A multi-role-oriented intelligent weeding operation management and control system and method

Abstract

This invention discloses a weeding operation control method and system based on path planning and emergency decision-making collaboration. The method acquires information on the route to be operated, the window time, the vehicle's initial battery level, and the initial quantity of working medium. An optimal path is generated through a multi-objective optimization algorithm, and simultaneously, predicted time series, predicted energy consumption series, and predicted working medium consumption series for each node position on the optimal path are generated. The globally optimal path and predicted data are loaded into the vehicle controller. During operation, the actual vehicle status data is collected in real time, and the deviation between the vehicle's instantaneous battery level and the corresponding node position in the predicted energy consumption series, the working medium remaining quantity deviation, and the time series deviation between the actual cumulative time and the corresponding node position in the predicted time series are calculated. If the data exceeds a corresponding preset safety threshold, a reinforcement learning decision model is activated to generate an emergency correction path in real time, and the vehicle is controlled to travel according to the emergency correction path.

Description

Technical Field

[0001] This invention belongs to the field of railway weeding technology, and in particular relates to a weeding operation control method and system based on path planning and emergency decision-making coordination. Background Technology

[0002] As the core of the modern transportation system, the operational safety of high-speed rail is closely related to the condition of the track. Weed growth along the tracks not only obstructs the ballast and affects drainage, but can also breed animals and interfere with signaling equipment, posing a serious threat to train safety. Therefore, regular and efficient weeding is a crucial aspect of high-speed rail maintenance. Traditional track weeding relies mainly on manual or semi-mechanized methods. Manual weeding is inefficient, labor-intensive, and requires workers to be exposed to the outdoor environment for extended periods, especially during maintenance windows, posing significant safety risks. While semi-mechanized equipment improves efficiency to some extent, it still requires close-range manual operation, has a low degree of automation, and cannot meet the large-scale, high-standard maintenance requirements of the high-speed rail network. Summary of the Invention

[0003] To address the aforementioned issues, the present invention aims to provide a weeding operation control method and system based on path planning and emergency decision-making collaboration, thereby enhancing the system's adaptability to unexpected operating conditions (such as sudden increases in energy consumption or equipment failure).

[0004] The technical solution provided by this invention is: a weeding operation control method based on path planning and emergency decision-making coordination, comprising the following steps: S1: Obtain the route information to be operated, the sunroof time, the initial battery power of the vehicle and the initial amount of working medium, generate a globally optimal path through a multi-objective optimization algorithm, and simultaneously generate the predicted time series, predicted energy consumption series and predicted working medium consumption series of each node position on the optimal path, and load the globally optimal path and its predicted data into the weeding vehicle controller. S2: During the operation of the weeding vehicle following the global optimal path, the actual status data of the vehicle is collected in real time, and the deviation ΔE between the vehicle's instantaneous power and the corresponding node position in the predicted energy consumption sequence, the deviation ΔL of the working medium remaining amount, and the time sequence deviation ΔT between the actual cumulative time and the corresponding node position in the predicted time sequence are calculated. S3: Determine whether the power deviation ΔE, working medium balance deviation ΔL, and timing deviation ΔT exceed the corresponding preset safety thresholds; if so, start the reinforcement learning decision model. S4: The reinforcement learning decision model generates an emergency correction path in real time based on the vehicle's current node position, remaining battery power, remaining working medium, and remaining sunroof time, and controls the vehicle to travel according to the emergency correction path.

[0005] Preferably, step S1, which generates a globally optimal path using a multi-objective optimization algorithm and simultaneously generates the predicted time series, predicted energy consumption series, and predicted working medium consumption series for each node position on the optimal path, further includes: The line to be operated is discretized into multiple track segments, and a complete operation path is encoded as a chromosome consisting of a sequence of track segment IDs; A fitness function is constructed with total energy consumption, total operating time, and total weeding coverage as objectives, and includes constraints on electricity consumption and operating medium consumption. The fitness function introduces density weights in the coverage calculation, so that individuals on paths covering high-density areas obtain higher fitness values, thereby driving the population to evolve in that direction. The chromosomes in the population are iteratively evolved through selection, crossover and mutation operations until the termination condition is met, and the chromosome with the highest fitness is output as the global optimal path. A physical model is generated for the predicted time series, predicted energy consumption series, and predicted working medium consumption series of each node location on the optimal path, including: Predicting time series ; Predicted energy consumption sequence ; Predicting the consumption sequence of working media ; Among them, L i G i C i D i These represent the length, slope, curvature, and weed density of the track segment along the path, respectively. i k is determined by the globally optimal path generated by the genetic algorithm, which is used to plan the speed. drive k medium denoted as energy consumption coefficient, and f, g, and h are preset physical relationship functions.

