A transport vehicle scheduling method for a smart port

Abstract

The application discloses a transport vehicle scheduling method for a smart port, and relates to the field of smart port logistics management.The method comprises the following steps: acquiring attribute information of each freight point and state information of available vehicles at a current time of the port in real time; calculating the urgency of goods to be transported in each freight point based on the attribute information of the freight point; constructing a grid map of the port and generating a basic path length from each available vehicle to each freight point based on a path search algorithm; constructing a multi-dimensional feature vector of each freight point and clustering the multi-dimensional feature vector to obtain a plurality of clustering clusters and calculate a scheduling path length from each available vehicle to each clustering cluster; calculating a scheduling tendency between each clustering cluster and each available vehicle, solving an optimal matching scheme between the available vehicles and the clustering clusters at the current time, and generating a corresponding task instruction to schedule the available vehicles.The application can realize optimal adaptation between the transport vehicles and the transport tasks, and improve the material circulation efficiency.

Description

Technical Field

[0001] This invention relates to the field of smart port logistics management technology, and in particular to a method for scheduling transport vehicles in smart ports. Background Technology

[0002] Existing smart port vehicle scheduling methods mostly rely on genetic algorithms, rolling time-domain frameworks, and hyperheuristic algorithms combining deep reinforcement learning and genetic programming for task allocation and path planning. The core idea is typically to collect vehicle and environmental data by integrating IoT sensors, the BeiDou Navigation Satellite System, and database servers, and then apply specific algorithmic models to perform calculations aimed at achieving vehicle path planning and task allocation. Many methods emphasize real-time data acquisition and dynamic response, such as reducing the solution scale through rolling time-domain control or improving response speed using distributed processing subsystems. However, many existing technologies focus more on optimizing the algorithm itself or primarily addressing the efficiency of single aspects, such as path search and task allocation, which limits the effectiveness of existing scheduling methods.

[0003] First, many methods fail to adequately consider the complex physical constraints of ports, such as the impact of obstacles on actual paths, which may lead to decreased efficiency in actual scheduling. Second, regarding adaptability to dynamic environments, although some methods mention real-time performance, their robustness and multi-dynamic response mechanisms still have room for improvement when dealing with multiple contingencies. Furthermore, many methods fail to adequately balance task requirements and vehicle status, either focusing on the urgency of the task itself or the vehicle's current position, lacking a decision-making model that deeply integrates various factors, which can easily lead to suboptimal scheduling results in real-world applications. Summary of the Invention

[0004] To achieve optimal matching between transport vehicles and transport tasks, and to address the problems of static planning, response lag, and uneven resource utilization in existing smart port vehicle scheduling systems, this invention provides a transport vehicle scheduling method for smart ports, the technical solution of which is as follows:

[0005] In a first aspect, the present invention provides a method for scheduling transport vehicles in a smart port, comprising the following steps: acquiring in real time the attribute information of each cargo point and the status information of available vehicles at the current time in the port; calculating the urgency of goods to be transported within each cargo point based on the attribute information of the cargo points; constructing a grid map of the port and generating the basic path length from each available vehicle to each cargo point based on a path search algorithm; constructing a multi-dimensional feature vector for each cargo point based on the attribute information, urgency, and basic path length of the cargo points and performing clustering to obtain several clusters and calculating the scheduling path length from each available vehicle to each cluster; calculating the scheduling tendency between each cluster and each available vehicle based on the attribute information and urgency of the cargo points within each cluster, as well as the status information and scheduling path length of the available vehicles; solving for the optimal matching scheme between the available vehicles and the clusters at the current time based on the scheduling tendency, and generating corresponding task instructions to schedule the available vehicles.

[0006] Preferably, the system collects attribute information of cargo points within the port in real time from the port business system, including the number and number of cargo points, the center coordinates of each cargo point, the weight of the goods to be transported at each cargo point, and the waiting time of the goods to be transported; it also collects the status information of each transport vehicle in real time from the vehicle's IoT device, including the current working status, current location coordinates, and current remaining battery percentage data, removes vehicle information with current remaining battery percentage data below a threshold, marks transport vehicles with an idle working status as available vehicles, counts the number of available vehicles and assigns them numbers, extracts the status information data of available vehicles, and obtains the current location coordinates and current remaining battery percentage data of each available vehicle.

