Logistics distribution route planning system and method based on artificial intelligence
Through the logistics distribution route planning method based on artificial intelligence, using reinforcement learning and Internet of Things technology, real-time adjustment of distribution routes and optimize vehicle scheduling, the problem of low logistics distribution efficiency is solved and more efficient distribution services are achieved.
Patent Information
- Application Number
- CN202411580486.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-07
- Publication Date
- 2025-02-14
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In the prior art, logistics distribution efficiency is low, users need to deliver multiple times when they are not at home, goods are out of stock, and large traffic flow and special weather, resulting in difficult delivery.
Adopt the logistics distribution route planning method based on artificial intelligence, through reinforcement learning and training of AI models, combining the Internet of Things and 5G technology to obtain dynamic information such as traffic, roads, weather in real time, adjust distribution routes, optimize vehicle and personnel scheduling, and improve distribution efficiency.
By adjusting the distribution routes in real time and optimizing vehicle and personnel scheduling, logistics distribution efficiency is improved, delivery delay is reduced, and user experience is improved.
Smart Images
Figure CN119444037A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of intelligent route planning, and in particular to a logistics distribution route planning method based on artificial intelligence. Background Art
[0002] Artificial Intelligence (AI) refers to the technology that simulates human intelligence through computers and algorithms, involving perception, learning, reasoning, decision-making and other capabilities. AI technology is currently widely used in many fields, such as automation, medical diagnosis, speech recognition, natural language processing, image recognition and autonomous driving. The progress of AI is driving innovation in various industries, such as medicine, manufacturing, finance, etc., especially combined with the power of big data and cloud computing, AI can more effectively solve complex problems and improve efficiency.
[0003] Currently, with the development of technology, the online shopping experience is more comprehensive. However, as more and more people start shopping online, the efficiency of logistics and delivery has become an important issue. The user is not at home when the goods are delivered, and multiple deliveries are required. The goods purchased by the user are out of stock in the warehouse due to high demand, and there are problems such as heavy traffic on special roads and special weather encountered during the delivery process. These have become huge troubles for the current logistics industry. Summary of the invention
[0004] The purpose of the present invention is to provide a logistics distribution route planning method based on artificial intelligence to solve the problems raised in the prior art.
[0005] To achieve the above object, the present invention provides the following technical solution: a logistics distribution route planning method based on artificial intelligence, the logistics distribution route planning method comprising the following steps: Step S1: According to the user's order situation, the order information, customer address, warehouse location, real-time traffic information and weather conditions of different order sources are collected; the order information records the user's order time and the type of purchased goods according to the time period; the customer address is the address to which the user needs to deliver the ordered goods, and data processing is performed; Step S2: Use reinforcement learning to train the AI model based on dynamically changing traffic conditions, road conditions, warehouse materials, vehicles, and personnel, and use the Internet of Things and 5G technology to obtain dynamic information on traffic, roads, weather, materials, vehicles, and personnel in real time, and transfer these data to the AI model in real time. The AI model adjusts the delivery route based on the real-time traffic conditions, road conditions, warehouse materials, vehicles, and personnel; Step S3: The AI model continuously self-learns and optimizes based on historical logistics distribution data and actual operation results, updates logistics strategies, and plans distribution routes and schedules. The AI model automatically allocates vehicles and personnel to optimize the material arrangement of the warehouse; In step S1, logistics distribution data is collected from multiple sources, and the specific steps for data processing are: Step S1-1, according to the user's order situation, collect order information, customer address, warehouse location, real-time traffic information and weather conditions from different order sources; the order information records the user's order time and the type of purchased goods according to the time period; the customer address is the address to which the user needs to deliver the ordered goods; Step S1-2, process the collected data, summarize the data according to the ordering users, form user purchase characteristics based on the time period, analyze the user addresses according to the same province, city and district, and update the real-time traffic information and weather conditions.