[0006] Preferably, in step S2, the current node location of the vehicle is determined by the Beidou / GPS positioning module integrated in the vehicle, and the predicted value corresponding to the node location is indexed from the predicted time series, predicted energy consumption series and predicted working medium consumption series, and used to calculate the time series deviation ΔT, power deviation ΔE and working medium balance deviation ΔL.

[0007] Preferably, the safety threshold in step S3 is dynamically adjusted according to a nonlinear decay function, as shown in the formula: ; in, The standard threshold at the start of the task. This is the minimum threshold before the task ends. Total skylight time, The remaining skylight time is n, where n is a sensitivity adjustment coefficient greater than 1. The sensitivity adjustment coefficient n is used to control the rate at which the threshold decays over time, so that the system maintains a high tolerance in the early stage of the operation, while the sensitivity to deviation increases sharply in the late stage of the operation.

[0008] Preferably, the reward function of the reinforcement learning decision model described in step S4 is: in, For safety rewards, a positive reward is given for safely evacuating or reaching a safe point before the end of the window, and a negative reward is given otherwise. The reward for completing the task is proportional to the effective working area. A basic efficiency bonus is set in proportion to the amount of work done per unit of time and the amount of work done per unit of energy consumption. The deviation penalty term is associated with the planning results of step S1. , where α, β, γ are penalty coefficients.

[0009] Preferably, the state space of the reinforcement learning decision model includes: The vehicle's current status information, remaining sunroof time, and the battery deviation ΔE, working medium remaining deviation ΔL, and timing deviation ΔT calculated by step S2; The action space of the reinforcement learning decision model is a hybrid action space, including: Discrete actions: {Continue operation as planned, suspend operation, execute return to base, proceed to the nearest safe point}; Continuous actions: {target driving speed, target spray flow rate}.

[0010] Preferably, after step S4, the method further includes: Record complete data for this operation, including actual path, actual energy consumption, actual operating medium consumption, actual operation time, and all triggered emergency correction paths; Based on the recorded data, the physical model parameters of the multi-objective optimization algorithm in step S1 and / or the policy network of the reinforcement learning decision model in step S4 are iteratively optimized and retrained, so that the system can continuously evolve in subsequent tasks.

[0011] Based on the same concept, the present invention also provides a weeding operation control system based on path planning and emergency decision-making coordination, comprising: The initial planning module is used to obtain the route information to be operated, the sunroof time, the initial battery power of the vehicle and the initial amount of working medium. It generates a globally optimal path through a multi-objective optimization algorithm, and simultaneously generates the predicted time series, predicted energy consumption series and predicted working medium consumption series of each node position on the optimal path. The globally optimal path and its predicted data are loaded into the weeding vehicle controller. The data acquisition and calculation module is used to collect the actual status data of the vehicle in real time during the operation of the weeding vehicle traveling along the global optimal path, and to calculate the deviation ΔE between the vehicle's instantaneous power and the corresponding node position in the predicted energy consumption sequence, the deviation ΔL of the working medium remaining amount, and the time sequence deviation ΔT between the actual cumulative time and the corresponding node position in the predicted time sequence. The monitoring module is used to determine whether the power deviation ΔE, the working medium balance deviation ΔL, and the timing deviation ΔT exceed the corresponding preset safety thresholds; if so, the reinforcement learning decision model is activated. The emergency decision-making module is used by the reinforcement learning decision-making model to generate an emergency correction path in real time based on the vehicle's current node position, remaining battery power, remaining working medium, and remaining sunroof time, and to control the vehicle to travel according to the emergency correction path.

[0012] Based on the same concept, the present invention also provides an electronic device, comprising: The memory is used to store the processing program; A processor, which, when executing the processing program, implements the weeding operation control method based on path planning and emergency decision-making coordination as described above.

