Engineering truck intelligent scheduling method based on artificial intelligence

Abstract

The invention discloses an intelligent engineering vehicle scheduling method based on artificial intelligence, and the method comprises the steps: constructing a scheduling objective function which at least comprises the engineering vehicle transportation cost, the task completion time cost, the path congestion cost and the energy consumption cost; establishing a multi-dimensional task demand prediction model based on historical task data and real-time traffic data, and outputting a task distribution prediction value and a traffic state prediction value in a future time period; inputting the model into an engineering vehicle dynamic scheduling model established based on a deep reinforcement learning algorithm for iterative optimization; and generating a real-time scheduling instruction based on the optimized dynamic scheduling model, dynamically allocating task paths and resources of the engineering vehicle, and monitoring an execution state in real time to adjust a scheduling strategy. According to the method, efficient, economical and environment-friendly intelligent scheduling of the engineering vehicle is realized through construction of a scheduling objective function, multi-dimensional demand prediction, deep reinforcement learning dynamic optimization and real-time scheduling and monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of construction vehicle management, and particularly to an intelligent scheduling method for construction vehicles based on artificial intelligence. Background Art

[0002] With the acceleration of the urbanization process and the continuous expansion of the scale of infrastructure construction, the demand for construction vehicles in engineering construction has increased rapidly, which has led to the problem of the operation management efficiency of construction vehicle fleets. In particular, the scheduling problem of construction vehicles has gradually become an important factor affecting project economy and execution efficiency. Traditional construction vehicle scheduling methods usually rely on manual experience-based scheduling methods. In the face of complex and changing engineering task requirements and real-time traffic conditions, it is difficult to achieve global optimization, showing obvious lag and inefficiency.

[0003] In recent years, with the development of artificial intelligence and machine learning technologies, some emerging scheduling methods have gradually introduced optimization algorithms in an attempt to solve the above problems. For example, engineering vehicle path planning and task allocation based on simple historical data analysis or heuristic algorithms can improve local efficiency to a certain extent. However, this method still lacks in-depth comprehensive consideration of multi-dimensional factors in complex scenarios and cannot dynamically adapt to real-time changing traffic conditions and task requirements. The overall optimization degree is still limited, and it is difficult to effectively balance the relationship between transportation costs, execution efficiency, and energy consumption.

[0004] Currently, the existing construction vehicle scheduling technologies generally have the following deficiencies: First, there is a lack of accurate prediction of future task requirements and traffic conditions, and scheduling decisions are made only based on the information at the current moment, resulting in strong blindness in task allocation and easily causing situations such as resource waste or task delays. Second, the scheduling scheme ignores the comprehensive impact of factors such as transportation costs, path congestion costs, and energy consumption, making it difficult to balance economy and environmental benefits. Third, the existing technologies have not fully utilized real-time feedback data for dynamic adaptive adjustment, and the stability and robustness of the model are poor, unable to effectively respond to real-time environmental changes, resulting in the construction vehicle scheduling decision not reaching the global optimal state.

[0005] In summary, the existing construction vehicle scheduling technologies urgently need to be improved in terms of prediction accuracy, comprehensive cost control, and dynamic decision optimization. Therefore, the existing construction vehicle scheduling technologies have problems such as low prediction accuracy of task requirements and traffic conditions, insufficient comprehensive optimization, and inability to adjust the scheduling strategy in real time dynamically. The present invention discloses an intelligent scheduling method for construction vehicles based on artificial intelligence. By constructing a scheduling objective function considering transportation costs, task completion time costs, path congestion costs, and energy consumption costs, and combining a multi-dimensional task demand prediction model and a dynamic scheduling model of deep reinforcement learning algorithm, the dynamic allocation of construction vehicle tasks and resources is realized. Summary of the Invention

[0006] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract of the specification and the title of the invention of this application to avoid blurring the purpose of this section, the abstract of the specification and the title of the invention, and such simplifications or omissions cannot be used to limit the scope of the invention.

[0007] In view of the above existing problems, the present invention is proposed.

[0008] To solve the above technical problems, the present invention provides the following technical solutions: constructing a scheduling objective function, wherein the scheduling objective function at least includes the transportation cost of the engineering vehicle, the task completion time cost, the path congestion cost, and the energy consumption cost;

[0009] Based on historical task data and real-time traffic data, a multi-dimensional task demand forecasting model is established to output task distribution forecasts and traffic status forecasts for future periods.

[0010] Inputting the task distribution prediction value and the traffic state prediction value into a dynamic scheduling model for engineering vehicles established based on a deep reinforcement learning algorithm, wherein the dynamic scheduling model for engineering vehicles uses the scheduling objective function as an optimization target and performs iterative optimization based on historical scheduling data and real-time feedback data;

[0011] Generate real-time scheduling instructions based on the optimized dynamic scheduling model, dynamically allocate engineering vehicle task paths and resources, and monitor the execution status in real time to adjust the scheduling strategy.

[0012] As a preferred solution of the intelligent scheduling method for engineering vehicles based on artificial intelligence according to the present invention, the scheduling objective function includes:

[0013] Let f(x) represent the scheduling objective function and x represent the engineering vehicle scheduling plan, then:

[0014] f(x)=α1·C trans (x)+α2·C time (x)+α3·C congestion (x)+α4·C energy (x)

[0015] Among them, α1, α2, α3, α4 are adjustable weight coefficients, C trans (x) is the transportation cost, C time (x) is the time cost of completing the task, C congestion (x) is the path congestion cost, C energy (x) is the energy consumption cost.