[0006] In step S2, the specific steps of training the AI model through reinforcement learning are: Step S2-1: construct a logistics state space based on the logistics distribution data that has been processed. The set of the logistics state space is: S t1 ={S t1 _car,S t1 _tri,S t1 _road,S t1 _re,S t1 _co,S t1 _po}; where S t1 Indicates the logistics status at t1; S t1 _car, S t1 _tri, S t1 _road, S t1 _re, S t1 _co and S t1 _po are the logistics state space set S t1 The elements in; t1 represents the time to obtain the logistics status; S t1 _car represents the position of the vehicle at time t1; S t1 _tri represents the traffic flow at time t1; S t1 _road represents the road traffic status at time t1; S t1 _re represents the delivery target that has not been completed at time t1; S t1 _co represents the remaining amount of materials stored at time t1; S t1 _po represents the vehicles and personnel available for delivery at time t1; Step S2-2: construct a logistics action space, the set of which is: A t2 ={A t2 _sec,A t2 _v,A t2 _road,A t2 _rec}; where A t2 Indicates the logistics action at t2; A t2 _sec, A t2 _v, A t2 _road and A t2 _rec is the logistics action space set A t2 The elements in represent the actions performed by the system; t2 represents the time for performing logistics actions; A t2 _sec indicates the choice of the next section of the road; A t2 _v means adjusting the current vehicle's speed; A t2 _rec means re-planning according to traffic; Step S2-3, designing a state transfer function, wherein the expression of the state transfer function is specifically: ;in, represents the probability of transitioning to the next state s' after executing action a in state s; p represents the probability operator; S t1 Indicates the logistics status at t1; t1 indicates the time when the logistics status is obtained; Indicates the logistics status at t1+1; A t2 represents the logistics action at t2; Step S2-4: Design a reward function and a selection strategy. The expression of the reward function is specifically: ; Where R(s,a) represents the immediate reward obtained by the system after performing action a in state s; L(s,a) represents the total length of the path selected by the system after performing action a in state s; T(s,a) represents the expected driving time in traffic congestion after the system performs action a in state s; G(s,a) represents the expected fuel consumption on the route after the system performs action a in state s; Respectively represent the influence weights of L(s,a), T(s,a) and G(s,a); the system decides the action to take in each state by selecting a strategy, the goal is to maximize the accumulated reward; the selection strategy is as follows: ;in, represents the probability of choosing a in state s; S t1 Indicates the logistics status at t1; A t2 Represents the logistics action at t2.
[0007] In step S2-3 and step S2-4, specifically: Through the state transfer function, the state value function is designed, and the expression of the state value function is specifically: ;in, Indicates that the strategy is followed in state s The expected value when ; E represents the expectation operator; represents the state s at time t t Next, perform action a t Rewards received after t represents the state at time t; a t represents the execution action at time t; Indicated in strategy The expected value under the condition of S0 represents the initial logistics state; Represents the discount factor, the value range is ; Design the action value function, specifically: ;in, Indicates usage strategy The expected return of selecting action a in state s; E represents the expected operator; A0 represents the initial logistics action; Represents the discount factor, the value range is ; Through the action value function, the value of each action is evaluated, so as to select the action with the largest Q value and find the corresponding update strategy , specifically: ;in, represents the action selected to obtain the maximum reward in state s, that is, the updated strategy; It represents the maximum expected reward that the system can obtain by following the update strategy after executing action a in state s.
[0008] In step S2, the Internet of Things and 5G technologies are used to obtain dynamic information on traffic, roads, weather, materials, vehicles and personnel, and these data are transmitted to the AI model in real time. The AI model adjusts the delivery route according to traffic conditions, road conditions, storage materials, vehicles and personnel as follows: Step S2-1-1, use weather forecast and navigation software to obtain traffic, road and weather data, record the dynamic information of materials, vehicles and personnel every time through the Internet of Things, and pass the traffic, road, weather and material data into the AI model for route planning; Step S2-1-2, the AI model calculates the state value function and the action value function, uses the Q-learning algorithm, and plans the route by updating the Q value. The specific expression of the Q-learning algorithm is: ; Where Q(s,a) is the expected return of executing action a in state s, which represents the expected return; Represents the learning rate, ranging from [0,1]; Represents the discount factor, the value range is ; R(s,a) represents the immediate reward obtained by the system after executing action a in state s; It represents the action with the maximum Q value among all possible actions a' in the next state s', and s' represents the next state after state s completes action a; Step S2-1-3: After obtaining the planned route by updating the Q value, the delivery vehicle and personnel corresponding to the state s perform delivery according to the planned route.
[0009] In step S3, the method of continuously self-learning and optimizing based on historical logistics distribution data and actual operation results to improve the accuracy of route planning is as follows: Design a gradient optimization function to optimize the strategy, the expression is: ; Indicates the strategy optimization adjustment parameters; Represents the learning rate, ranging from [0,1]; Indicates the parameters The gradient obtained after differentiation; is the expected return of the strategy, which means that under the current strategy parameters The expected cumulative reward under Specifically: ;in, represents the state s at time t t Next, perform action a t Rewards received after t represents the state at time t; a t represents the execution action at time t; Indicated in strategy The adjustment parameters are Expected value; Represents the discount factor, the value range is .
[0010] Step S4: Equip the delivery vehicle with a positioning device and use the GPS positioning system to achieve real-time positioning of the delivery vehicle. By embedding maps in mobile applications and web interfaces, users can see the real-time location and movement trajectory of the delivery vehicle and understand the transportation dynamics of the package. By predicting the arrival time based on the historical average logistics transportation time, users can view the delivery status in real time through the mobile application and continuously optimize the delivery service based on the logistics distribution data.