[0013] Based on the same concept, the present invention also provides a readable storage medium storing a processing program, which, when executed by a processor, implements the weeding operation control method based on path planning and emergency decision coordination described above.

[0014] This invention, by employing the above technical solutions, possesses the following advantages and positive effects compared to existing technologies: The technical solution of this invention comprehensively considers time, energy consumption, and operational medium consumption through a multi-objective optimization algorithm to generate a globally optimal path and predicted sequence. Through real-time monitoring and dynamic deviation calculation, the system can accurately perceive the deviation between the plan and the actual situation. Once the deviation exceeds the dynamic safety threshold, a reinforcement learning model is activated for intelligent emergency decision-making. The reinforcement learning model can generate the optimal corrected path in real time based on the current state, rather than simply interrupting or returning to the starting point, thus improving the system's adaptability to unexpected operating conditions (such as sudden increases in energy consumption or equipment failure). By closely integrating "macro-optimal planning" with "micro-intelligent response," the problem of high risk of operational interruption under complex constraints is effectively solved, ensuring that weeding operations are completed safely, efficiently, and reliably within the strict window of opportunity. Attached Figure Description

[0015] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings, wherein: Figure 1 This is a flowchart of the weeding operation control method based on path planning and emergency decision-making coordination of the present invention; Figure 2 This is a schematic diagram of the structure of the weeding vehicle of the present invention.

[0016] Explanation of reference numerals in the attached diagram: 100-Weeding vehicle; 1-Vehicle control system; 2-Sprayer truck; 3-Storage tank; 4-Sprayer head. Detailed Implementation

[0017] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. The advantages and features of the present invention will become clearer from the following description and claims. It should be noted that the drawings are all in a very simplified form and use non-precise ratios, and are only used to facilitate and clarify the illustration of the embodiments of the present invention.

[0018] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0019] First Embodiment See Figure 1 This embodiment provides a weeding operation control method based on path planning and emergency decision-making coordination, including the following steps: S1: Obtain the route information to be operated, the sunroof time, the initial battery power of the vehicle and the initial amount of working medium, generate a globally optimal path through a multi-objective optimization algorithm, and simultaneously generate the predicted time series, predicted energy consumption series and predicted working medium consumption series of each node position on the optimal path, and load the globally optimal path and its predicted data into the weeding vehicle controller. S2: During the operation of the weeding vehicle following the global optimal path, the actual status data of the vehicle is collected in real time, and the deviation ΔE between the vehicle's instantaneous power and the corresponding node position in the predicted energy consumption sequence, the deviation ΔL of the working medium remaining amount, and the time sequence deviation ΔT between the actual cumulative time and the corresponding node position in the predicted time sequence are calculated. S3: Determine whether the power deviation ΔE, working medium balance deviation ΔL, and timing deviation ΔT exceed the corresponding preset safety thresholds; if so, start the reinforcement learning decision model. S4: The reinforcement learning decision model generates an emergency correction path in real time based on the vehicle's current node position, remaining battery power, remaining working medium, and remaining sunroof time, and controls the vehicle to travel according to the emergency correction path.

[0020] The technical solution in this embodiment comprehensively considers time, energy consumption, and working medium consumption through a multi-objective optimization algorithm to generate a globally optimal path and predicted sequence. Through real-time monitoring and dynamic deviation calculation, the system can accurately perceive the deviation between the plan and the actual situation. Once the deviation exceeds the dynamic safety threshold, a reinforcement learning model is activated for intelligent emergency decision-making. The reinforcement learning model can generate the optimal corrected path in real time based on the current state, rather than simply interrupting or returning to the starting point, thus improving the system's adaptability to unexpected working conditions (such as sudden increases in energy consumption or equipment failure). By closely combining "macro-optimal planning" with "micro-intelligent response," the problem of high risk of work interruption under complex constraints is effectively solved, ensuring that weeding operations are completed safely, efficiently, and reliably within the strict time window.

