or LoRa wireless transmission units) deployed on the robots.

[0049] The system has established a complete closed-loop feedback mechanism. This mechanism compares and analyzes the task execution data collected during actual operation (such as path completion time, energy consumption, number of task interruptions, etc.) with the ideal results predicted by the scheduling model. When a deviation exceeds a preset threshold (e.g., congestion on a path or a robot malfunction), the system immediately triggers a response mechanism, automatically calculating new alternative paths and reallocating affected tasks. Simultaneously, it adjusts the workload of relevant robots to balance the overall load and minimize the impact on task progress.

[0050] To achieve long-term adaptive optimization of the system, this invention introduces a reinforcement learning algorithm to continuously iteratively update the scheduling parameters. This algorithm trains a reinforcement learning model based on historical task data and real-time feedback data, enabling it to autonomously identify the optimal scheduling strategy under specific environments. During training, the system defines a multi-dimensional state space (including task type, environmental conditions, and resource status), an action space (including path changes, robot reassignment, and job density adjustments), and a reward function (based on indicators such as task completion efficiency, resource utilization, and user satisfaction). The scheduling decision logic is continuously optimized using advanced algorithms such as Deep Q-Network (DQN) or Proximal Policy Optimization (PPO).

[0051] As the number of task executions increases and feedback data accumulates, the scheduling model's decision-making ability continuously improves, and the system gradually acquires the ability to learn from experience and proactively adapt to new scenarios. Ultimately, the system can achieve efficient collaborative control of a distributed car wash robot cluster without human intervention, significantly improving task completion efficiency and service quality.

[0052] This step achieves intelligent control and self-evolution of the entire car wash task by dynamically adjusting the robot path and task density, constructing a closed-loop feedback mechanism, and introducing reinforcement learning to continuously optimize scheduling parameters. This fully demonstrates the technological advancement and engineering practicality of this invention in path scheduling, resource coordination, and system self-adaptation.

[0053] Please see Figure 2 As shown, a distributed intelligent scheduling and decision-making control system for car wash robots includes: A weighted graph construction module is used to construct a weighted graph model that includes vehicle location, obstacle distribution, and path reachability, and to set the weights of the edges based on path length, energy consumption, and time; wherein, the vehicle location is defined as a node; The preliminary optimization module uses the minimum spanning tree algorithm to perform preliminary optimization on the weighted graph, identifies the minimum total weight path structure covering all nodes, and calculates the path efficiency index to characterize the overall scheduling efficiency of each candidate path. The local optimal evaluation module, based on the subgraph after minimum spanning tree screening, applies the shortest path algorithm to determine the local optimal path between the starting point and the target point, and calculates the dynamic adaptation coefficient in combination with dynamic environment information to reflect the path's responsiveness to environmental changes. The intelligent scheduling model application module integrates the path efficiency index and the dynamic adaptation coefficient to form a comprehensive scheduling feature vector, which is then input into the trained intelligent scheduling model to predict scheduling strategies based on task priority and resource status. The dynamic adjustment and closed-loop feedback optimization module dynamically adjusts the robot's movement path and work density based on the output results of the scheduling model, and iteratively optimizes the scheduling parameters through a closed-loop feedback mechanism to achieve efficient collaborative control of the entire car wash task.