[0011] A logistics distribution route planning system based on artificial intelligence, the logistics distribution route planning system includes a data acquisition and processing module, a state and action space construction module, a reinforcement learning module, a vehicle and personnel scheduling module, a user interface module and a self-optimization module; The data collection and processing module is used to collect and process logistics and delivery data from different sources; the module receives inputs such as user orders, customer addresses, warehouse locations, real-time traffic information and weather conditions, and the processed logistics and delivery data serves as input for subsequent modules; The state and action space construction module is used to construct the logistics state space and action space according to the logistics data; the logistics state space includes vehicle position, traffic flow, road circulation status and unfinished delivery targets; the action space includes road selection, speed adjustment, and traffic re-planning execution actions; the constructed state and action space provides input for the subsequent reinforcement learning module; The reinforcement learning module uses reinforcement learning algorithms to perform dynamic route planning and real-time optimization based on state and action space; updates Q values based on state transfer functions and reward functions to find routes that maximize cumulative returns; implements real-time route adjustments, considers dynamic changes such as traffic flow and road conditions, and dynamically optimizes delivery routes to provide input for the vehicle scheduling module; The vehicle and personnel scheduling module automatically schedules vehicles and personnel according to the optimized route output by the reinforcement learning module; The warehouse management module is used to optimize the arrangement and management of warehouse materials by analyzing order volume and route planning; dynamically adjust the storage location of materials in the warehouse, generate replenishment orders in advance, ensure sufficient inventory, and reduce delivery delays; The user interface module is used to provide users with an interface for tracking and querying logistics status, including mobile applications and web pages; real-time display of delivery routes, order status, estimated arrival time and other information; users can view the delivery progress of orders through the interface and receive relevant notifications; The self-optimization module is based on historical logistics data and actual operation results. It continuously improves the accuracy of route planning through the gradient optimization algorithm, continuously updates the parameters in the reinforcement learning strategy, improves the self-learning ability of the model, optimizes according to actual delivery results and feedback, and gradually improves delivery efficiency.
[0012] The data collection and processing module includes an order data collection unit, a traffic and weather data collection unit, a data cleaning unit and a data analysis unit; The order data collection unit is used to obtain order data from the user's order system, including the user's order time, purchased goods, and customer address information; The traffic and weather data collection unit is used to collect external data such as traffic information, road conditions, weather conditions, etc. in real time as a basis for logistics decision-making; The data cleaning unit is used to clean and format the collected raw data, remove noise and redundant data, and ensure data quality; The data analysis unit is used to analyze user purchasing behavior, warehouse location and order distribution, generate purchasing characteristics and prediction information, and provide support for subsequent modules.
[0013] The state and action space construction module includes a state space definition unit, an action space definition unit and a state and action mapping unit; The state space definition unit is used to construct a logistics state space based on logistics data, including vehicle positions, traffic flow, road conditions and unfinished delivery targets, and define state variables; The action space definition unit is used to define possible actions in the logistics system, including path selection, speed adjustment and re-planning, and provides an optional action set for reinforcement learning; The state and action mapping unit is used to map the current state to the corresponding action, ensuring that the system finds the execution action according to the current logistics state.
[0014] The self-optimization module includes a policy gradient optimization unit, a historical data analysis unit and a model updating unit; The policy gradient optimization unit is used to adjust system parameters and optimize the distribution strategy through the policy gradient optimization algorithm, so that the system can self-learn and improve performance; The historical data analysis unit is used to analyze historical delivery data, evaluate the effectiveness of previous logistics strategies, and improve future route planning; The model updating unit is used to periodically update the reinforcement learning model.
[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention collects user order data, analyzes the data, obtains the user's purchase characteristics, and through periodic analysis, delivers the predicted goods to the nearest warehouse in advance when the user's purchase frequency is high; updates traffic and weather information in real time, and uses reinforcement learning methods to train the logistics route planning model. The delivery route takes into account traffic flow, vehicle speed and weather conditions, and delivers goods in the same area at the same time, which greatly improves the delivery efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 A schematic diagram of the steps of a logistics distribution route planning method based on artificial intelligence of the present invention; Figure 2 This is a structural schematic diagram of an artificial intelligence-based logistics distribution route planning system of the present invention. DETAILED DESCRIPTION
[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0018] Example: Figure 1-Figure 2 As shown, the present invention provides a technical solution, a logistics distribution route planning method based on artificial intelligence, such as Figure 1 As shown, the logistics distribution route planning method includes the following steps: A logistics company needs to deliver goods from warehouse A to three delivery points B, C, and D in different areas. The following data is known: Order requirements: Orders come from delivery points B, C, and D, requiring the delivery of 100, 150, and 200 packages respectively.
