IoT-based dynamic transportation path optimization method

Abstract

The invention discloses a dynamic transportation path optimization method based on IoT, and belongs to the technical field of intelligent transportation and transportation scheduling, and the method comprises the following steps: 1, collecting multi-source data; step 2, data preprocessing; step 3, path planning; step 4, dynamic event response and re-planning; and step 5, multi-vehicle cooperative scheduling optimization. By means of the IoT-based dynamic transportation path optimization method, all links of transportation path optimization can be covered, the real-time position and speed of a vehicle are concerned, multi-dimensional data such as instantaneous oil consumption, real-time load and road congestion are also included, a genetic algorithm and reinforcement learning are deeply fused to achieve path dynamic adjustment, and the dynamic transportation path optimization efficiency is improved. The path planning is more suitable for an actual transportation scene, and a more scientific decision basis is provided for transportation scheduling.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent transportation and transportation scheduling, more particularly to a dynamic transportation path optimization method based on IoT. BACKGROUND

[0002] Dynamic transportation path optimization refers to a decision-making process of dynamically adjusting transportation paths based on real-time traffic information, vehicle status and external environmental changes to achieve comprehensive optimization of efficiency, cost and timeliness. However, existing transportation path planning systems mostly use static path planning algorithms, which are difficult to cope with real-time traffic congestion, weather changes and vehicle emergencies, resulting in reduced transportation timeliness, increased fuel consumption and waste of vehicle resources. Therefore, it is urgent to introduce path planning methods with dynamic response and real-time optimization capabilities to adapt to the current complex and changing transportation environment.

[0003] The patent document with the authorization announcement number CN120218533A discloses a "transportation vehicle operation path and scheduling optimization method and system", which generates a path library and globally searches for a particle swarm search algorithm based on the path library. During the execution of the particle swarm search algorithm, a local search is cyclically executed based on the current global optimal solution, i.e., a new solution is generated using a destruction and repair strategy and conflict adjustment is performed, then the global optimal solution is updated based on the new solution, then the particle position and velocity are updated, and the particle swarm iterative search is performed again, i.e., global iteration is performed. After reaching the iteration number, the optimal solution is obtained, and a scheduling scheme is generated by decoding.

[0004] The patent document with the authorization announcement number CN113139774B discloses a "vehicle path optimization method for multi-warehouse transportation", which includes calculating the distance of customer nodes to all warehouses, assigning customer nodes to the nearest warehouse, after the assignment is completed, for each warehouse and its assigned customer nodes, using an artificial bee colony algorithm to find partial optimal solutions; merging the vehicle scheduling route set for a single warehouse node into a set to form a total vehicle scheduling route set for the multi-warehouse vehicle path problem, and optimizing the vehicle path.

[0005] While existing technologies can improve computational efficiency by reducing multi-warehouse problems to single-warehouse problems, avoid premature convergence by using artificial bee colony algorithms, or resolve spatiotemporal conflicts among multiple vehicles and task points through path libraries and particle swarm search algorithms combined with conflict adjustment mechanisms, thus optimizing scheduling schemes, most existing technologies rely on static data or preset paths and do not deeply integrate with real-time IoT data. The data fusion dimensions are limited, and they are only applicable to specific scenarios (multi-warehouse allocation, fixed task point scheduling), lacking dynamic replanning capabilities. They cannot quickly respond to real-time changes such as sudden traffic events and sudden changes in vehicle status during transportation. In addition, the optimization algorithms do not dynamically adjust parameters based on real-time road conditions, and road load balancing and global fuel consumption costs are not fully considered in multi-vehicle collaborative scheduling, resulting in insufficient real-time performance and scalability for large-scale fleet scheduling. Summary of the Invention

[0006] This invention mainly provides a dynamic transportation route optimization method based on IoT, which can solve the problems mentioned in the background.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a dynamic transportation route optimization method based on IoT, comprising: S1. Real-time collection of operational data of transport vehicles and road traffic data; S2. The collected data is preprocessed through edge computing nodes, and the preprocessed key data is uploaded to the cloud server. S3. An optimization model combining reinforcement learning and genetic algorithms is used to generate the optimal route scheme that maximizes transportation efficiency. S4. When traffic congestion or emergencies are detected, the transportation route is replanned in real time and route update instructions are issued to the transportation vehicles. S5. Coordinate and schedule multiple transport vehicles to optimize the overall operational efficiency of the transportation network.