[0021] See Figure 2 The weeding vehicle 100 for implementing the control method of the present invention includes a tractor carrying the vehicle control system 1 and a spraying vehicle 2. The spraying vehicle 2 is equipped with a liquid storage tank 3 and a nozzle 4. The liquid storage tank 3 is used to store the working medium, and the nozzle 4 is used to spray the working medium to a preset range.

[0022] Preferably, step S1, which generates a globally optimal path using a multi-objective optimization algorithm and simultaneously generates the predicted time series, predicted energy consumption series, and predicted working medium consumption series for each node position on the optimal path, further includes: The line to be operated is discretized into multiple track segments, and a complete operation path is encoded as a chromosome consisting of a sequence of track segment IDs; A fitness function is constructed with total energy consumption, total operating time, and total weeding coverage as objectives, and includes constraints on electricity consumption and operating medium consumption. The fitness function introduces density weights in the coverage calculation, so that individuals on paths covering high-density areas obtain higher fitness values, thereby driving the population to evolve in that direction. The chromosomes in the population are iteratively evolved through selection, crossover and mutation operations until the termination condition is met, and the chromosome with the highest fitness is output as the global optimal path. A physical model is generated for the predicted time series, predicted energy consumption series, and predicted working medium consumption series of each node location on the optimal path, including: Predicting time series ; Predicted energy consumption sequence ; Predicting the consumption sequence of working media ; Among them, L i G i C i D i These represent the length, slope, curvature, and weed density of the track segment along the path, respectively. i k is determined by the globally optimal path generated by the genetic algorithm, which is used to plan the speed. drive k medium denoted as energy consumption coefficient, and f, g, and h are preset physical relationship functions.

[0023] This embodiment utilizes a genetic algorithm, using total energy consumption, operation time, and weed coverage as optimization objectives, while strictly adhering to constraints related to power consumption and the weeding medium. The fitness function incorporates weed density weights, intelligently guiding the algorithm to prioritize weed-prone areas, thus improving the targeting of the operation and the final weeding effect. By synchronously generating predicted time, energy consumption, and medium consumption sequences through a physical model, a precise benchmark is provided for subsequent real-time monitoring and emergency decision-making, enhancing the accuracy and controllability of the entire operation system.

[0024] In some embodiments, f, g, h are linear or quadratic functions or empirical formulas fitted based on historical data.

[0025] In some embodiments, the time function f(L) i G i C i D i V i Considering the special kinematic characteristics of high-speed rail track maintenance vehicles: in, Introducing a safety reduction factor η safety This is designed to meet minimum mandatory safety requirements for high-speed rail operations. The piecewise function in the formula handles steep gradients (>G). max Nonlinear velocity decay at this time, curve radius R i Directly affects throughput speed (high-speed rail has strict speed limits on small-radius curves), weed density has an impact factor η weed An exponential decay method is used to simulate the decrease in operating speed in high-density areas. An additional time τ is set to account for the preparation time of visual inspection, robotic arm positioning, etc. setup .

[0026] By introducing a mandatory safety reduction factor, absolute safety and compliance of high-speed rail operations are ensured at the model level. Secondly, the function utilizes piecewise functions, curve radius factors, and exponential decay models to simulate the nonlinear effects of complex conditions such as gradient, curvature, and weed density on operating speed, overcoming the limitations of traditional linear models. Furthermore, the preparation time term makes the predictions more closely resemble actual operational processes.

[0027] In some embodiments, the energy consumption function g(L) i G i C i D i V i Considering the complex energy consumption structure of high-speed rail operating vehicles: in: m vehicle η is the total mass of the operating vehicle. regen : For regenerative braking recovery efficiency (energy can be recovered when the railcar goes downhill), P light It is the lighting power inside the tunnel, which is equal to the tunnel length T. tunnel The function, P climate Here, represents the dehumidification power, and represents the ambient temperature T. amb The function, ε spray Energy consumption per unit of medium spraying.

[0028] This model decomposes the complex energy consumption structure into three parts: traction, auxiliary, and operation. It systematically deconstructs the energy flow and fully considers the regenerative braking energy recovery unique to rail vehicles. At the same time, it incorporates external factors such as tunnel lighting and ambient temperature into the auxiliary energy consumption calculation, making the model highly consistent with actual working conditions.

