CN-122022631-A - Logistics vehicle intelligent scheduling method and system based on big data

CN122022631ACN 122022631 ACN122022631 ACN 122022631ACN-122022631-A

Abstract

The invention relates to the technical field of logistics big data and discloses an intelligent logistics vehicle scheduling method and system based on big data. The method is characterized in that multisource heterogeneous data from vehicle-mounted sensors, traffic monitoring platforms and the like are collected in real time, and a vehicle-environment state feature vector is generated by utilizing a data fusion technology after distributed cleaning, standardized processing and outlier detection. And then, constructing a vehicle scheduling decision model based on the deep reinforcement learning framework, generating and optimizing a scheduling sequence through a strategy network and a multi-objective optimization algorithm, and finally encoding into an executable instruction set to drive a vehicle to execute. The system also realizes closed-loop dynamic optimization, ensures self-adaptive environment change of a scheduling strategy, and improves logistics efficiency.

Inventors

CHENG CHAORAN
ZHANG XU
LIU CHANG
SONG BING
PEI JUNCHAO

Assignees

安徽邮谷快递智能科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260202

Claims (10)

1. The logistics vehicle intelligent scheduling method based on big data is characterized by comprising the following steps: Acquiring data related to logistics vehicle dispatching in real time from a plurality of heterogeneous data sources, wherein the data sources comprise vehicle-mounted sensors, a traffic monitoring platform, a cargo management database and a historical operation record library, and acquiring real-time position data, cargo distribution demand data, traffic flow data and environmental factor data of the vehicles; Carrying out distributed cleaning and standardization processing on the acquired data, eliminating data abnormality through an outlier detection algorithm, and integrating multi-source data into a unified vehicle-environment state feature vector by adopting a data fusion technology, wherein the vehicle-environment state feature vector comprises vehicle state features, environment features and position features, the vehicle state features comprise vehicle load rate, running speed and residual fuel quantity, the environment features comprise traffic jam indexes and weather condition indexes, and the position features comprise real-time longitude and latitude coordinates and road section identifications; Based on the vehicle-environment state feature vector, constructing a vehicle dispatching decision model by utilizing a deep reinforcement learning framework, generating candidate strategies of a vehicle path and a dispatching sequence through a strategy network, and optimally selecting the candidate strategies by adopting a multi-objective optimization algorithm, wherein the multi-objective optimization algorithm simultaneously considers transportation cost, time efficiency and resource utilization rate; encoding the scheduling sequence into an executable instruction set, and transmitting the executable instruction set to a vehicle-mounted control unit of a corresponding logistics vehicle through a wireless communication network to drive the vehicle to execute a scheduling task; And monitoring vehicle position data, task execution state and external environment change in real time, collecting feedback data, and updating parameters of a vehicle scheduling decision model based on time difference errors to realize closed-loop dynamic optimization.
2. The intelligent logistics vehicle scheduling method based on big data as set forth in claim 1, wherein the outlier detection algorithm specifically includes: Processing a time sequence data stream by adopting a local outlier factor algorithm based on dynamic time regularity improvement, wherein the time sequence data stream comprises vehicle real-time position data and traffic stream data; Dividing the data stream into continuous segments by a local outlier factor algorithm through a sliding window, and calculating local density deviation of each data point in the segments by adopting a k-nearest neighbor algorithm; initializing a detection threshold according to the distribution characteristics of the historical data, and carrying out self-adaptive updating through an exponentially weighted dynamic mechanism; and (3) for the identified outliers, performing data restoration by adopting a probability interpolation method based on a Gaussian mixture model, and maintaining the integrity of the data sequence.
3. The intelligent logistics vehicle scheduling method based on big data as set forth in claim 2, wherein the data fusion technique comprises: carrying out normalization processing on the data after the data abnormality is eliminated, and eliminating dimension differences; dividing the normalized data into time sequence dynamic characteristic data and static characteristic data; The time sequence dynamic characteristic data comprises vehicle real-time position data, traffic flow data and dynamic environment characteristics derived from environment factor data, and the static characteristic data comprises vehicle inherent attribute data and cargo fixing information; Aiming at time sequence dynamic characteristic data, a multi-head self-attention network is input, and the weight coefficient of each characteristic in the fusion process is dynamically distributed by calculating the correlation score among the characteristics, wherein the multi-head self-attention network integrated position encoder reserves the time sequence attribute of the data; aiming at static characteristic data, a fully-connected neural network is adopted to perform characteristic transformation and fusion; the processed time sequence dynamic characteristics and static characteristics are spliced and compressed into a unified vehicle-environment state characteristic vector.