[0008] Furthermore, in S1, the operation-related data includes the vehicle's real-time location, instantaneous speed, instantaneous fuel consumption, and real-time load, while the road traffic-related data includes road congestion levels, types of emergencies, and their impact range. The data is collected through vehicle-mounted GPS, speed sensors, fuel consumption sensors, load sensors, vehicle-mounted cameras, and roadside IoT devices.

[0009] Furthermore, in S2, the preprocessing includes compressing the data using the LZ77 lossless compression algorithm and detecting abnormal data using the 3σ principle; key data is uploaded via the MQTT protocol, and data encryption uses the AES-128 algorithm.

[0010] Furthermore, in S3, the optimization model adopts a hierarchical fusion mode of genetic algorithm and reinforcement learning. The model is generated by training through historical transportation data. The training process combines the characteristics of different transportation scenarios. When the model is running, the parameters are dynamically adjusted based on the key data collected in real time.

[0011] Furthermore, in S4, the triggering conditions for replanning include the road congestion level reaching a preset threshold, the impact range of a sudden event meeting a preset standard, abnormal vehicle fuel consumption, changes in vehicle load reaching a preset proportion, and actual travel time exceeding the planned time by a preset proportion; the route update command is issued through the MQTT protocol, and the vehicle navigation system synchronizes the updated optimal route in real time.

[0012] Furthermore, in S5, the constraints of collaborative scheduling include that the vehicle load does not exceed the maximum load, the real-time number of vehicles on the road does not exceed the maximum number of vehicles passing through, and the actual delivery time of the vehicle does not exceed the latest delivery time; a distributed task allocation algorithm is adopted in the scheduling process to take into account both road load balancing and matching of return trip goods.

[0013] Furthermore, the method supports real-time route optimization for large-scale transport fleets, adopting an edge computing + cloud collaborative architecture. The edge computing nodes have local emergency route planning capabilities, while the cloud server adopts a microservice architecture and a distributed database.

[0014] The beneficial effects of the IoT-based dynamic transportation route optimization method of this invention are as follows: This method covers all aspects of transportation route optimization, focusing not only on real-time vehicle location and speed but also incorporating multi-dimensional data such as instantaneous fuel consumption, real-time load, and road congestion. It deeply integrates genetic algorithms and reinforcement learning to achieve dynamic route adjustment, making route planning more aligned with actual transportation scenarios and providing a more scientific basis for transportation scheduling decisions. Furthermore, through edge computing and cloud collaborative architecture technologies, effective collaboration between vehicle terminals and cloud systems is achieved, breaking down data and computing power barriers. This enables the system to process large-scale vehicle route optimization tasks in real time, improving the efficiency and accuracy of transportation scheduling and enhancing the real-time performance and adaptability of the route optimization model. Simultaneously, through dynamic programming triggering and multi-vehicle collaborative scheduling technologies, it provides faster and more flexible route adjustment capabilities, enabling timely responses to traffic congestion and emergencies. Furthermore, based on real-time data and fusion algorithms, it can rapidly generate the overall optimal route in complex transportation environments, significantly improving the operational efficiency and resource utilization of the transportation network. Attached Figure Description

[0015] The present invention will now be described in further detail with reference to the accompanying drawings and specific implementation methods.