[0019] Vehicle situation: There are two delivery vehicles, each capable of carrying 300 packages, and the starting position is warehouse A.
[0020] Traffic and road information: The distance from A to B is 20 kilometers. The current traffic is congested and the estimated travel time is 40 minutes. The distance from A to C is 30 kilometers. If the road is clear, the estimated driving time is 30 minutes. The distance from A to D is 25 kilometers, with slight congestion, and the estimated driving time is 35 minutes; Real-time weather information: The weather is sunny and suitable for delivery.
[0021] Step S1: According to the user's order situation, the order information, customer address, warehouse location, real-time traffic information and weather conditions of different order sources are collected; the order information records the user's order time and the type of purchased goods according to the time period; the customer address is the address to which the user needs to deliver the ordered goods, and data processing is performed; Step S1-1, according to the user's order situation, collect order information, customer address, warehouse location, real-time traffic information and weather conditions from different order sources; the order information records the user's order time and the type of purchased goods according to the time period; the customer address is the address to which the user needs to deliver the ordered goods; Step S1-2, process the collected data, summarize the data according to the ordering users, form user purchase characteristics based on the time period, analyze the user addresses according to the same province, city and district, and update the real-time traffic information and weather conditions.
[0022] Logistics order data: According to order requirements, the order distribution is delivery point B (100 packages), delivery point C (150 packages), and delivery point D (200 packages); Traffic data: Based on real-time traffic information, traffic from A to B is congested, the road from A to C is unobstructed, and traffic from A to D is slightly congested; Weather data: Good weather, no serious impacts.
[0023] Step S2: Use reinforcement learning to train the AI model based on dynamically changing traffic conditions, road conditions, warehouse materials, vehicles, and personnel, and use the Internet of Things and 5G technology to obtain dynamic information on traffic, roads, weather, materials, vehicles, and personnel in real time, and transfer these data to the AI model in real time. The AI model adjusts the delivery route based on the real-time traffic conditions, road conditions, warehouse materials, vehicles, and personnel; In step S2, the specific steps of reinforcement learning are: Step S2-1: construct a logistics state space based on the logistics distribution data that has been processed. The set of the logistics state space is: S t1 ={S t1 _car,S t1 _tri,S t1 _road,S t1 _re,S t1 _co,S t1 _po}; where S t1 Indicates the logistics status at t1; S t1 _car, S t1 _tri, S t1 _road, S t1 _re, S t1 _co and S t1 _po are the logistics state space set S t1 The elements in; t1 represents the time to obtain the logistics status; S t1 _car represents the position of the vehicle at time t1; S t1 _tri represents the traffic flow at time t1; S t1 _road represents the road traffic status at time t1; S t1 _re represents the delivery target that has not been completed at time t1; S t1 _co represents the remaining amount of materials stored at time t1; S t1 _po represents the vehicles and personnel available for delivery at time t1; Step S2-2: construct a logistics action space, the set of which is: At2 ={A t2 _sec,A t2 _v,A t2 _road,A t2 _rec}; where A t2 Indicates the logistics action at t2; A t2 _sec, A t2 _v, A t2 _road and A t2 _rec is the logistics action space set A t2 The elements in represent the actions performed by the system; t2 represents the time for performing logistics actions; A t2 _sec indicates the choice of the next section of the road; A t2 _v means adjusting the current vehicle's speed; A t2 _rec means re-planning according to traffic; Step S2-3, designing a state transfer function, wherein the expression of the state transfer function is specifically: ;in, represents the probability of transitioning to the next state s' after executing action a in state s; p represents the probability operator; S t1 Indicates the logistics status at t1; t1 indicates the time when the logistics status is obtained; Indicates the logistics status at t1+1; A t2 represents the logistics action at t2; Step S2-4: Design a reward function and a selection strategy. The expression of the reward function is specifically: ; Where R(s,a) represents the immediate reward obtained by the system after performing action a in state s; L(s,a) represents the total length of the path selected by the system after performing action a in state s; T(s,a) represents the expected driving time in traffic congestion after the system performs action a in state s; G(s,a) represents the expected fuel consumption on the route after the system performs action a in state s; Respectively represent the influence weights of L(s,a), T(s,a) and G(s,a); the system decides the action to take in each state by selecting a strategy, the goal is to maximize the accumulated reward; the selection strategy is as follows: ;in, represents the probability of choosing a in state s; S t1 Indicates the logistics status at t1; A t2 Represents the logistics action at t2.