[0029] In some embodiments, the media consumption function h(Li, Gi, Ci, Di, Vi) takes into account the special needs of weed control along high-speed railway tracks: in: W eff For the effective spray width (considering the shape of the track cross-section), f humidity (RH_i) is a humidity correction function; the dosage needs to be increased when the humidity is low. φ wind This is a wind field influencing factor, related to wind speed v_wind and wind direction θ_wind. opt To achieve the optimal spraying speed and ensure droplet settling effect, μ surfactant The surfactant addition ratio (for use when evaporating rapidly at high temperatures), μ dye Adjust the proportion of fluorescent agent added (for visualization purposes during nighttime operations).

[0030] Preferably, in step S2, the current node location of the vehicle is determined by the Beidou / GPS positioning module integrated in the vehicle, and the predicted value corresponding to the node location is indexed from the predicted time series, predicted energy consumption series and predicted working medium consumption series, and used to calculate the time series deviation ΔT, power deviation ΔE and working medium balance deviation ΔL.

[0031] By integrating a BeiDou / GPS module, precise real-time vehicle positioning was achieved. The real-time location was dynamically matched with the prediction sequence generated in step S1 to index the prediction baseline value of the current node.

[0032] Preferably, the safety threshold in step S3 is dynamically adjusted according to a nonlinear decay function, as shown in the formula: ; in, The standard threshold at the start of the task. This is the minimum threshold before the task ends. Total skylight time, The remaining skylight time is n, where n is a sensitivity adjustment coefficient greater than 1. The sensitivity adjustment coefficient n is used to control the rate at which the threshold decays over time, so that the system maintains a high tolerance in the early stage of the operation, while the sensitivity to deviation increases sharply in the late stage of the operation.

[0033] By introducing a nonlinear decay function containing a sensitivity adjustment coefficient n, the system can precisely control the rate at which the threshold changes over time. This allows the system to maintain high tolerance in the early stages of operation, avoiding frequent emergency decisions triggered by minor fluctuations and ensuring operational stability. Towards the end of the operation, the system's sensitivity rapidly increases, ensuring a quick response to deviations when time is of the essence.

[0034] Preferably, the reward function of the reinforcement learning decision model described in step S4 is: in, For safety rewards, a positive reward is given for safely evacuating or reaching a safe point before the end of the window, and a negative reward is given otherwise. The reward for completing the task is proportional to the effective working area. A basic efficiency bonus is set in proportion to the amount of work done per unit of time and the amount of work done per unit of energy consumption. The deviation penalty term is associated with the planning results of step S1. , where α, β, γ are penalty coefficients.

[0035] The safety reward item prioritizes "safe evacuation before the end of the window," setting a safety baseline for the model's behavior and ensuring absolute operational safety. Task and efficiency rewards drive the model to maximize effective work area and improve resource utilization efficiency while maintaining safety, ensuring the effectiveness of emergency decisions. A deviation penalty item closely links emergency decisions to the overall planning results, penalizing deviations from the original plan. This ensures that the emergency paths generated by the model are not random responses but intelligent corrections based on the original optimal path, guaranteeing system robustness.

[0036] α, β, and γ are penalty coefficients. In some embodiments, the penalty term makes the reinforcement learning decision model tend to choose a strategy that can reduce or correct the current deviation when generating emergency correction paths, thus reflecting the corrective and synergistic effect of emergency decision on global planning.

[0037] Preferably, the state space of the reinforcement learning decision model includes: The vehicle's current status information (such as location, remaining battery power, remaining working medium), remaining sunroof time, and the battery power deviation ΔE, working medium remaining quantity deviation ΔL, and timing deviation ΔT calculated by step S2; The action space of the reinforcement learning decision model is a hybrid action space, including: Discrete actions: {Continue operation as planned, suspend operation, execute return to base, proceed to the nearest safe point}; Continuous actions: {target driving speed, target spray flow rate}.

[0038] The state space includes the vehicle's basic status such as position and battery level, as well as remaining sunroof time. It integrates real-time calculated battery level, timing, and medium deviation, enabling the emergency decision-making model to comprehensively and accurately capture the degree of deviation between the current situation and the original plan, thereby making targeted, data-driven judgments. In the action space, discrete actions (such as returning to base or pausing) determine the strategic direction of the emergency response, while continuous actions (such as adjusting speed or flow rate) achieve fine-grained control.