[0016] Figure 1 This is a schematic diagram of the method flow for the IoT-based dynamic transportation route optimization method of the present invention. Detailed Implementation

[0017] To make the technical solution of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] Example 1 like Figure 1 As shown, a technical solution is provided: a dynamic transportation route optimization method based on IoT, including: Step 1: Multi-source data acquisition Real-time collection of operational data related to transport vehicles and road traffic data; Specifically, operational data includes real-time vehicle location, instantaneous speed, instantaneous fuel consumption, and real-time load, while road traffic data includes road congestion levels, types of emergencies, and their impact range. The data is collected through vehicle-mounted GPS, speed sensors, fuel consumption sensors, load sensors, vehicle-mounted cameras, and roadside IoT devices.

[0019] The vehicle-mounted GPS continuously collects the vehicle's real-time latitude and longitude coordinates at a sampling frequency of 1Hz, with positioning accuracy controlled within ±5m, ensuring the real-time nature and accuracy of the location data. The speed sensor collects the vehicle's instantaneous driving speed with a measurement accuracy of ±0.1m / s, generating one valid speed data point every second and synchronously recording the speed change trend. The fuel consumption sensor collects instantaneous and cumulative fuel consumption data with an accuracy of ±0.01L / km, and correlates the vehicle's driving status with fuel consumption in real time. The load sensor captures changes in the vehicle's load in real time with a measurement accuracy of ±50kg, providing data support for subsequent fuel cost calculation and load balancing scheduling. Meanwhile, the vehicle-mounted camera continuously captures the road scene ahead and combines it with the actual road traffic conditions to identify traffic density and road occupancy ratio, thereby judging the degree of road congestion. It does not rely on additional third-party traffic data interfaces. The roadside IoT device and the vehicle-mounted data collection device establish a short-range communication link to push the specific situation of emergencies on the surrounding roads in real time, including key information such as the location of the accident, the scope of the construction section, and the area affected by severe weather. This allows vehicles to obtain the dynamics of road traffic in the entire area in a timely manner, breaking the limitation of the field of view of a single vehicle-mounted device.

[0020] In addition, during the data collection process, the data from each sensor are cross-checked. For example, changes in speed data will correspond to the vehicle's driving status. If abnormal data features are found, they will be initially marked for further screening in the subsequent preprocessing stage. At the same time, it is ensured that the collected data can fully reflect the actual situation of vehicle operation and road environment, avoid data loss caused by the failure of a single sensor, ensure the reliability and effectiveness of data collection, and provide high-quality input data for subsequent model optimization. Finally, through the collaborative linkage mechanism between devices, the scattered data collected by vehicle GPS, various sensors, vehicle cameras and roadside IoT devices are initially integrated to ensure that vehicle operation data corresponds one-to-one with road traffic data, forming a complete data collection chain, and providing comprehensive, coherent and actual transportation scenario-based data support for the subsequent preprocessing stage of edge nodes.

[0021] Step 2: Data Preprocessing The collected data is preprocessed by edge computing nodes, and the preprocessed key data is uploaded to the cloud server. Specifically, preprocessing includes compressing the data using the LZ77 lossless compression algorithm and detecting abnormal data using the 3σ principle; critical data is uploaded via the MQTT protocol, and data encryption uses the AES-128 algorithm.