4. The logistics vehicle intelligent scheduling method based on big data according to claim 1, wherein the vehicle scheduling decision model adopts a framework based on deep reinforcement learning, and comprises a feature extraction network, a strategy network and a value network; The feature extraction network is used for extracting space-time feature vectors from vehicle-environment state feature vectors, and is formed by connecting a one-dimensional convolutional neural network and a long-term memory network in series, wherein the one-dimensional convolutional neural network is used for extracting local space modes of the vehicle-environment state feature vectors, and the long-term memory network is used for capturing time dependence of features; the strategy network is used for generating a vehicle dispatching strategy, and the vehicle dispatching strategy comprises a vehicle path and a dispatching sequence; The value network is used for evaluating long-term benefits of the scheduling strategy; The multi-objective optimization algorithm adopts an improved non-dominant ordering genetic algorithm, maintains a pareto optimal solution set in each generation of evolution, maintains the diversity of solutions by calculating the crowdedness of the solutions, and finally selects a scheduling sequence with the highest comprehensive utility from the solution set as output.
5. The intelligent logistics vehicle scheduling method based on big data as set forth in claim 4, wherein said strategy network adopts encoder-decoder architecture, and its input layer receives the space-time feature vector extracted by the feature extraction network; The encoder is composed of a plurality of full-connection layers, and a batch normalization layer and a ReLU activation function are arranged behind each full-connection layer; the decoder adopts a circulating neural network structure based on an attention mechanism to gradually generate a vehicle path sequence and a scheduling sequence; And the output layer of the strategy network uses a Softmax function to respectively output the probability distribution of vehicle path selection at each decision moment and the probability distribution of vehicle scheduling sequence arrangement.
6. The intelligent logistics vehicle scheduling method based on big data as set forth in claim 5, wherein the inputs of the value network include the space-time feature vector extracted by the feature extraction network and the vehicle scheduling policy outputted by the policy network, and the output is a long-term value estimation scalar for the current state-action pair; the value network comprises a plurality of fully connected layers for realizing nonlinear mapping from input features to value estimation; The value network is trained through a time difference learning algorithm, and the mean square error between the predicted value and the target value is minimized.
7. The intelligent logistics vehicular scheduling method based on big data as set forth in claim 6, wherein the improved non-dominant ordering genetic algorithm comprises the following steps: Initializing a population, performing non-dominant sorting on individuals of the population, and calculating crowding degree; Making a tournament selection based on the leading edge level and the crowdedness of the individual; generating a child population by adopting simulated binary crossover and polynomial mutation operation; The cross probability and the variation probability are adaptively adjusted according to the diversity index of the current generation of the population; after each generation of evolution, applying local search operation to individuals in the pareto optimal solution set, wherein the local search adopts disturbance operators related to a problem domain; and the algorithm calculates the comprehensive utility value of each solution from the pareto optimal solution set of the final population according to a preset preference rule or weight coefficient, and selects a scheduling sequence with the highest utility value as output.
8. The intelligent logistics vehicular scheduling method of claim 1, wherein said encoding a scheduling sequence into an executable instruction set comprises: according to a preset dispatching operation specification library, resolving a dispatching sequence into a series of atomic operation instructions, wherein each atomic operation instruction corresponds to a basic action executable by a vehicle; According to a standard communication protocol of the vehicle-mounted control unit, the atomic operation instruction is serialized into a structured data message, and the data message is packaged in a JSON format and is added with a time stamp and a vehicle identity; and compressing the data message by using a data compression algorithm, and ensuring the transmission reliability by adding a forward error correction code to form a final executable instruction set.
9. The intelligent logistics vehicle scheduling method based on big data as set forth in claim 8, wherein the closed loop dynamic optimization process specifically includes: after receiving and analyzing the executable instruction set, the vehicle-mounted control unit drives the vehicle to execute corresponding actions; Meanwhile, continuously collecting vehicle positioning data, engine state parameters and real-time traffic information as feedback data through a vehicle-mounted sensor; And the feedback data is transmitted back to the central processing system, an online learning process of the model is triggered based on the time difference error of the actual reward signal and the value network predicted value, and the strategy network and the value network parameters are updated in an increment mode by using a reinforcement learning algorithm, so that closed-loop optimization of the scheduling strategy is realized.
10. A big data based logistics vehicular intelligent scheduling system, characterized in that the system is used for implementing a big data based logistics vehicular intelligent scheduling method as set forth in any one of claims 1 to 9, comprising: the data acquisition module acquires data related to logistics vehicle dispatching in real time from a plurality of heterogeneous data sources; the data preprocessing and fusion module is used for cleaning, standardizing and outlier detection of the collected data, and performing multi-source data fusion to generate the vehicle-environment state feature vector; The scheduling decision calculation module is used for running a vehicle scheduling decision model and a multi-objective optimization algorithm to generate a scheduling sequence; the instruction coding and communication module codes the scheduling sequence into an executable instruction set and transmits the executable instruction set through a wireless communication network; And the monitoring and closed-loop feedback module is used for monitoring the vehicle execution state, collecting feedback data and driving the dynamic update of the scheduling decision model.