[0024] =1, =1; =1; A to B, L = 20 km, T = 40 min, G = 8 L; R(A,B)=-(20+40+8)=-68; A to C, L = 30 km, T = 30 min, G = 9 L; R(A,C)=-(30+30+9)=-69; A to D, L = 25 km, T = 35 min, G = 10 L; R(A,C)=-(25+35+10)=-70; In step S2-3 and step S2-4, specifically: Through the state transfer function, the state value function is designed, and the expression of the state value function is specifically: ;in, Indicates that the strategy is followed in state s The expected value when ; E represents the expectation operator; represents the state s at time t t Next, perform action a t Rewards received after t represents the state at time t; a t represents the execution action at time t; Indicated in strategy The expected value under the condition of S0 represents the initial logistics state; Represents the discount factor, the value range is ; Design the action value function, specifically: ;in, Indicates usage strategy The expected return of selecting action a in state s; E represents the expected operator; A0 represents the initial logistics action; Represents the discount factor, the value range is ; Through the action value function, the value of each action is evaluated, so as to select the action with the largest Q value and find the corresponding update strategy , specifically: ;in, represents the action selected to obtain the maximum reward in state s, that is, the updated strategy; It represents the maximum expected reward that the system can obtain by following the update strategy after executing action a in state s.
[0025] In step S2, the Internet of Things and 5G technologies are used to obtain dynamic information on traffic, roads, weather, materials, vehicles and personnel, and these data are transmitted to the AI model in real time. The AI model adjusts the delivery route according to traffic conditions, road conditions, storage materials, vehicles and personnel as follows: Step S2-1-1, use weather forecast and navigation software to obtain traffic, road and weather data, record the dynamic information of materials, vehicles and personnel every time through the Internet of Things, and pass the traffic, road, weather and material data into the AI model for route planning; Step S2-1-2, the AI model calculates the state value function and the action value function, uses the Q-learning algorithm, and plans the route by updating the Q value. The specific expression of the Q-learning algorithm is: ; Where Q(s,a) is the expected return of executing action a in state s, which represents the expected return; Represents the learning rate, ranging from [0,1]; Represents the discount factor, the value range is ; R(s,a) represents the immediate reward obtained by the system after executing action a in state s; It represents the action with the maximum Q value among all possible actions a' in the next state s', and s' represents the next state after state s completes action a; Step S2-1-3: After obtaining the planned route by updating the Q value, the delivery vehicle and personnel corresponding to the state s perform delivery according to the planned route.
[0026] =0.1, =0.9; Q value update: ; Q(B,D)=-65; ; Step S3: The AI model continuously self-learns and optimizes based on historical logistics distribution data and actual operation results, updates logistics strategies, and plans distribution routes and schedules. The AI model automatically allocates vehicles and personnel to optimize the material arrangement of the warehouse; In step S3, the method of continuously self-learning and optimizing based on historical logistics distribution data and actual operation results to improve the accuracy of route planning is as follows: Design a gradient optimization function to optimize the strategy, the expression is: ; Indicates the strategy optimization adjustment parameters; Represents the learning rate, ranging from [0,1]; Indicates the parameters The gradient obtained after differentiation; is the expected return of the strategy, which means that under the current strategy parameters The expected cumulative reward under Specifically: ;in, represents the state s at time t t Next, perform action a t Rewards received after t represents the state at time t; a t represents the execution action at time t; Indicated in strategy The adjustment parameters are Expected value; Represents the discount factor, the value range is .
[0027] Step S4: Equip the delivery vehicle with a positioning device and use the GPS positioning system to achieve real-time positioning of the delivery vehicle. By embedding maps in mobile applications and web interfaces, users can see the real-time location and movement trajectory of the delivery vehicle and understand the transportation dynamics of the package. By predicting the arrival time based on the historical average logistics transportation time, users can view the delivery status in real time through the mobile application and continuously optimize the delivery service based on the logistics distribution data.