[0054] The working principle of this invention: This invention aims to achieve efficient collaborative scheduling and path optimization of multiple car wash robots in complex dynamic environments. The method includes five core steps: 1) Constructing a weighted graph model including vehicle location, obstacle distribution, and path reachability, and setting edge weights based on path length, energy consumption, and time; 2) Initially optimizing the weighted graph using an improved minimum spanning tree algorithm, and calculating a path efficiency index by fusing fuzzy logic and neural networks to evaluate the overall scheduling efficiency of candidate paths; 3) Based on the subgraph after minimum spanning tree selection, applying a shortest path algorithm to determine locally optimal paths, and calculating a dynamic adaptation coefficient based on dynamic environment information to measure the robustness of paths in changing environments; 4) Fusing the path efficiency index and dynamic adaptation coefficient into a comprehensive scheduling feature vector, inputting it into a trained deep learning scheduling model, and predicting the optimal scheduling strategy by combining task priority and robot resource status; 5) Dynamically adjusting robot movement paths and task density based on the scheduling model output, continuously collecting execution data through a closed-loop feedback mechanism, and iteratively optimizing scheduling parameters using reinforcement learning algorithms to improve the system's adaptability and overall collaborative efficiency. This invention significantly improves the scheduling accuracy, response speed, and long-term stability of car wash robot systems in complex scenarios through multi-dimensional indicator fusion, real-time environmental perception, and online model update mechanisms, and has good engineering application prospects and promotion value.

[0055] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A distributed car washing robot intelligent scheduling decision control method, characterized in that, The method comprises the following steps: S1: constructing a weighted graph model comprising vehicle positions, obstacle distribution, and path accessibility, and setting the weight of the edge according to the path length, energy consumption, and time; wherein the vehicle position is defined as a node; S2: using a minimum spanning tree algorithm to preliminarily optimize the weighted graph, identifying the minimum total weight path structure covering all nodes, and calculating the path efficiency index for representing the overall scheduling efficiency of each candidate path; S3: on the basis of the subgraph screened by the minimum spanning tree, applying a shortest path algorithm to determine the locally optimal path between the starting point and the target point, and combining dynamic environmental information to calculate a dynamic adaptation coefficient to reflect the response ability of the path to environmental changes; S4: fusing the path efficiency index and the dynamic adaptation coefficient to form a comprehensive scheduling feature vector, inputting the vector into the trained intelligent scheduling model, and combining the task priority and the resource state to predict the scheduling strategy; S5: dynamically adjusting the robot moving path and the work density according to the output result of the scheduling model, and iteratively optimizing the scheduling parameters through a closed-loop feedback mechanism to realize efficient collaborative control of the entire car washing task.

2. The distributed car washing robot intelligent scheduling decision control method according to claim 1, characterized in that, The construction process of the weighted graph model is as follows: Real-time acquisition of vehicle positions and obstacle distribution information through an integrated sensor network, and analysis of path accessibility using geographic information system technology to ensure that all vehicle positions are accurately identified and located; Defining each vehicle position, obstacle position, and predetermined task point as a node in the graph; for any two directly reachable nodes, an edge is established between them, and the weight of the edge is calculated according to the preset standard, including the straight-line distance between the two points, the expected travel time, and the energy consumption estimate; Using a dynamic weight adjustment algorithm, dynamically adjusting the weight value of each edge based on real-time traffic conditions, weather conditions, and the current power state of the robot to reflect changes in the actual operating environment; at the same time, different weight adjustment strategies are set for priority consideration of time efficiency to construct the weighted graph model.

3. The distributed car washing robot intelligent scheduling decision control method according to claim 1, characterized in that, The preliminary optimization of the weighted graph using the minimum spanning tree algorithm specifically includes: Real-time monitoring of environmental changes, including obstacle movement and newly added vehicle positions, and immediately updating the weights of the relevant nodes and edges when changes are detected; Using a priority queue to store all candidate edges, with the priority set according to the weight of the edge to ensure that the current optimal edge is selected each time to reduce the amount of calculation; For the partially constructed minimum spanning tree, only the newly added nodes and their adjacent nodes are incrementally updated, rather than recalculating the minimum spanning tree of the entire graph, thereby reducing the computational complexity and speeding up the processing.