[0016] As a preferred solution of the intelligent scheduling method for engineering vehicles based on artificial intelligence according to the present invention, it specifically includes:

[0017] The transportation cost is measured according to the number of vehicles, the transportation distance, and the cost per unit transportation mileage. Define the cost per unit distance of a single vehicle as β1 and the number of vehicles as N. Then the total transportation cost under the scheduling plan x is expressed as:

[0018]

[0019] Among them, d i (x) represents the total driving mileage of the i-th engineering vehicle to perform the task;

[0020] The task completion time cost is comprehensively calculated according to the execution duration, waiting duration, and default duration of each task. Define the delay penalty coefficient as β2. Then the task completion time cost under the scheduling plan x is expressed as:

[0021]

[0022] Among them, M is the total number of tasks to be executed, and T j (x) represents the total time consumption of the j-th task;

[0023] The path congestion cost is weighted and calculated according to the congestion degree of different roads at different time periods. Define the congestion cost as β3. Then the path congestion cost under the scheduling plan x is expressed as:

[0024]

[0025] Among them, K is the total number of all available roads, and ρ k is the congestion coefficient of road k, and d k (x) is the driving distance of the engineering vehicle on road k;

[0026] The energy consumption cost comprehensively considers vehicle fuel / electricity consumption. Define the energy consumption cost per unit vehicle as β4. Then the energy consumption cost under the scheduling plan x is expressed as:

[0027]

[0028] Among them, E i (x) represents the total energy consumption of the i-th engineering vehicle under the scheduling plan x.

[0029] As a preferred solution of the intelligent scheduling method for engineering vehicles based on artificial intelligence according to the present invention, the multi-dimensional task demand prediction model includes:

[0030] The historical task data at least includes the time stamp, geographical location, task type, and vehicle usage of the task;

[0031] The real-time traffic data at least includes road traffic flow, congestion index, traffic accidents, and traffic restrictions information;

[0032] Perform missing value filling, outlier removal, and data format unification processing on the historical task data and the real-time traffic data;

[0033] Rearrange the historical task data in time series, and extract time series features of different periods of day, week, and month;

[0034] Adopt a time series model based on LSTM to jointly predict the future short-term and medium- to long-term task demand trends. Among them, calculate the root mean square error of the time series model on the validation set as an evaluation index. If the evaluation index is higher than the first threshold, automatically perform model parameter tuning, otherwise proceed to the next step;

[0035] According to the geographical location, perform grid processing on the historical task data to obtain several spatial sub-regions. For each sub-region, combine the task occurrence frequency and vehicle driving distance characteristics to construct a spatial distribution prediction model. Among them, within each sub-region, compare the deviation between the predicted value and the actual historical data. If the deviation exceeds the second threshold, retrain the spatial distribution prediction model, otherwise keep the existing model;

[0036] Based on the real-time collected road traffic flow and congestion index, adopt a regression model to dynamically predict the future congestion situation of different road sections, and superimpose traffic accidents and traffic restrictions information to correct the congestion probability of each road section under different time windows. Among them, if some road sections are in a highly congested state for a long time in the prediction, mark them as risk road sections and increase the congestion cost weight, otherwise process them according to the conventional congestion coefficient;

[0037] According to the engineering vehicle task category, use a classification model to independently predict the demand for various tasks, and perform multi-dimensional fusion on the judgment results obtained in the above steps to output the task distribution prediction value and traffic state prediction value within the future time period;

[0038] If the overall prediction error after fusion is greater than the third threshold, re-perform iterative training, otherwise output the prediction result.

[0039] As a preferred solution of the engineering vehicle intelligent scheduling method based on artificial intelligence described in the present invention, the engineering vehicle dynamic scheduling model includes:

[0040] Select a deep Q-network as the deep reinforcement learning algorithm;

[0041] Use the task distribution prediction value and the traffic state prediction value as the environmental state vector of the deep reinforcement learning algorithm;

[0042] Define the executable action space in the deep reinforcement learning algorithm, where the action space includes the scheduling and allocation of vehicles and path selection;

[0043] Split the rewards based on the scheduling objective function. Specifically, if the scheduling decision reduces the transportation cost, time cost, congestion cost, and vehicle energy consumption, a positive reward is given; if the scheduling decision causes delays, congestion, and excessive energy consumption, a negative reward is given;

[0044] After each scheduling decision cycle ends, the cumulative reward is used to update the model parameters.