[0022] The LZ77 lossless compression algorithm scans the collected continuous data sequence, identifies and marks recurring data segments, and reduces the data volume by recording the position and length of the recurring segments without changing the core information of the original data. It focuses on compressing continuous and close position data when the vehicle is driving smoothly and speed data under constant speed, which greatly reduces the communication bandwidth occupied during data transmission, while ensuring that the original data can be completely restored during subsequent data parsing without affecting the calculation accuracy of the path optimization model. At the same time, the 3σ principle is used to verify each item of the collected data. The validity of the data is judged in combination with the actual logic of vehicle operation. For example, when a speed data is significantly different from the speed data of the preceding and following time periods and there is no corresponding change in driving status to support it, it is judged as abnormal data. For abnormal values ​​in fuel consumption data that are significantly outside the normal driving range, the same principle is used to filter them. Then, invalid abnormal data is directly removed to prevent erroneous data from entering the subsequent calculation process and to ensure that the key data uploaded to the cloud can truly reflect the actual situation of vehicle operation and road environment. In addition, the MQTT protocol is used in the critical data upload stage. This protocol can adapt to the remote communication scenario between the vehicle equipment and the cloud server, support the efficient and low-latency transmission of data, and ensure the stability of data transmission even when the network signal fluctuates. In the data encryption process, the AES-128 algorithm is used to encrypt the pre-processed critical data throughout the entire process. From the moment the data leaves the edge computing node to the moment the cloud server receives and decrypts it, a complete encrypted transmission link is formed, which effectively prevents the data from being stolen or tampered with during transmission and ensures the security and privacy of transportation data. Finally, after completing data compression and anomaly detection, the edge computing nodes further filter key data, extracting only core vehicle operation data and key road status data closely related to path planning and collaborative scheduling. Non-critical data is stored locally on the edge nodes for easy traceability and verification later. At the same time, the key data after filtering is verified for integrity to ensure that no core information is missing before being uploaded to the cloud server through an encrypted transmission link. The entire preprocessing process is completed quickly on the edge nodes, avoiding transmission delays caused by directly uploading large amounts of raw data, and laying the foundation for real-time calculation of the subsequent cloud optimization model.

[0023] Step 3, Path Planning An optimization model combining reinforcement learning and genetic algorithms is used to generate the optimal route scheme that maximizes transportation efficiency. Specifically, the optimization model adopts a hierarchical fusion mode of genetic algorithm and reinforcement learning. The model is generated by training on historical transportation data. The training process incorporates the characteristics of different transportation scenarios. During the model's operation, the parameters are dynamically adjusted based on key data collected in real time.