4. The distributed car washing robot intelligent scheduling decision control method according to claim 1, characterized in that, The path efficiency index acquisition process is as follows: Defining a set of key quantitative indicators that affect path efficiency, including the total path length, expected travel time, energy consumption, and obstacle bypass frequency, each of which is calculated through the attribute values of the corresponding nodes and edges in the weighted graph; Designing a fuzzy logic system containing multiple input variables and output variables, each input variable being mapped to a fuzzy set and being assigned a corresponding membership function to reflect its influence on path efficiency; Based on historical data and real-time feedback, an adaptive mechanism is used to dynamically adjust the weight coefficients of each quantitative indicator, ensuring that the actual efficiency of the path can be accurately reflected under different conditions; Through a pre-trained neural network model, the quantitative indicators processed by fuzzy logic are integrated into a single path efficiency index.

5. The distributed car washing robot intelligent scheduling decision control method according to claim 4, characterized in that, The process of integrating the quantitative indicators processed by fuzzy logic into a single path efficiency index through a pre-trained neural network model specifically includes: Based on historical data, a neural network model for path efficiency evaluation is constructed and trained, where the model input is a multi-dimensional quantitative indicator processed by fuzzy logic, and the output is a single numerical value representing the overall scheduling efficiency of the path; A historical path planning dataset is collected, which includes path length, travel time, energy consumption, and detour frequency indicators, and the corresponding path efficiency scores are obtained through system feedback. The data is cleaned, missing values are filled, and normalized to form standardized training samples; A neural network model with a multi-layer perception structure is constructed, which includes an input layer, at least two hidden layers, and an output layer. The input layer receives the preprocessed feature vector, the hidden layer uses the ReLU activation function for non-linear mapping, and the output layer outputs the path efficiency index; The mean square error loss function is used as the optimization objective, and the model parameters are adjusted through the backpropagation algorithm. The K-fold cross-validation mechanism is introduced during training to improve the model's generalization ability. Model integration strategies are applied, and the Bagging method is used to fuse the prediction results of multiple neural network models. Hyperparameter tuning techniques are used to optimize learning rate, batch size, and hidden layer node number to improve path efficiency prediction accuracy. The trained neural network is deployed in the scheduling control system. During task execution, real-time environmental state and task demand information are input to dynamically generate path efficiency indexes. Through a closed-loop feedback mechanism, execution effect data is continuously collected for online model updating and iterative optimization.

6. The distributed car washing robot intelligent scheduling decision control method according to claim 1, characterized in that, Based on the subgraph filtered by the minimum spanning tree, the shortest path algorithm is used to determine the locally optimal path between the starting point and the target point, specifically including: Based on the subgraph structure optimized by the minimum spanning tree, the starting node and target node of the current task are identified, and the path search range is set; Dijkstra's algorithm is used to calculate the locally optimal path between two points while ensuring path connectivity. The path selection criteria include at least one of path length, estimated travel time, and energy consumption. A candidate path set is generated based on the path planning results, and topological structure analysis is performed on each path to mark key intermediate nodes and potential conflict areas, providing basic data support for subsequent dynamic adaptation coefficient calculation; Through path historical execution records and real-time environmental feedback information, the stability and accessibility of each candidate path are evaluated, and a number of feasible paths suitable for the current state are preliminarily selected.

7. The distributed car washing robot intelligent scheduling decision control method according to claim 1, characterized in that, The process of obtaining the dynamic adaptation coefficient is as follows: Current environmental dynamic information is collected, including but not limited to traffic flow changes, temporary obstacle position updates, weather influence factors, and robot state parameters. The current environmental dynamic information is mapped to the key nodes and edges of the candidate path, a dynamic weight adjustment model is constructed, which is used to simulate the performance of the path under different environmental disturbances; A calculation function of the dynamic adaptation coefficient is defined, which takes the stability, recoverability and re-planning cost of the path in the dynamic environment as input variables, and outputs a comprehensive score reflecting the robustness of the path; According to the dynamic adaptation coefficient, the candidate paths are sorted, and a multi-dimensional evaluation system is formed by combining the path efficiency index, which provides decision support for subsequent path optimization and scheduling strategy adjustment.