[0028] like Figure 2 As shown, a logistics distribution route planning system based on artificial intelligence, the logistics distribution route planning system includes a data acquisition and processing module, a state and action space construction module, a reinforcement learning module, a vehicle and personnel scheduling module, a user interface module and a self-optimization module; The data collection and processing module is used to collect and process logistics and delivery data from different sources; the module receives inputs such as user orders, customer addresses, warehouse locations, real-time traffic information and weather conditions, and the processed logistics and delivery data serves as input for subsequent modules; The state and action space construction module is used to construct the logistics state space and action space according to the logistics data; the logistics state space includes vehicle position, traffic flow, road circulation status and unfinished delivery targets; the action space includes road selection, speed adjustment, and traffic re-planning execution actions; the constructed state and action space provides input for the subsequent reinforcement learning module; The reinforcement learning module uses reinforcement learning algorithms to perform dynamic route planning and real-time optimization based on state and action space; updates Q values based on state transfer functions and reward functions to find routes that maximize cumulative returns; implements real-time route adjustments, considers dynamic changes such as traffic flow and road conditions, and dynamically optimizes delivery routes to provide input for the vehicle scheduling module; The vehicle and personnel scheduling module automatically schedules vehicles and personnel according to the optimized route output by the reinforcement learning module; The warehouse management module is used to optimize the arrangement and management of warehouse materials by analyzing order volume and route planning; dynamically adjust the storage location of materials in the warehouse, generate replenishment orders in advance, ensure sufficient inventory, and reduce delivery delays; The user interface module is used to provide users with an interface for tracking and querying logistics status, including mobile applications and web pages; real-time display of delivery routes, order status, estimated arrival time and other information; users can view the delivery progress of orders through the interface and receive relevant notifications; The self-optimization module is based on historical logistics data and actual operation results. It continuously improves the accuracy of route planning through the gradient optimization algorithm, continuously updates the parameters in the reinforcement learning strategy, improves the self-learning ability of the model, optimizes according to actual delivery results and feedback, and gradually improves delivery efficiency.
[0029] The data collection and processing module includes an order data collection unit, a traffic and weather data collection unit, a data cleaning unit and a data analysis unit; The order data collection unit is used to obtain order data from the user's order system, including the user's order time, purchased goods, and customer address information; The traffic and weather data collection unit is used to collect external data such as traffic information, road conditions, weather conditions, etc. in real time as a basis for logistics decision-making; The data cleaning unit is used to clean and format the collected raw data, remove noise and redundant data, and ensure data quality; The data analysis unit is used to analyze user purchasing behavior, warehouse location and order distribution, generate purchasing characteristics and prediction information, and provide support for subsequent modules.
[0030] The state and action space construction module includes a state space definition unit, an action space definition unit and a state and action mapping unit; The state space definition unit is used to construct a logistics state space based on logistics data, including vehicle positions, traffic flow, road conditions and unfinished delivery targets, and define state variables; The action space definition unit is used to define possible actions in the logistics system, including path selection, speed adjustment and re-planning, and provides an optional action set for reinforcement learning; The state and action mapping unit is used to map the current state to the corresponding action, ensuring that the system finds the execution action according to the current logistics state.
[0031] The self-optimization module includes a policy gradient optimization unit, a historical data analysis unit and a model updating unit; The policy gradient optimization unit is used to adjust system parameters and optimize the distribution strategy through the policy gradient optimization algorithm, so that the system can self-learn and improve performance; The historical data analysis unit is used to analyze historical delivery data, evaluate the effectiveness of previous logistics strategies, and improve future route planning; The model updating unit is used to periodically update the reinforcement learning model.
[0032] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.
Claims
1. A logistics distribution route planning method based on artificial intelligence, characterized by: The logistics distribution route planning method comprises the following steps: Step S1: According to the user's order situation, the order information, customer address, warehouse location, real-time traffic information and weather conditions of different order sources are collected; the order information records the user's order time and the type of purchased goods according to the time period; the customer address is the address to which the user needs to deliver the ordered goods, and data processing is performed; Step S2: Use reinforcement learning to train the AI model based on dynamically changing traffic conditions, road conditions, warehouse materials, vehicles, and personnel, use the Internet of Things and 5G technologies to obtain dynamic information on traffic, roads, weather, materials, vehicles, and personnel, and transfer these data to the AI model in real time. The AI model adjusts the delivery route based on traffic conditions, road conditions, warehouse materials, vehicles, and personnel; Step S3: self-learn and optimize based on historical logistics distribution data and actual operation results, update logistics strategies, and optimize AI models; Step S4: Equip the delivery vehicle with a positioning device and use the GPS positioning system to achieve real-time positioning of the delivery vehicle. By embedding maps in mobile applications and web interfaces, users can see the real-time location and movement trajectory of the delivery vehicle and understand the transportation dynamics of the package. By predicting the arrival time based on the historical average logistics transportation time, users can view the delivery status in real time through the mobile application.
2. The method for logistics distribution route planning based on artificial intelligence according to claim 1, characterized in that: In step S1, logistics distribution data is collected from multiple sources, and the specific steps for data processing are: Step S1-1, according to the user's order situation, collect order information, customer address, warehouse location, real-time traffic information and weather conditions from different order sources; The order information records the user's order time and the type of purchased goods according to the time period; The customer address is the address to which the goods need to be delivered when the user places an order; Step S1-2, process the collected data, summarize the data according to the ordering users, form user purchase characteristics based on the time period, analyze the user addresses according to the same province, city and district, and update the real-time traffic information and weather conditions.