[0045] As a preferred solution of the intelligent scheduling method for engineering vehicles based on artificial intelligence according to the present invention, the engineering vehicle dynamic scheduling model takes the scheduling objective function as the optimization goal and performs iterative optimization with historical scheduling data and real-time feedback data, including:

[0046] Conduct preliminary training with historical scheduling data to obtain an initial policy network. During the training process, if it is found that after several consecutive iterations, the average reward of the model in the verification environment does not meet the fourth threshold, the learning rate is adjusted; otherwise, continue training until convergence;

[0047] Deploy the engineering vehicle dynamic scheduling model after preliminary training to the scheduling system, and obtain the scheduling decision by inputting the environmental state vector;

[0048] Compare the objective function value under the scheduling decision with the historical average value. If the overall cost drops significantly, proceed to the next step; otherwise, continue model iteration;

[0049] Among them, the objective function value includes transportation cost, time cost, congestion cost, and energy consumption;

[0050] Obtain the running status of the engineering vehicle and the completion status of task execution in real time, and write the real-time scheduling feedback data into the data buffer pool for data update;

[0051] Perform online fine-tuning on the engineering vehicle dynamic scheduling model after preliminary training through the real-time scheduling feedback data;

[0052] If the increase in the average cumulative reward obtained from the new round of training is less than the fifth threshold compared to the previous round and this persists for several consecutive iterations, automatically adjust the network structure parameters; if the average cumulative reward continues to increase, continue iterative training;

[0053] Repeat this process until the optimal level of the cost target is reached.

[0054] As a preferred solution of the intelligent dispatching method for engineering vehicles based on artificial intelligence according to the present invention, real-time dispatching instructions are generated based on the optimized dynamic dispatching model, and the task paths and resources of engineering vehicles are dynamically allocated, including:

[0055] Obtain the current optimal dispatching decision from the optimized dynamic dispatching model;

[0056] For each task to be executed, generate a dispatching instruction in combination with the current state of the vehicle. The dispatching instruction includes the start and end positions of the task, the passing route, the estimated execution duration, and the required resource configuration. Among them, the start and end times, the required mileage, and the remaining load of the vehicle for each task are quickly verified. If the resources are insufficient or there are time conflicts, substitute vehicles are called or the task is split. Otherwise, the verification passes;

[0057] Based on the predicted traffic state value, initially select a path with low congestion risk and short driving distance. Among them, if it is detected that there are traffic restrictions, serious traffic accidents, or severe congestion along the path, the replanning strategy is automatically triggered to select a sub-optimal path. Otherwise, execute according to the original path;

[0058] Consider the functional characteristics of the vehicle type and the fuel type, and perform optimal matching according to the task attributes. When the function of a single vehicle cannot meet the requirements, the task is split and multi-vehicle collaborative dispatching is adopted. Among them, if the additional cost of collaborative dispatching is higher than the expected benefit, the single-vehicle dispatching scheme is rolled back. Otherwise, the multi-vehicle collaborative scheme is executed.

[0059] As a preferred solution of the intelligent dispatching method for engineering vehicles based on artificial intelligence according to the present invention, the execution state is monitored in real time to adjust the dispatching strategy, including:

[0060] Real-time collect the vehicle's geographical coordinates by satellite positioning, and record the running trajectory of the vehicle during the task execution;

[0061] Monitor the time consumption and completion progress of the vehicle in different stages of loading, driving, and unloading, and count the task completion percentage;

[0062] Collect the vehicle load information, fuel / battery remaining information. If it is detected that the vehicle's energy consumption is too high / the battery remaining is insufficient, trigger the resource replenishment operation;

[0063] Compare with the estimated execution time, energy consumption level, and path conditions in the dispatching instruction;

[0064] If the measured execution time exceeds the preset delay threshold, trigger the emergency dispatching strategy;

[0065] If the actual energy consumption of the vehicle exceeds the preset energy consumption threshold, use this record as a negative sample and feed it back to the dynamic dispatching model;

[0066] If the execution results are basically consistent with the model prediction results, they will be used as positive samples to consolidate the model strategy.

[0067] As a preferred solution of the intelligent scheduling method for engineering vehicles based on artificial intelligence according to the present invention, the adjustment of the scheduling strategy includes the following steps for dynamically correcting the scheduling of engineering vehicles according to the monitoring results of the execution status:

[0068] When a vehicle encounters sudden congestion, an accident, or a vehicle anomaly during its execution, the dynamic scheduling model recalculates the optimal path based on the new environmental state vector. If the reduction in total cost after replanning is greater than the switching cost caused by the replanning, the new scheduling plan is implemented; otherwise, the existing strategy is maintained.

[0069] If delays in multiple batches of tasks or serious imbalance in resource allocation are detected, the matching relationship, driving path and execution order of all available vehicles and tasks will be recalculated, and new global scheduling instructions will be deployed.

[0070] Beneficial effects of the present invention:

[0071] 1. By constructing a scheduling objective function that comprehensively considers the transportation cost, task completion time cost, route congestion cost, and energy consumption cost of engineering vehicles, it quantifies and weighs various important cost factors in the engineering vehicle scheduling process. This provides a unified and comprehensive optimization goal for intelligent scheduling, more effectively controls comprehensive costs during the actual scheduling process, and improves the economic efficiency and environmental friendliness of engineering vehicle operation.

[0072] 2. By establishing a task demand forecasting model based on historical task data and real-time traffic data, encompassing multiple dimensions including time, space, traffic status, and task type, this model enables accurate prediction of future engineering task demands and traffic conditions. This makes engineering vehicle scheduling more predictable and forward-looking, effectively reducing the blindness and uncertainty of resource allocation during task execution, and helping to prevent route congestion, task delays, and resource waste risks in advance.