Description

Logistics vehicle intelligent scheduling method and system based on big data Technical Field The invention relates to the technical field of logistics big data, in particular to an intelligent logistics vehicle scheduling method and system based on big data. Background Traditional logistics scheduling relies on manual experience or simple rules, and it is difficult to integrate multi-source heterogeneous data such as vehicle states, traffic environments, cargo demands and the like in real time, so that path planning is unreasonable and resource allocation is unbalanced. The existing method mostly adopts a static model, cannot dynamically adapt to sudden factors such as traffic jams, weather changes and the like, and is high in transportation cost and poor in timeliness. While the basic data analysis is introduced into part of the system, the system lacks deep learning and multi-objective optimization capability, and is difficult to realize global optimization among cost, efficiency and resource utilization rate. With the expansion of the logistics scale, the traditional method cannot meet the intelligent scheduling requirement in a complex scene. Disclosure of Invention Aiming at the technical defects, the invention aims to provide a logistics vehicle intelligent scheduling method and system based on big data, which solve the problems of low efficiency, poor resource utilization rate and weak adaptability of the existing logistics scheduling method. In order to solve the technical problems, the invention adopts the following technical scheme: In a first aspect, the invention provides a logistics vehicle intelligent scheduling method based on big data, which comprises the following steps: Acquiring data related to logistics vehicle dispatching in real time from a plurality of heterogeneous data sources, wherein the data sources comprise vehicle-mounted sensors, a traffic monitoring platform, a cargo management database and a historical operation record library, and acquiring real-time position data, cargo distribution demand data, traffic flow data and environmental factor data of the vehicles; Carrying out distributed cleaning and standardization processing on the acquired data, eliminating data abnormality through an outlier detection algorithm, and integrating multi-source data into a unified vehicle-environment state feature vector by adopting a data fusion technology, wherein the vehicle-environment state feature vector comprises vehicle state features, environment features and position features, the vehicle state features comprise vehicle load rate, running speed and residual fuel quantity, the environment features comprise traffic jam indexes and weather condition indexes, and the position features comprise real-time longitude and latitude coordinates and road section identifications; Based on the vehicle-environment state feature vector, constructing a vehicle dispatching decision model by utilizing a deep reinforcement learning framework, generating candidate strategies of a vehicle path and a dispatching sequence through a strategy network, and optimally selecting the candidate strategies by adopting a multi-objective optimization algorithm, wherein the multi-objective optimization algorithm simultaneously considers transportation cost, time efficiency and resource utilization rate; encoding the scheduling sequence into an executable instruction set, and transmitting the executable instruction set to a vehicle-mounted control unit of a corresponding logistics vehicle through a wireless communication network to drive the vehicle to execute a scheduling task; And monitoring vehicle position data, task execution state and external environment change in real time, collecting feedback data, and updating parameters of a vehicle scheduling decision model based on time difference errors to realize closed-loop dynamic optimization. In a possible implementation manner of the first aspect, the outlier detection algorithm is specifically: Processing a time sequence data stream by adopting a local outlier factor algorithm based on dynamic time regularity improvement, wherein the time sequence data stream comprises vehicle real-time position data and traffic stream data; Dividing the data stream into continuous segments by a local outlier factor algorithm through a sliding window, and calculating local density deviation of each data point in the segments by adopting a k-nearest neighbor algorithm; initializing a detection threshold according to the distribution characteristics of the historical data, and carrying out self-adaptive updating through an exponentially weighted dynamic mechanism; and (3) for the identified outliers, performing data restoration by adopting a probability interpolation method based on a Gaussian mixture model, and maintaining the integrity of the data sequence. In a possible implementation manner of the first aspect, the data fusion technique includes: carrying out normalization process