3. The method for logistics distribution route planning based on artificial intelligence according to claim 2, characterized in that: In step S2, the specific steps of training the AI model through reinforcement learning are: Step S2-1: construct a logistics state space based on the logistics distribution data that has been processed. The set of the logistics state space is: S t1 ={S t1 _car,S t1 _tri,S t1 _road,S t1 _re,S t1 _co,S t1 _po}; where S t1 Indicates the logistics status at t1; S t1 _car, S t1 _tri, S t1 _road, S t1 _re, S t1 _co and S t1 _po are the logistics state space set S t1 The elements in; t1 represents the time to obtain the logistics status; S t1 _car represents the position of the vehicle at time t1; S t1 _tri represents the traffic flow at time t1; S t1 _road represents the road traffic status at time t1; S t1 _re represents the delivery target that has not been completed at time t1; S t1 _co represents the remaining amount of materials stored at time t1; S t1 _po represents the vehicles and personnel available for delivery at time t1; Step S2-2: construct a logistics action space, the set of which is: A t2 ={A t2 _sec,A t2 _v,A t2 _road,A t2 _rec}; where A t2 Indicates the logistics action at t2; A t2 _sec, A t2 _v, A t2 _road and A t2 _rec is the logistics action space set A t2 The elements in represent the actions performed by the system; t2 represents the time for performing logistics actions; A t2 _sec indicates the choice of the next section of the road; A t2 _v means adjusting the current vehicle's speed; A t2 _rec means re-planning according to traffic; Step S2-3, designing a state transfer function, wherein the expression of the state transfer function is specifically: ;in, represents the probability of transitioning to the next state s' after executing action a in state s; p represents the probability operator; S t1 Indicates the logistics status at t1; t1 indicates the time when the logistics status is obtained; Indicates the logistics status at t1+1; A t2 represents the logistics action at t2; Step S2-4: Design a reward function and a selection strategy. The expression of the reward function is specifically: ; Where R(s,a) represents the immediate reward obtained by the system after performing action a in state s; L(s,a) represents the total length of the path selected by the system after performing action a in state s; T(s,a) represents the expected driving time in traffic congestion after the system performs action a in state s; G(s,a) represents the expected fuel consumption on the route after the system performs action a in state s; Respectively represent the influence weights of L(s,a), T(s,a) and G(s,a); the system decides the action to take in each state by selecting a strategy, the goal is to maximize the accumulated reward; the selection strategy is as follows: ;in, represents the probability of choosing a in state s; S t1 Indicates the logistics status at t1; A t2 Represents the logistics action at t2.
4. The method for logistics distribution route planning based on artificial intelligence according to claim 3 is characterized in that: In step S2-3 and step S2-4, specifically: Through the state transfer function, the state value function is designed, and the expression of the state value function is specifically: ;in, Indicates following the strategy in state s The expected value when ; E represents the expectation operator; represents the state s at time t t Next, perform action a t Rewards received after t represents the state at time t; a t represents the execution action at time t; Indicated in strategy The expected value under the condition of S0 represents the initial logistics state; Represents the discount factor, the value range is ; Design the action value function, specifically: ;in, Indicates usage strategy The expected return of selecting action a in state s; E represents the expected operator; A0 represents the initial logistics action; Represents the discount factor, the value range is ; Through the action value function, the value of each action is evaluated, so as to select the action with the largest Q value and find the corresponding update strategy , specifically: ;in, represents the action selected to obtain the maximum reward in state s, that is, the updated strategy; It represents the maximum expected reward that the system can obtain by following the update strategy after executing action a in state s.
5. The method for logistics distribution route planning based on artificial intelligence according to claim 4 is characterized in that: In step S2, the Internet of Things and 5G technologies are used to obtain dynamic information on traffic, roads, weather, materials, vehicles and personnel, and these data are transmitted to the AI model in real time. The AI model adjusts the delivery route according to traffic conditions, road conditions, storage materials, vehicles and personnel as follows: Step S2-1-1, use weather forecast and navigation software to obtain traffic, road and weather data, record the dynamic information of materials, vehicles and personnel every time through the Internet of Things, and pass the traffic, road, weather and material data into the AI model for route planning; Step S2-1-2, the AI model calculates the state value function and the action value function, uses the Q-learning algorithm, and plans the route by updating the Q value. The specific expression of the Q-learning algorithm is: ; Where Q(s,a) is the expected return of executing action a in state s, which represents the expected return; Represents the learning rate, ranging from [0,1]; Represents the discount factor, the value range is ; R(s,a) represents the immediate reward obtained by the system after executing action a in state s; It represents the action with the maximum Q value among all possible actions a' in the next state s', and s' represents the next state after state s completes action a; Step S2-1-3: After obtaining the planned route by updating the Q value, the delivery vehicle and personnel corresponding to the state s perform delivery according to the planned route.