[0024] The genetic algorithm is responsible for the initial global path planning. It first abstracts the entire transportation network as a weighted undirected graph containing a start point, destination, and waypoints / intersections. The weight of each road is calculated using a comprehensive cost formula, which takes into account physical distance, fuel cost, and travel time. Specifically: In the formula, For the roadside physical distance, For the fuel consumption cost of roadside vehicle e, Let e ​​be the travel time at roadside e. , , The weighting coefficients are calculated by combining the road fuel consumption cost with the vehicle's real-time load and speed to ensure that the weights accurately reflect the actual consumption during transportation, avoid path planning deviations caused by single-dimensional considerations, and provide a realistic calculation basis for searching the globally optimal path. Then, the genetic algorithm transforms different path schemes into computable chromosomes through path sequence encoding. For example, for a path from the starting point through multiple waypoints to the ending point, the node order is directly used as the chromosome encoding, and a fitness function is constructed simultaneously. In the formula, For overall cost, For each road edge e in the current transportation path, the lower the overall path cost, the higher the fitness function value. This serves as the core basis for selecting the optimal path. During the iteration process, the roulette wheel selection method is used to select chromosomes with high fitness. The gene recombination of different path schemes is achieved through partial matching crossover. Mutation operations are performed by randomly exchanging non-starting / ending nodes. The path scheme is continuously optimized until the fitness value fluctuations stabilize after several generations. The iteration stops and the global initial optimal path is output. Reinforcement learning is responsible for real-time fine-tuning of the initial optimal path. First, a complete state space is defined, encompassing key information such as the vehicle's current location, surrounding road congestion, remaining fuel consumption and load, and waypoints on incomplete routes. This ensures a comprehensive understanding of dynamic changes during transportation. Then, an action space is defined, including options such as maintaining the original path and switching to multiple preset alternative paths. Simultaneously, a reward function is constructed. In the formula, This refers to the additional passage time added after the action is executed. To increase fuel costs, This is the criterion for determining whether the latest delivery time is met. , , To reward the weight coefficients, this function guides the model to select the path adjustment action that is more time-efficient and cost-effective, while also meeting timeliness requirements. The training process uses the Q-learning algorithm to continuously update the Q-value, allowing the model to learn the optimal adjustment strategy under different dynamic scenarios. The formula is shown below: In the formula, As a discount factor, The learning rate; In addition, the model training phase makes full use of historical transportation data accumulated over a long period of time, covering route planning cases and corresponding actual operation data for different seasons, time periods, and transportation scenarios (such as express delivery and long-distance logistics). At the same time, it conducts targeted training based on the core needs and characteristics of each scenario. For example, in the express delivery scenario, the training of time-related parameters is strengthened, and in the long-distance logistics scenario, the model parameters related to fuel consumption cost are optimized. During actual operation, the model will receive key data such as real-time vehicle speed, road congestion level, and instantaneous fuel consumption uploaded by edge nodes in real time, and dynamically adjust the weight coefficients in the comprehensive cost formula and the reward weight coefficients in the reward function, so that the route planning can adapt to changes in the transportation environment and vehicle status in real time. Finally, the global initial optimal path output by the genetic algorithm serves as the initial path reference for reinforcement learning. Reinforcement learning continuously monitors the path's operational status based on real-time collected data, evaluating the path's overall cost and timeliness at fixed intervals. If it finds that the current path's cost has increased or timeliness cannot be guaranteed due to dynamic changes, a fine-tuning operation is immediately triggered. The optimal adjustment action is quickly selected using a well-trained Q-value table. The two work together, ensuring the global optimality of the path through the genetic algorithm and enabling rapid adaptation to real-time dynamic changes through reinforcement learning. Ultimately, an optimal path scheme that balances global efficiency and real-time adaptability, maximizing transportation efficiency, is generated. The scheme clearly marks key information such as the estimated travel time, fuel consumption, and nodes along the route for each segment, providing clear and accurate guidance for vehicle navigation.

[0025] Step 4: Dynamic Event Response and Replanning When traffic congestion or emergencies are detected, transportation routes are replanned in real time, and route update instructions are issued to the transport vehicles. Specifically, the triggering conditions for replanning include road congestion level reaching a preset threshold, the impact range of a sudden event meeting a preset standard, abnormal vehicle fuel consumption, changes in vehicle load reaching a preset proportion, and actual travel time exceeding the preset proportion of the planned time; the route update command is issued through the MQTT protocol, and the vehicle navigation system synchronizes the updated optimal route in real time.