[0073] 3. By dynamically responding to and optimizing the complex and ever-changing real-time traffic and engineering task status, the model continuously updates its learning ability for optimal decisions in different scenarios. During the actual scheduling process, the model can adaptively make better decisions, continuously optimize its own scheduling strategy, and improve the robustness and stability of the scheduling model to external disturbances.

[0074] 4. It is efficiently and accurately implemented through the intelligent scheduling strategy, and at the same time, the actual effect is fed back to the dynamic scheduling model in real time, continuously correcting and improving the scheduling strategy, further improving the efficiency of construction vehicle resource allocation, the timeliness of task completion, the rationality of path planning, and the energy utilization efficiency, effectively avoiding real-time emergencies and task execution deviations that cannot be handled by traditional static scheduling methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings. Among them:

[0076] Figure 1 It is a flowchart of an intelligent scheduling method for construction vehicles based on artificial intelligence shown in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0077] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will describe the specific embodiments of the present invention in detail with reference to the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them.

[0078] Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.

[0079] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0080] According to an embodiment of the present invention, in combination with Figure 1 the flowchart shown, an intelligent scheduling method for construction vehicles based on artificial intelligence specifically includes the following steps:

[0081] S【1】. Construct a scheduling objective function, which at least includes the transportation cost of construction vehicles, the task completion time cost, the path congestion cost, and the energy consumption cost. Among them, it should be noted in this step that:

[0082] The transportation cost is measured according to the number of vehicles, the transportation distance, and the unit transportation mileage cost. Define the single-vehicle distance cost as β【1】, and the number of vehicles as N. Then the total transportation cost under the scheduling plan x is expressed as:

[0083]

[0084] Among them, d i (x) represents the total driving mileage of the i-th construction vehicle to perform tasks;

[0085] The task completion time cost is comprehensively calculated based on the execution duration, waiting duration, and default duration of each task. Define the delay fine coefficient as β2, then the task completion time cost under the scheduling scheme x is expressed as:

[0086]

[0087] Among them, M is the total number of tasks to be executed, T j (x) represents the total time consumption of the j-th task;

[0088] The path congestion cost is calculated by weighting according to the congestion degree of different roads at different time periods. Define the congestion cost as β3, then the path congestion cost under the scheduling scheme x is expressed as:

[0089]

[0090] Among them, K is the total number of all available roads, ρ k is the congestion coefficient of road k, d k (x) is the driving distance of the construction vehicle on road k;

[0091] The energy consumption cost comprehensively considers vehicle fuel / electricity consumption. Define the energy consumption cost per vehicle as β4, then the energy consumption cost under the scheduling scheme x is expressed as:

[0092]

[0093] Among them, E i (x) represents the total energy consumption of the i-th construction vehicle under the scheduling scheme x.

[0094] As an example, the scheduling objective function is expressed by the following formula. Let f(x) represent the scheduling objective function and x represent the construction vehicle scheduling scheme, then:

[0095] f(x) = α1·C trans (x) + α2·C time (x) + α3·C congestion (x) + α4·C energy (x)

[0096] Among them, α1, α2, α3, α4 are adjustable weight coefficients, C trans (x) is the transportation cost, C time (x) is the task completion time cost, C congestion (x) is the path congestion cost, Cenergy (x) is the energy consumption cost.

[0097] S2. Based on historical task data and real-time traffic data, establish a multi-dimensional task demand prediction model, and output the predicted values of task distribution and traffic status in the future time period. Among them, it should be noted in this step that:

[0098] The historical task data at least includes the time mark of the task (such as date, hour), geographical location (such as longitude and latitude, area number), task type, and vehicle usage (such as the number of vehicles, vehicle load, scheduling frequency);

[0099] The real-time traffic data at least includes road traffic flow, congestion index, traffic accidents, and traffic restriction information;

[0100] Perform missing value filling, outlier removal, and data format unification processing on historical task data and real-time traffic data;

[0101] Rearrange the historical task data according to the time series, and extract the time series characteristics of different cycles of day, week, and month;

[0102] Adopt a time series model based on LSTM to jointly predict the task demand trends in the short-term and medium-long term in the future. Among them, calculate the root mean square error of the time series model on the validation set as the evaluation index. If the evaluation index is higher than the first threshold, automatically perform model parameter tuning, otherwise enter the next step;

[0103] Exemplarily, let X t represent several historical time series data before time t, represent the predicted values of the demand for time t and several subsequent time periods by the model;

[0104] Adopt a long short-term memory (LSTM) network to model X t and output

[0105] Evaluate the root mean square error RMSE of the model on the validation set. If:

[0106]

[0107] then automatically perform model parameter tuning;

[0108] If RMSE ≤ TH1, enter the next step;

[0109] The first threshold (TH1) is set to 1.2 - 1.5 times the mean RMSE on the past training set. If it exceeds this ratio, it indicates that the model prediction deviation is too large and the learning rate needs to be adjusted.