[0039] Preferably, after step S4, the method further includes: Record complete data for this operation, including actual path, actual energy consumption, actual operating medium consumption, actual operation time, and all triggered emergency correction paths; Based on the recorded data, the physical model parameters of the multi-objective optimization algorithm in step S1 and / or the policy network of the reinforcement learning decision model in step S4 are iteratively optimized and retrained, so that the system can continuously evolve in subsequent tasks.

[0040] By comprehensively recording actual operational data, including path, energy consumption, and all emergency corrections, the system can accurately compare the actual execution results with the planning predictions of S1 and the emergency decisions of S4. This allows for iterative optimization of the path planning model parameters and retraining of the reinforcement learning decision model.

[0041] In some embodiments, the system further includes: 1. Access Control Module: This module is used to hierarchically assign access permissions (Priority I, II, III) to the command center, management personnel, and operational personnel, and to assign corresponding operational permissions. The command center has permissions to view operational status, control weeders, and make automated emergency decisions; management personnel have permissions to view operational status, control weeders, make automated emergency decisions, and handle emergency situations such as remote braking / starting / returning; operational personnel, in addition to having the same basic operational and emergency handling permissions as management personnel, also have permissions to input operational information (skylight start time, vehicle number information, route information, operational personnel / safety officer, etc.).

[0042] The access control module works in real time with the information display module, emergency decision-making module, and operation information input module to assign operation permissions and information display content according to role identity.

[0043] 2. Information Display Module: This module provides differentiated information display based on role and access permissions. For the command center and management personnel, it displays the vehicle's remaining battery power (percentage), operating status (forward, reverse, parked, working), weeder operating status, speed, mileage, and estimated completion time. For operators, it additionally displays on-site monitoring video, on-site alarm information (text + voice), BeiDou positioning information, obstacle avoidance radar information, operation information, sunroof start time, and vehicle number. The information display module communicates in real-time with vehicle sensors (battery sensor, speed sensor, positioning module, etc.), monitoring equipment, and alarm devices, collecting data and then filtering and displaying it according to role and access permissions.

[0044] 3. Operation Information Input Module: This module allows operators to input operation information, including the start and end times of maintenance windows, vehicle number, route information (starting mileage, up / down direction, mileage increments / decrements), operator and safety officer information, etc. It is linked to the access control module, allowing only operators to access it. Inputted information is synchronized to the information display module and the management backend.

[0045] Based on the same concept, the present invention provides an electronic device, comprising: The memory is used to store the processing program; A processor, which, when executing the processing program, implements the weeding operation control method based on path planning and emergency decision-making coordination as described above.

[0046] Based on the same concept, the present invention provides a readable storage medium storing a processing program, which, when executed by a processor, implements the weeding operation control method based on path planning and emergency decision coordination described above.

[0047] The weeding operation control method based on path planning and emergency decision-making coordination, if implemented as program instructions and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment, essentially, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in software. This computer software is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0048] Second Embodiment Based on the same concept, the present invention also provides a weeding operation control system based on path planning and emergency decision-making coordination, comprising: The initial planning module is used to obtain the route information to be operated, the sunroof time, the initial battery power of the vehicle and the initial amount of working medium. It generates a globally optimal path through a multi-objective optimization algorithm, and simultaneously generates the predicted time series, predicted energy consumption series and predicted working medium consumption series of each node position on the optimal path. The globally optimal path and its predicted data are loaded into the weeding vehicle controller. The data acquisition and calculation module is used to collect the actual status data of the vehicle in real time during the operation of the weeding vehicle traveling along the global optimal path, and to calculate the deviation ΔE between the vehicle's instantaneous power and the corresponding node position in the predicted energy consumption sequence, the deviation ΔL of the working medium remaining amount, and the time sequence deviation ΔT between the actual cumulative time and the corresponding node position in the predicted time sequence. The monitoring module is used to determine whether the power deviation ΔE, the working medium balance deviation ΔL, and the timing deviation ΔT exceed the corresponding preset safety thresholds; if so, the reinforcement learning decision model is activated. The emergency decision-making module is used by the reinforcement learning decision-making model to generate an emergency correction path in real time based on the vehicle's current node position, remaining battery power, remaining working medium, and remaining sunroof time, and to control the vehicle to travel according to the emergency correction path.