[0026] Each trigger condition is precisely determined based on real-time collected IoT data and preset logic: the road congestion level is calculated by combining the traffic density and road occupancy ratio identified by the vehicle camera with the ratio of real-time vehicle speed to the design speed. When the ratio is lower than the preset standard, it is determined that the congestion trigger condition is met. The impact range of the sudden event is determined by comprehensively judging the event location pushed by the roadside IoT device and the scene captured by the vehicle camera to determine the impact distance of the event on the surrounding roads. When it reaches the preset range, the response is triggered. The abnormal vehicle fuel consumption is continuously monitored by the 3σ principle for instantaneous fuel consumption data. When the fuel consumption exceeds the stable value of the normal driving range, the judgment is initiated. The change of vehicle load is captured in real time by the load sensor. When the loading and unloading causes the load to fluctuate significantly and reach the preset ratio, the trigger is initiated. The actual travel time is continuously compared by real-time speed data and route planning time. When it exceeds the preset ratio and there is no reasonable data support, the trigger process is initiated to ensure that all trigger conditions are consistent with the actual transportation scenario and avoid false triggering or missed triggering. Then, once the edge computing node detects that any trigger condition is met, it immediately starts the response mechanism to organize the core information of the event, including key information such as event type, location of occurrence, scope of impact, and current vehicle status. Without waiting for the complete data cycle, it directly uploads the information to the cloud server via the MQTT protocol and simultaneously synchronizes it to the edge gateway of the region. This enables rapid synchronization of event information across multiple devices, ensuring that the cloud can obtain complete and accurate event details in the shortest possible time, thus buying time for subsequent replanning. Simultaneously, by calling the mature genetic algorithm and reinforcement learning fusion optimization model trained in step 3 through the cloud server, based on the latest uploaded event information, global road traffic data, and real-time status data of multiple vehicles, the global optimization calculation of the new route is quickly carried out. The model prioritizes updating the weight of the road segment affected by the event in the comprehensive cost formula of the route, re-evaluates the travel time and fuel consumption cost of each candidate route, quickly searches for the globally optimal alternative route through the genetic algorithm, and then fine-tunes it to adapt to the real-time traffic conditions through reinforcement learning. The entire calculation process makes full use of the computing power advantage of the cloud service architecture to ensure that the new route is generated in a very short time. At the same time, if the scope of the event involves the overlapping area of ​​the routes of multiple transport vehicles, the multi-vehicle collaborative scheduling logic is linked in sync to adjust the route plan of the relevant vehicles to avoid multiple vehicles rushing into the alternative route and causing secondary congestion, thus ensuring the feasibility and efficiency of the new route. Finally, after the cloud server generates the optimal updated route, it sends a route update command to the target vehicle via the MQTT protocol. The command contains core information such as the nodes along the new route, the estimated travel time, and key road segment prompts. The data transmission process uses the AES-128 algorithm for encryption to ensure that the command is secure and tamper-proof. After the vehicle receives the command, the navigation system updates the route in real time and prompts the driver to switch routes through both voice and visual prompts.

[0027] Step 5: Multi-vehicle collaborative scheduling optimization Coordinate and schedule multiple transport vehicles to optimize the overall operational efficiency of the transportation network; Specifically, the constraints of collaborative scheduling include that the vehicle load does not exceed the maximum load, the real-time number of vehicles on the road does not exceed the maximum number of vehicles passing through, and the actual delivery time of the vehicle does not exceed the latest delivery time. During the scheduling process, a distributed task allocation algorithm is adopted to take into account both road load balancing and matching of goods on the return route.

[0028] Each constraint is strictly implemented through real-time IoT data and dynamic monitoring mechanisms: vehicle load constraints are achieved by collecting data in real time through load sensors and continuously comparing it with the maximum load threshold of vehicles. When allocating transportation tasks, the dispatching system calculates the weight of goods and the remaining load capacity of vehicles in advance to ensure that the vehicle load does not exceed the limit after task allocation. Road real-time vehicle number constraints are achieved by collecting traffic flow data from roadside IoT devices and vehicle cameras, combined with the road design capacity to determine the maximum number of vehicles that can pass, and real-time statistics of the current vehicle density of each road are used as the core basis for dispatching decisions. Vehicle delivery time constraints are achieved by comparing the estimated travel time in the route planning with the latest delivery time, and dynamically adjusting the driving rhythm in combination with real-time road conditions to ensure that the actual delivery time of each vehicle does not exceed the preset limit. The three constraints are interconnected and constitute the basic guarantee for collaborative dispatching. The distributed task allocation algorithm achieves precise task decomposition and matching through global data collaboration: the cloud server aggregates all transportation task information (such as cargo weight, delivery location, and time requirements) and all vehicle status data (such as real-time location, remaining load, current fuel consumption, and planned routes), and decomposes the overall transportation task into several sub-tasks according to region, time, and load requirements. Then, it allocates tasks based on the matching degree between vehicles and sub-tasks. The matching degree considers core factors such as the distance between the vehicle and the starting point of the sub-task, the compatibility of the remaining load with the cargo weight, and the fuel consumption cost of completing the sub-task. It does not require centralized calculation of the combination relationship between all vehicles and tasks, but rather processes the sub-task allocation in parallel through distributed nodes, which greatly improves scheduling efficiency and avoids overloading of a single vehicle or idle resources. Furthermore, road load balancing achieves comprehensive optimization through dynamic monitoring and route adjustment: the cloud tracks changes in the number of vehicles on each road in real time. When the real-time number of vehicles on a certain road approaches the preset proportion of the maximum number of vehicles that can pass, the load balancing mechanism is triggered. For related vehicles that have not yet entered that road, alternative routes are replanned to guide traffic flow to roads with less traffic pressure. At the same time, in return trip scheduling, the system matches the return route of vehicles with the information of goods waiting to be transported along the way. By comparing the degree of contiguity between the return route of the vehicle and the origin of the goods transportation, and the matching degree between the weight of the goods and the remaining load of the vehicle, the system prioritizes the allocation of goods along the way to return vehicles, thereby reducing empty mileage from the source and achieving efficient and circular utilization of transportation resources.