8. The distributed car washing robot intelligent scheduling decision control method according to claim 1, characterized in that, The scheduling strategy prediction is combined with the task priority and resource state, specifically including: Standardize the path efficiency index and dynamic adaptation coefficient to have the same dimension and range, and fuse them into a comprehensive scheduling feature vector; An intelligent scheduling model based on deep learning is constructed, which includes an input layer, multiple hidden layers and an output layer. The input layer receives the comprehensive scheduling feature vector as input, the hidden layer uses the ReLU activation function to enhance the non-linear expression ability, and the output layer outputs the predicted scheduling strategy. A large amount of historical scheduling data is used to pre-train the intelligent scheduling model, and the mean square error loss function is used to measure the difference between the predicted value and the actual value of the model, and the model parameters are continuously adjusted until convergence through the back propagation algorithm; Define the task priority evaluation standard, including but not limited to task urgency, customer satisfaction influence factor and economic benefit evaluation index, and quantify and integrate these factors into the comprehensive scheduling feature vector; Real-time monitoring of the resource state of each robot in the system, including power level, workload and maintenance needs, and adding it as an additional input feature to the intelligent scheduling model to ensure that the scheduling strategy fully considers resource availability and rationality of distribution; Based on the comprehensive scheduling feature vector and real-time resource state information, the intelligent scheduling model outputs specific scheduling strategies for each task, including robot selection, path planning and execution order arrangement; A closed-loop feedback mechanism is established to continuously monitor the execution effect of the scheduling strategy and dynamically adjust the model parameters and scheduling strategy according to the feedback data to ensure efficient operation and continuous optimization of the system.

9. The distributed car washing robot intelligent scheduling decision control method according to claim 1, characterized in that, S5 further includes: Analyzing the scheduling strategy output by the scheduling model, extracting the robot moving path, priority order and job density information corresponding to each task; generating a detailed execution plan based on the job density information to ensure that each robot operates according to the optimal path and work intensity; Real-time monitoring of the running state of each robot and environmental changes, including but not limited to the current position of the robot, power level, obstacle position and traffic flow; Collecting these data using a sensor network and communication module, and transmitting them to the central control system for analysis and processing; A closed-loop feedback mechanism is established to continuously monitor the execution effect of the scheduling strategy and dynamically adjust the model parameters and scheduling strategy according to the feedback data to ensure efficient operation and continuous optimization of the system. The reinforcement learning algorithm is used to continuously optimize the scheduling parameters; based on historical data and real-time feedback information, the reinforcement learning model is trained to identify the best scheduling strategy, and through continuous iteration of model parameters, the adaptability and response speed of the system are improved, ensuring efficient collaborative control of the entire car washing task.

10. A distributed car washing robot intelligent scheduling decision control system, characterized in that, The distributed car washing robot intelligent scheduling decision control method according to any one of claims 1-9, comprising: A weighted graph construction module is used to construct a weighted graph model containing vehicle position, obstacle distribution and path accessibility, and set the edge weight according to path length, energy consumption and time; wherein the vehicle position is defined as a node; A preliminary optimization module uses a minimum spanning tree algorithm to preliminarily optimize the weighted graph, identifies the minimum total weight path structure covering all nodes, calculates the path efficiency index, and uses it to represent the overall scheduling efficiency of each candidate path; A local optimal evaluation module determines the local optimal path between the starting point and the target point based on the subgraph after minimum spanning tree screening, and calculates the dynamic adaptation coefficient combined with dynamic environment information to reflect the response ability of the path to environmental changes; An intelligent scheduling model application module fuses the path efficiency index and the dynamic adaptation coefficient to form a comprehensive scheduling feature vector, which is input into the trained intelligent scheduling model, and the scheduling strategy is predicted combined with task priority and resource state; A dynamic adjustment and closed-loop feedback optimization module dynamically adjusts the robot movement path and work density according to the scheduling model output result, and iteratively optimizes the scheduling parameters through a closed-loop feedback mechanism to achieve efficient collaborative control of the entire car washing task.

Images