[0110] In an alternative embodiment, historical task data is partitioned into grids according to geographical locations to obtain a number of spatial sub-regions, and each sub-region corresponds to a road region in a city;

[0111] For each sub-region, its historical task frequency, vehicle driving distance, and vehicle demand characteristics are counted;

[0112] Let F z represent the historical feature vector of sub-region z (such as the task frequency in the past period and the average vehicle running distance), and represent the predicted value of the task demand in this sub-region in the future period;

[0113] A multiple linear regression model is used to estimate the demand of each sub-region;

[0114] Compare the predicted value with the deviation of the actual historical data D z ; If Δ z > the second threshold (TH2), it indicates that the prediction accuracy of the model in this sub-region is poor and needs to be retrained;

[0115] If Δ z ≤ TH2, the existing model is maintained;

[0116] The setting of the second threshold (TH2) is determined based on the importance of each sub-region's impact on the overall scheduling decision.

[0117] In an alternative embodiment, let G k,t represent the traffic feature vector of road k collected at time t (including road traffic flow, congestion index, whether an accident or traffic restriction occurs), and represent the predicted value of the congestion degree of road k in the future period t;

[0118] A random forest regression model is used to estimate ;

[0119] If, for some roads k, exceeds the preset congestion critical value for a long time during prediction, road k is marked as a high-risk congested section, and its congestion cost weight is increased in subsequent scheduling decisions;

[0120] If remains at a normal level, it is handled according to the normal congestion coefficient;

[0121] When the system detects that an accident has occurred on road k or there is a temporary traffic restriction announcement, is corrected to output a congestion prediction value that is more in line with the actual situation.

[0122] Further, according to the task categories of engineering vehicles, the classification model is used to independently predict various task requirements, and the judgment results obtained in the above steps are fused in multiple dimensions to output the predicted values of task distribution and traffic state in the future time period;

[0123] If the overall prediction error after fusion is greater than the third threshold, iterative training is carried out again; otherwise, the prediction results are output, that is, the predicted values of task distribution and traffic state in the future time period are output;

[0124] Among them, the task categories of engineering vehicles include at least earthwork transportation, equipment handling, and concrete delivery.

[0125] As an example, the multi-dimensional task requirement prediction model is expressed as:

[0126]

[0127] Among them, X t is a high-dimensional input vector that synthesizes time, space, traffic, and task type information, represents the predicted value of task distribution in the future time period t, represents the predicted value of traffic state in the future time period t;

[0128] It should be noted that the time dimension prediction and the space dimension prediction are combined by weighting to obtain the traffic state prediction and the task type prediction are concatenated in the feature space to obtain a comprehensive prediction of different types of task requirements under various road congestion conditions;

[0129] Finally, the above multi-dimensional results are integrated into a unified prediction output:

[0130]

[0131] Select a comprehensive error index (such as weighted multi-dimensional RMSE) to evaluate the fused and ;

[0132] If the comprehensive error > the third threshold TH3, iterative training is carried out again;

[0133] If the comprehensive error ≤ the third threshold TH3, the prediction results are output as the input of the subsequent dynamic scheduling model.

[0134] S3. Input the task distribution prediction value and the traffic state prediction value into the engineering vehicle dynamic scheduling model established based on the deep reinforcement learning algorithm. The engineering vehicle dynamic scheduling model takes the scheduling objective function as the optimization goal and performs iterative optimization with historical scheduling data and real-time feedback data. It should be noted in this step that:

[0135] Let π θ represent the decision-making policy function of the engineering vehicle dynamic scheduling model, where θ is the learnable parameter of the model. In each decision-making cycle, the model outputs the scheduling allocation and path selection action a for the vehicle based on the current environmental state s t ; t

[0136] The environmental state s t is composed of the task distribution prediction value and the traffic state prediction value output by step S2;

[0137] In DQN (Deep Q-Network), a function Q(s, a; θ) is defined to estimate the expected return value that can be obtained by taking action a in state s;

[0138] The value of θ is continuously updated through backpropagation to make the estimation of Q(s, a; θ) approximate the true return.

[0139] In an alternative implementation, the scheduling objective function f(x) constructed in step S1 of this embodiment, where x represents the engineering vehicle scheduling plan, and the function includes at least transportation cost, task completion time cost, path congestion cost, and energy consumption cost;

[0140] When the model successfully reduces one or more of the above cost factors in a single scheduling decision, a positive reward is given; if it causes delays, congestion, or excessive energy consumption, a negative reward is given;

[0141] It should be noted that the reward value r t is calculated by comparing the improvement degree of this scheduling decision on several cost indicators;

[0142] Let A be the action space in the present invention. Each scheduling decision includes the task allocation selection and path selection for several engineering vehicles;

[0143] In DQN, the max a∈A Q(s t , a; θ) is used to determine the current best action in order to maximize the cumulative reward in the long term;

[0144] Construct the training sample set in the format of <s, a, r, s'> with historical scheduling data (such as the vehicle scheduling decisions, cost expenditures, and traffic congestion levels actually recorded in past projects);

[0145] This data serves as the experience replay pool for DQN offline training, helping the model learn the initial Q function;

[0146] Randomly extract small batches of samples from the training sample set repeatedly, and iteratively update Q(s,a;θ);

[0147] After several rounds of training, obtain a preliminary policy network The average reward value on the validation set is denoted as

[0148] Let TH4 (the fourth threshold) represent the minimum average reward requirement in the validation environment. If, after several consecutive iterations, it is always lower than TH4, then perform learning rate adjustment; if then it indicates that the model has reached an available level and enters the next stage of deployment and online training;

[0149] The setting of TH4 is determined based on the average cost level of historical decisions.