[0029] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A dynamic transportation route optimization method based on IoT, characterized in that: S1. Real-time collection of operational data of transport vehicles and road traffic data; S2. The collected data is preprocessed through edge computing nodes, and the preprocessed key data is uploaded to the cloud server. S3. An optimization model combining reinforcement learning and genetic algorithms is used to generate the optimal route scheme that maximizes transportation efficiency. S4. When traffic congestion or emergencies are detected, the transportation route is replanned in real time and route update instructions are issued to the transportation vehicles. S5. Coordinate and schedule multiple transport vehicles to optimize the overall operational efficiency of the transportation network.

2. The IoT-based dynamic transportation route optimization method according to claim 1, characterized in that: In S1, the operation-related data includes the vehicle's real-time location, instantaneous speed, instantaneous fuel consumption, and real-time load, while the road traffic-related data includes road congestion level, type of emergency, and scope of impact. The data is collected through vehicle-mounted GPS, speed sensor, fuel consumption sensor, load sensor, vehicle-mounted camera, and roadside IoT devices.

3. The IoT-based dynamic transportation route optimization method according to claim 1, characterized in that: In step S2, preprocessing includes compressing the data using the LZ77 lossless compression algorithm and detecting abnormal data using the 3σ principle; key data is uploaded via the MQTT protocol, and data encryption uses the AES-128 algorithm.

4. The IoT-based dynamic transportation route optimization method according to claim 1, characterized in that: In S3, the optimization model employs a hierarchical fusion of genetic algorithms and reinforcement learning. The model is generated through training on historical transportation data, and the training process incorporates features from different transportation scenarios. The model dynamically adjusts parameters based on key data collected in real time during runtime.

5. The IoT-based dynamic transportation route optimization method according to claim 1, characterized in that: In S4, the triggering conditions for replanning include: the road congestion level reaching a preset threshold, the impact range of a sudden event meeting a preset standard, abnormal vehicle fuel consumption, changes in vehicle load reaching a preset proportion, and actual travel time exceeding the planned time by a preset proportion. The route update command is sent via the MQTT protocol, and the vehicle navigation system synchronizes the updated optimal route in real time.

6. The IoT-based dynamic transportation route optimization method according to claim 1, characterized in that: In S5, the constraints of collaborative scheduling include that the vehicle load does not exceed the maximum load, the real-time number of vehicles on the road does not exceed the maximum number of vehicles passing through, and the actual delivery time of the vehicle does not exceed the latest delivery time. During the scheduling process, a distributed task allocation algorithm is adopted to take into account both road load balancing and matching of return trip goods.

7. The IoT-based dynamic transportation route optimization method according to claim 1, characterized in that: The method supports real-time route optimization for large-scale transport fleets and adopts an edge computing + cloud collaborative architecture. The edge computing nodes have local emergency route planning functions, and the cloud server adopts a microservice architecture and a distributed database.

Images