[0049] The technical solution in this embodiment comprehensively considers time, energy consumption, and working medium consumption through a multi-objective optimization algorithm to generate a globally optimal path and predicted sequence. Through real-time monitoring and dynamic deviation calculation, the system can accurately perceive the deviation between the plan and the actual situation. Once the deviation exceeds the dynamic safety threshold, a reinforcement learning model is activated for intelligent emergency decision-making. The reinforcement learning model can generate the optimal corrected path in real time based on the current state, rather than simply interrupting or returning to the starting point, thus improving the system's adaptability to unexpected working conditions (such as sudden increases in energy consumption or equipment failure). By closely combining "macro-optimal planning" with "micro-intelligent response," the problem of high risk of work interruption under complex constraints is effectively solved, ensuring that weeding operations are completed safely, efficiently, and reliably within the strict time window.

[0050] Those skilled in the art will readily understand that, for the sake of convenience and brevity, the specific identification content executed by the systems and devices described above can be referred to the corresponding processes in the foregoing method embodiments. The embodiments of the present invention have been described in detail above with reference to the accompanying drawings; however, the present invention is not limited to the above embodiments. Even if various modifications are made to the present invention, if these modifications fall within the scope of the claims of the present invention and their equivalents, they shall still fall within the protection scope of the present invention.

Claims

1. A weeding operation control method based on path planning and emergency decision-making coordination, characterized in that, Includes the following steps: S1: Obtain the route information to be operated, the sunroof time, the initial battery power of the vehicle and the initial amount of working medium, generate a globally optimal path through a multi-objective optimization algorithm, and simultaneously generate the predicted time series, predicted energy consumption series and predicted working medium consumption series of each node position on the optimal path, and load the globally optimal path and its predicted data into the weeding vehicle controller. S2: During the operation of the weeding vehicle following the global optimal path, the actual status data of the vehicle is collected in real time, and the deviation ΔE between the vehicle's instantaneous power and the corresponding node position in the predicted energy consumption sequence, the deviation ΔL of the working medium remaining amount, and the time sequence deviation ΔT between the actual cumulative time and the corresponding node position in the predicted time sequence are calculated. S3: Determine whether the power deviation ΔE, working medium balance deviation ΔL, and timing deviation ΔT exceed the corresponding preset safety thresholds; if so, start the reinforcement learning decision model. S4: The reinforcement learning decision model generates an emergency correction path in real time based on the vehicle's current node position, remaining battery power, remaining working medium, and remaining sunroof time, and controls the vehicle to travel according to the emergency correction path.

2. The weeding operation control method based on path planning and emergency decision-making coordination according to claim 1, characterized in that, Step S1, which generates a globally optimal path using a multi-objective optimization algorithm and simultaneously generates the predicted time series, predicted energy consumption series, and predicted working medium consumption series for each node position on the optimal path, further includes: The line to be operated is discretized into multiple track segments, and a complete operation path is encoded as a chromosome consisting of a sequence of track segment IDs; A fitness function is constructed with total energy consumption, total operating time, and total weeding coverage as objectives, and includes constraints on electricity consumption and operating medium consumption. The fitness function introduces density weights in the coverage calculation, so that individuals on paths covering high-density areas obtain higher fitness values, thereby driving the population to evolve in that direction. The chromosomes in the population are iteratively evolved through selection, crossover and mutation operations until the termination condition is met, and the chromosome with the highest fitness is output as the global optimal path. A physical model is generated for the predicted time series, predicted energy consumption series, and predicted working medium consumption series of each node location on the optimal path, including: Predicting time series ; Predicted energy consumption sequence ; Predicting the consumption sequence of working media ; Among them, L i G i C i D i These represent the length, slope, curvature, and weed density of the track segment along the path, respectively. i k is determined by the globally optimal path generated by the genetic algorithm, which is used to plan the speed. drive k medium denoted as energy consumption coefficient, and f, g, and h are preset physical relationship functions.