[0150] In an alternative implementation, the obtained through preliminary training is

[0151] Deployed in the engineering vehicle scheduling system; t , the model outputs the action a t ;

[0152] If the objective function value (such as the comprehensive cost) obtained after the current scheduling decision is significantly lower than the historical average, it indicates that the decision has been significantly improved; otherwise, it indicates that the model needs to continue iterating;

[0153] The system continuously obtains the vehicle operation status and the task execution completion situation, and records the newly generated <s t ,a t ,r t ,s t+1 > in the data buffer pool to prepare for subsequent online training;

[0154] Let represent the average reward in the nth round of the online training stage;

[0155] After each cycle ends, perform a small batch update on θ according to the real-time scheduling feedback data, so that Q(s,a;θ) gradually adapts to the newly emerging environmental states and task requirements;

[0156] Let TH5 (the fifth threshold) represent the average reward improvement amplitude threshold in consecutive multiple rounds of iteration of the model, used to judge whether the model continues to maintain effective improvement;

[0157] If the obtained in the new round of training is relative to the previous round Improvement If it continues to be lower than TH5 for a long time, it indicates that the improvement of the model has slowed down and there is no significant improvement after multiple iterations; then the network structure parameters are automatically adjusted (such as adding or deleting network layers);

[0158] If it still remains greater than TH5, it means that the model is still in the rapid convergence stage and no large-scale structure adjustment is required;

[0159] After multiple iterations of online training, if it remains less than TH5, it indicates that the model tends to converge and a relatively mature dynamic scheduling strategy for construction vehicles is obtained;

[0160] Let θ * represent the optimal parameter set obtained after the above iterative optimization, then the final dynamic scheduling model is denoted as

[0161] Exemplarily, deployed on the weights θ of the deep neural network * " when receiving the environmental state s t the model outputs the optimal action through the forward propagation of the network

[0162]

[0163] Thus, tasks are dynamically assigned to construction vehicles and paths are planned.

[0164] S4. Generate real-time scheduling instructions based on the optimized dynamic scheduling model, dynamically allocate the task paths and resources of construction vehicles, and monitor the execution status in real time to adjust the scheduling strategy. Among them, it should be noted in this step that:

[0165] Obtain the current optimal scheduling decision from the optimized dynamic scheduling model;

[0166] For each task to be executed, generate a scheduling instruction in combination with the current state of the vehicle. The scheduling instruction includes the start and end positions of the task, the route to pass through, the estimated execution duration, and the required resource configuration; among them, quickly verify the start and end times, the required mileage, and the remaining load of each vehicle. If the resources are insufficient or there are time conflicts, call in substitute vehicles or split the task, otherwise the verification passes;

[0167] Based on the predicted traffic state value, initially select a path with low congestion risk and short driving distance; among them, if it is detected that there are traffic restrictions, serious traffic accidents, or severe congestion along the path, automatically trigger a replanning strategy and select a sub-optimal path, otherwise execute according to the original path;

[0168] Exemplarily, the replanning strategy is to determine whether to switch to a sub-optimal path. If Δf(x) > Δcost switch , then perform a path switch and switch to the sub-optimal path;

[0169] where Δf(x) represents the overall cost reduction due to changing the path, and Δcost switch represents the additional burden caused by the path switch.

[0170] Considering the functional characteristics of the vehicle type and the fuel type, perform optimal matching according to the task attributes. When the function of a single vehicle cannot meet the requirements, split the task and adopt multi-vehicle collaborative scheduling. Among them, if the additional cost of collaborative scheduling is higher than the expected revenue, fallback to the single-vehicle scheduling scheme; otherwise, execute the multi-vehicle collaborative scheme;

[0171] Exemplarily, if ΔR co-op > ΔC coordination , then execute the multi-vehicle collaborative scheme;

[0172] where ΔR co-op represents the overall revenue increment after multi-vehicle collaboration, and ΔC coordination represents the additional overhead of fleet collaborative scheduling.

[0173] In an alternative embodiment, monitor the execution status in real time to adjust the scheduling strategy, including:

[0174] Collect the vehicle's geographical coordinates in real time through satellite positioning and record the vehicle's running trajectory during the task execution;

[0175] Monitor the time consumption and completion progress of the vehicle at different stages of loading, driving, and unloading, and count the task completion percentage;

[0176] Collect the vehicle load information and fuel / battery remaining information. If it is detected that the vehicle has high energy consumption or insufficient battery remaining, trigger a resource replenishment operation;

[0177] Compare with the estimated execution time, energy consumption level, and path conditions in the scheduling instruction;

[0178] If the measured execution time exceeds the preset delay threshold, trigger an emergency scheduling strategy;

[0179] If the actual energy consumption of the vehicle exceeds the preset energy consumption threshold, use this record as a negative sample and feedback it to the dynamic scheduling model;

[0180] If the execution result is basically consistent with the model prediction result, use it as a positive sample to consolidate the model strategy.