6. The method for logistics distribution route planning based on artificial intelligence according to claim 5 is characterized in that: In step S3, the method of self-learning and optimization based on historical logistics distribution data and actual operation results is as follows: Design a gradient optimization function to optimize the strategy, the expression is: ; Indicates the strategy optimization adjustment parameters; Represents the learning rate, ranging from [0,1]; Indicates the parameters The gradient obtained after differentiation; is the expected return of the strategy, which means that under the current strategy parameters The expected cumulative reward under Specifically: ;in, represents the state s at time t t Next, perform action a t Rewards received after t represents the state at time t; a t represents the execution action at time t; Indicated in strategy The adjustment parameters are Expected value; Represents the discount factor, the value range is .
7. A logistics distribution route planning system based on artificial intelligence, characterized in that: The logistics distribution route planning system includes a data collection and processing module, a state and action space construction module, a reinforcement learning module, a vehicle and personnel scheduling module, a warehouse management module, a user interface module and a self-optimization module; The data collection and processing module is used to collect and process logistics and delivery data from different sources; the module receives inputs such as user orders, customer addresses, warehouse locations, real-time traffic information and weather conditions, and the processed logistics and delivery data serves as input for subsequent modules; The state and action space construction module is used to construct the logistics state space and action space according to the logistics data; the logistics state space includes vehicle position, traffic flow, road circulation status and unfinished delivery targets; the action space includes road selection, speed adjustment, and traffic re-planning execution actions; the constructed state and action space provides input for the subsequent reinforcement learning module; The reinforcement learning module adopts reinforcement learning algorithm to perform dynamic route planning and real-time optimization according to the state and action space; updates Q value according to the state transfer function and reward function to find the route that maximizes the cumulative return; Realize real-time route adjustment, consider dynamic changes in traffic flow, road conditions, etc., dynamically optimize delivery routes, and provide input for the vehicle dispatch module; The vehicle and personnel scheduling module automatically schedules vehicles and personnel according to the optimized route output by the reinforcement learning module; The warehouse management module is used to optimize the arrangement and management of warehouse materials by analyzing order volume and route planning; dynamically adjust the storage location of materials in the warehouse, generate replenishment orders in advance, ensure sufficient inventory, and reduce delivery delays; The user interface module is used to provide an interface for users to track and query logistics status, including mobile applications and web pages; Real-time display of delivery routes, order status, estimated arrival time and other information; Users can check the delivery progress of their orders through the interface and receive relevant notifications; The self-optimization module is based on historical logistics data and actual operation results. It continuously improves the accuracy of route planning through the gradient optimization algorithm, continuously updates the parameters in the reinforcement learning strategy, improves the self-learning ability of the model, optimizes according to actual delivery results and feedback, and gradually improves delivery efficiency.
8. The artificial intelligence-based logistics distribution route planning system according to claim 7, characterized in that: The data collection and processing module includes an order data collection unit, a traffic and weather data collection unit, a data cleaning unit and a data analysis unit; The order data collection unit is used to obtain order data from the user's order system, including the user's order time, purchased goods, and customer address information; The traffic and weather data collection unit is used to collect external data such as traffic information, road conditions, weather conditions, etc. in real time as a basis for logistics decision-making; The data cleaning unit is used to clean and format the collected raw data, remove noise and redundant data, and ensure data quality; The data analysis unit is used to analyze user purchasing behavior, warehouse location and order distribution, generate purchasing characteristics and prediction information, and provide support for subsequent modules.
9. The artificial intelligence-based logistics distribution route planning system according to claim 8, characterized in that: The state and action space construction module includes a state space definition unit, an action space definition unit and a state and action mapping unit; The state space definition unit is used to construct a logistics state space based on logistics data, including vehicle locations, traffic flow, road conditions and unfinished delivery targets, and define state variables; The action space definition unit is used to define possible actions in the logistics system, including path selection, speed adjustment and re-planning, and provides an optional action set for reinforcement learning; The state and action mapping unit is used to map the current state to the corresponding action, ensuring that the system finds the execution action according to the current logistics state.
10. The artificial intelligence-based logistics distribution route planning system according to claim 9, characterized in that: The self-optimization module includes a policy gradient optimization unit, a historical data analysis unit and a model updating unit; The policy gradient optimization unit is used to adjust system parameters and optimize the distribution strategy through the policy gradient optimization algorithm, so that the system can self-learn and improve performance; The historical data analysis unit is used to analyze historical delivery data, evaluate the effectiveness of previous logistics strategies, and improve future route planning; The model updating unit is used to periodically update the reinforcement learning model.