3. The weeding operation control method based on path planning and emergency decision-making coordination according to claim 1, characterized in that, In step S2, the current node position of the vehicle is determined by the Beidou / GPS positioning module integrated in the vehicle, and the predicted value corresponding to the node position is indexed from the predicted time series, predicted energy consumption series and predicted working medium consumption series, which are used to calculate the time series deviation ΔT, power deviation ΔE and working medium balance deviation ΔL.

4. The weeding operation control method based on path planning and emergency decision-making coordination according to claim 1, characterized in that, The safety threshold in step S3 is dynamically adjusted according to a nonlinear decay function, the formula of which is: ； in, The standard threshold at the start of the task. This is the minimum threshold before the task ends. Total skylight time, The remaining skylight time is n, where n is a sensitivity adjustment coefficient greater than 1. The sensitivity adjustment coefficient n is used to control the rate at which the threshold decays over time, so that the system maintains a high tolerance in the early stage of the operation, while the sensitivity to deviation increases sharply in the late stage of the operation.

5. The weeding operation control method based on path planning and emergency decision-making coordination according to claim 1, characterized in that, The reward function of the reinforcement learning decision model described in step S4 is: in, For safety rewards, a positive reward is given for safely evacuating or reaching a safe point before the end of the window, and a negative reward is given otherwise. The reward for completing the task is proportional to the effective working area. A basic efficiency bonus is set in proportion to the amount of work done per unit of time and the amount of work done per unit of energy consumption. The deviation penalty term is associated with the planning results of step S1. , where α, β, γ are penalty coefficients.

6. The weeding operation control method based on path planning and emergency decision-making coordination according to claim 5, characterized in that, The state space of the reinforcement learning decision model includes: The vehicle's current status information, remaining sunroof time, and the battery deviation ΔE, working medium remaining deviation ΔL, and timing deviation ΔT calculated by step S2; The action space of the reinforcement learning decision model is a hybrid action space, including: Discrete actions: {Continue operation as planned, suspend operation, execute return to base, proceed to the nearest safe point}; Continuous actions: {target driving speed, target spray flow rate}.

7. The weeding operation control method based on path planning and emergency decision-making coordination according to claim 1, characterized in that, The process after step S4 also includes: Record complete data for this operation, including actual path, actual energy consumption, actual operating medium consumption, actual operation time, and all triggered emergency correction paths; Based on the recorded data, the physical model parameters of the multi-objective optimization algorithm in step S1 and / or the policy network of the reinforcement learning decision model in step S4 are iteratively optimized and retrained, so that the system can continuously evolve in subsequent tasks.

8. A weeding operation control system based on path planning and emergency decision-making coordination, characterized in that, include: The initial planning module is used to obtain the route information to be operated, the sunroof time, the initial battery power of the vehicle and the initial amount of working medium. It generates a globally optimal path through a multi-objective optimization algorithm, and simultaneously generates the predicted time series, predicted energy consumption series and predicted working medium consumption series of each node position on the optimal path. The globally optimal path and its predicted data are loaded into the weeding vehicle controller. The data acquisition and calculation module is used to collect the actual status data of the vehicle in real time during the operation of the weeding vehicle traveling along the global optimal path, and to calculate the deviation ΔE between the vehicle's instantaneous power and the corresponding node position in the predicted energy consumption sequence, the deviation ΔL of the working medium remaining amount, and the time sequence deviation ΔT between the actual cumulative time and the corresponding node position in the predicted time sequence. The monitoring module is used to determine whether the power deviation ΔE, working medium balance deviation ΔL, and timing deviation ΔT exceed the corresponding preset safety thresholds; if so, the reinforcement learning decision model is activated. The emergency decision-making module is used by the reinforcement learning decision-making model to generate an emergency correction path in real time based on the vehicle's current node position, remaining battery power, remaining working medium, and remaining sunroof time, and to control the vehicle to travel according to the emergency correction path.

9. An electronic device, characterized in that, include: The memory is used to store the processing program; A processor, which, when executing the processing program, implements the weeding operation control method based on path planning and emergency decision-making coordination as described in any one of claims 1 to 7.

10. A readable storage medium, characterized in that, The readable storage medium stores a processing program, which, when executed by a processor, implements the weeding operation control method based on path planning and emergency decision coordination as described in any one of claims 1 to 7.

Images