[0181] Specifically, the adjustment of the scheduling strategy includes the following steps for dynamically correcting the engineering vehicle scheduling according to the monitoring results of the execution status:

[0182] When the vehicle encounters sudden congestion, accidents or vehicle anomalies during operation, the dynamic scheduling model recalculates the optimal path according to the new environmental state vector; among them, if the reduction in the total cost after replanning is greater than the switching cost brought by replanning, the new scheduling plan is executed, otherwise the existing strategy is maintained;

[0183] If it is detected that multiple batches of tasks are delayed or the resource allocation is seriously unbalanced, recalculate the matching relationship, driving path and execution order of all available vehicles and each task, and deploy new global scheduling instructions.

[0184] The above methods for filling missing values, removing outliers and unifying data formats of the collected data can be implemented by means or methods in the existing technology, and will not be elaborated herein.

[0185] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. An intelligent dispatching method for engineering vehicles based on artificial intelligence, characterized in that, Including: Construct a scheduling objective function, which at least includes the transportation cost of engineering vehicles, the task completion time cost, the path congestion cost, and the energy consumption cost; Based on historical task data and real-time traffic data, establish a multi-dimensional task demand prediction model, and output the predicted values of task distribution and traffic status in the future time period; Input the predicted values of task distribution and traffic status into the dynamic scheduling model of engineering vehicles established based on the deep reinforcement learning algorithm. The dynamic scheduling model of engineering vehicles takes the scheduling objective function as the optimization target and is iteratively optimized with historical scheduling data and real-time feedback data; Generate real-time scheduling instructions based on the optimized dynamic scheduling model, dynamically allocate the task paths and resources of engineering vehicles, and monitor the execution status in real time to adjust the scheduling strategy.

2. The intelligent scheduling method for engineering vehicles based on artificial intelligence according to claim 1, wherein The scheduling objective function includes: Let f(x) represent the scheduling objective function and x represent the engineering vehicle scheduling plan, then: f(x) = α1·C trans (x) + α2·C time (x) + α3·C congestion (x) + α4·C energy (x), where α1, α2, α3, α4 are adjustable weight coefficients, C trans (x) is the transportation cost, C time (x) is the task completion time cost, C congestion (x) is the path congestion cost, C energy (x) is the energy consumption cost.

3. The intelligent dispatching method for engineering vehicles based on artificial intelligence according to claim 2, wherein Specifically including: The transportation cost is measured according to the number of vehicles, the transportation distance, and the unit transportation mileage cost. Define the single-vehicle distance cost as β1 and the number of vehicles as N. Then the total transportation cost under the scheduling plan x is expressed as: Among them, d i (x) represents the total driving mileage of the i-th construction vehicle to perform tasks; The task completion time cost is comprehensively calculated according to the execution duration, waiting duration, and default duration of each task. Define the delay penalty coefficient as β2. Then the task completion time cost under the scheduling plan x is expressed as: where M is the total number of tasks to be executed, and T j (x) represents the total time consumption of the j-th task; The path congestion cost is weighted and calculated according to the congestion degree of different roads at different times. Define the congestion cost as β3. Then the path congestion cost under the scheduling plan x is expressed as: Among them, K is the total number of all available roads, and ρ k is the congestion coefficient of road k, and d k (x) is the driving distance of the engineering vehicle on road k; The energy consumption cost comprehensively considers the fuel / electricity consumption of vehicles. Define the single-vehicle energy consumption cost as β4. Then the energy consumption cost under the scheduling plan x is expressed as: Among them, E i (x) represents the total energy consumption of the i-th engineering vehicle under the scheduling plan x.

4. The intelligent scheduling method for engineering vehicles based on artificial intelligence according to claim 1, wherein The multi-dimensional task demand prediction model includes: The historical task data at least includes the time mark, geographical location, task type, and vehicle usage of the task; The real-time traffic data at least includes road traffic flow, congestion index, traffic accidents, and traffic restriction information; Perform missing value filling, outlier removal, and data format unification processing on the historical task data and the real-time traffic data; Rearrange the historical task data in time series and extract the time series features of different cycles of day, week, and month; Adopt a time series model based on LSTM to jointly predict the task demand trends in the short-term and medium- to long-term future. Among them, calculate the root mean square error of the time series model on the validation set as the evaluation index. If the evaluation index is higher than the first threshold, automatically perform model parameter tuning, otherwise enter the next step; According to the geographical location, perform grid processing on the historical task data to obtain several spatial sub-regions. For each sub-region, combine the task occurrence frequency and vehicle driving distance characteristics to construct a spatial distribution prediction model. Among them, within each sub-region, compare the deviation between the predicted value and the actual historical data. If the deviation exceeds the second threshold, retrain the spatial distribution prediction model, otherwise keep the existing model; Based on the real-time collected road traffic flow and congestion index, a regression model is used to dynamically predict the future congestion situation of different road sections. By overlaying traffic accident and traffic restriction information, the congestion probability of each road section under different time windows is corrected. Among them, if some road sections are in a high congestion state for a long time in the prediction, they are marked as risk road sections and the congestion cost weight is increased, otherwise they are processed according to the conventional congestion coefficient; According to the task categories of engineering vehicles, a classification model is used to independently predict the demand for various tasks, and the judgment results obtained in the above steps are fused in multiple dimensions to output the predicted values of task distribution and traffic state in the future time period; If the overall prediction error after fusion is greater than the third threshold, iterative training is carried out again, otherwise the prediction result is output.

5. The intelligent dispatching method for engineering vehicles based on artificial intelligence according to claim 1 or 4, characterized in that, The engineering vehicle dynamic scheduling model includes: Select the deep Q-network as the deep reinforcement learning algorithm; Take the predicted value of the task distribution and the predicted value of the traffic state as the environmental state vector of the deep reinforcement learning algorithm; Define the executable action space in the deep reinforcement learning algorithm, and the action space includes the scheduling allocation and path selection of vehicles; Based on the scheduling objective function, the reward is split. Among them, if the scheduling decision reduces the transportation cost, time cost, congestion cost and vehicle energy consumption, a positive reward is given. If the scheduling decision causes delays, congestion and excessive energy consumption, a negative reward is given; After the end of each scheduling decision cycle, the cumulative reward is used to update the model parameters.

6. The intelligent dispatching method for engineering vehicles based on artificial intelligence according to claim 5, characterized in that The engineering vehicle dynamic scheduling model takes the scheduling objective function as the optimization goal and performs iterative optimization with historical scheduling data and real-time feedback data, including: Through preliminary training with historical scheduling data, an initial policy network is obtained. Among them, when it is found during the training process that after several consecutive iterations, the average reward of the model in the verification environment does not meet the fourth threshold, the learning rate is adjusted, otherwise the training continues until convergence; Deploy the engineering vehicle dynamic scheduling model after preliminary training into the scheduling system, and obtain the scheduling decision by inputting the environmental state vector; Compare the objective function value under the scheduling decision with the historical average value. If the overall cost drops significantly, proceed to the next step, otherwise continue the model iteration; Among them, the objective function value includes transportation cost, time cost, congestion cost and energy consumption; Obtain the running state of the engineering vehicle and the completion situation of task execution in real time, and write the real-time scheduling feedback data into the data buffer pool for data update; Perform online fine-tuning on the engineering vehicle dynamic scheduling model after preliminary training through the real-time scheduling feedback data; If the increase in the average cumulative reward obtained from the new round of training is less than the fifth threshold compared with the previous round and this continues for several iterations, the network structure parameters are automatically adjusted. If the average cumulative reward continues to increase, continue the iterative training; Repeat this process until the optimal level of the cost target is reached.

7. The intelligent dispatching method for engineering vehicles based on artificial intelligence according to claim 6, wherein Generate real-time scheduling instructions based on the optimized dynamic scheduling model, and dynamically allocate the task paths and resources of engineering vehicles, including: Obtain the current optimal scheduling decision from the optimized dynamic scheduling model; For each task to be executed, a scheduling instruction is generated in combination with the current state of the vehicle. The scheduling instruction includes the start and end positions of the task, the route to pass through, the estimated execution duration, and the required resource configuration. Among them, the start and end times, the required mileage, and the remaining load of each task are quickly verified. If the resources are insufficient or there are time conflicts, substitute vehicles are called or the tasks are split. Otherwise, the verification passes. Based on the predicted traffic state value, a path with low congestion risk and short driving distance is initially selected. Among them, if it is detected that there are traffic restrictions, serious traffic accidents, or severe congestion along the path, the re-planning strategy is automatically triggered to select the sub-optimal path. Otherwise, it is executed according to the original path. Considering the functional characteristics of the vehicle type and the fuel type, the optimal matching is carried out according to the task attributes. When the function of a single vehicle cannot meet the requirements, the task is split and multi-vehicle collaborative scheduling is adopted. Among them, if the additional cost of collaborative scheduling is higher than the expected revenue, it falls back to the single-vehicle scheduling plan. Otherwise, the multi-vehicle collaborative plan is executed.

8. The intelligent dispatching method for engineering vehicles based on artificial intelligence according to claim 7, characterized in that The execution status is monitored in real time to adjust the scheduling strategy, including: The geographical coordinates of the vehicle are collected in real time through satellite positioning, and the running track of the vehicle during the execution of the task is recorded. Monitor the time consumption and completion progress of the vehicle in different stages of loading, driving, and unloading, and count the task completion percentage. Collect the vehicle load information, fuel / battery remaining information. If it is detected that the vehicle energy consumption is too high / the battery remaining is insufficient, trigger the resource replenishment operation. Compare with the estimated execution time, energy consumption level, and path conditions in the scheduling instruction. If the measured execution time exceeds the preset delay threshold, trigger the emergency scheduling strategy. If the actual energy consumption of the vehicle exceeds the preset energy consumption threshold, this record is fed back to the dynamic scheduling model as a negative sample. If the execution result is basically consistent with the model prediction result, it is used as a positive sample to consolidate the model strategy.

9. The intelligent dispatching method for engineering vehicles based on artificial intelligence according to claim 8, characterized in that The adjustment of the scheduling strategy includes the following steps for dynamically correcting the engineering vehicle scheduling according to the monitoring results of the execution status: When the vehicle encounters sudden congestion, accidents, or vehicle abnormalities during the execution process, the dynamic scheduling model recalculates the optimal path according to the new environmental state vector. Among them, if the reduction in the total cost after re-planning is greater than the switching cost brought by re-planning, the new scheduling plan is executed. Otherwise, the existing strategy is maintained. If it is detected that multiple batches of tasks are delayed or the resource allocation is seriously unbalanced, recalculate the matching relationship, driving path, and execution order of all available vehicles and each task, and deploy a new global scheduling instruction.

Images