[0007] Preferably, based on the waiting time of the goods to be transported at each freight point, the Sigmoid function is used to quantify the urgency of the goods to be transported at each freight point, so as to obtain the priority of the goods to be transported at each freight point. The priority of the goods to be transported at each freight point is normalized, and the normalized value is used as the urgency of the goods to be transported at each freight point. The urgency of the goods to be transported at each freight point is calculated in turn.

[0008] Preferably, based on the actual layout of the port, the entire port area is divided into several identical grid units to construct a rasterized map containing obstacle information. The size of each grid unit in the map is set according to the scene accuracy requirements. Areas where transport vehicles can pass normally are marked as passable grids, and areas where transport vehicles cannot pass are marked as obstacle grids. The current position coordinates of each available vehicle and the center position coordinates of each cargo point are extracted. The current position coordinates of each available vehicle are used as the path start point, and the center position coordinates of each cargo point are used as the path end point. The A* algorithm is used to generate the shortest passable path from each path start point to each path end point, which is used as the shortest collision-free path from the corresponding available vehicle to each cargo point. Based on the rasterized map, the passable grid sequence of each path is obtained. Based on the set grid unit size, the product of the number of passable grids and the grid unit size of each path is used as the basic path length of the corresponding path. The basic path length from each available vehicle to each cargo point is calculated sequentially.

[0009] Preferably, the basic path length from each available vehicle to each freight point is extracted, and the average of the basic path lengths corresponding to each freight point is used as the path feature value of the corresponding freight point. The center location coordinates, weight of the goods to be transported, urgency of the goods to be transported, and path feature values ​​of each freight point are merged to obtain the feature vector of each freight point. The DBSCAN clustering algorithm is used to cluster the feature vectors of all freight points to obtain several clusters. The center location coordinates of each freight point in each cluster are extracted, and the average of the center location coordinates of each freight point in each cluster is used as the center coordinates of the corresponding cluster. Based on the current location coordinates of each available vehicle, the A* algorithm is used to generate the shortest travel path from each available vehicle to the center coordinates of each cluster. Then, based on the grid map of the port, the scheduling path length from each available vehicle to each cluster is calculated.

[0010] Preferably, the weight and urgency of the goods to be transported corresponding to each shipping point are extracted. Based on the clustering results, the total weight of the goods to be transported in each cluster is taken as the load of the corresponding cluster, and the average urgency of the goods to be transported in each cluster is taken as the average time urgency of the corresponding cluster. The maximum value among all the loads of the clusters is extracted, and the ratio between the load of the cluster and the maximum value is taken as the load feature value of the cluster. The product between the load feature value of each cluster and the corresponding average time urgency is taken as the attraction of the corresponding cluster.

[0011] Preferably, the current remaining battery percentage data of each available vehicle and the scheduling path length data from each available vehicle to each cluster are extracted. The ratio between the scheduling path length from an available vehicle to a certain cluster and the current remaining battery percentage data of that available vehicle is calculated, and the ratio is normalized. The normalized value is used as the repulsive force between the available vehicle and this cluster. Similarly, the repulsive force between each available vehicle and each cluster is obtained. Based on the attractiveness of each cluster and the repulsive force between each cluster and each available vehicle, the difference between the attractiveness of the cluster and the repulsive force between each available vehicle is calculated sequentially, and the difference is mapped using a natural exponential function to obtain the scheduling tendency between each cluster and each available vehicle.

[0012] Preferably, the rated load data of port transport vehicles is obtained. Based on the total weight of the goods to be transported corresponding to each cluster, the total weight of the goods to be transported in each cluster is decomposed into several standardized task units. Each unit represents the rated load of a transport vehicle when fully loaded. When the weight of the remaining goods to be transported is less than the rated load of the transport vehicle, it is still processed as a standardized task unit. A bipartite graph model is constructed, where the left nodes represent available vehicles, and each available vehicle corresponds to a left node. The right nodes represent task units, and each task unit corresponds to a right node. The edge weights are set based on the scheduling preference between each cluster and each available vehicle. When the number of available vehicles is not equal to the number of task units, a virtual node mechanism is used to add virtual vehicles or task units and set their edge weights to zero to balance the bipartite graph. The KM algorithm is used to solve the maximum weight matching of the balanced bipartite graph to obtain the optimal matching scheme between available vehicles and clusters at the current time.

[0013] Preferably, the optimal matching scheme between available vehicles and clusters at the current moment is transformed into an executable scheduling task instruction and issued to each available vehicle. Virtual matching scenarios where vehicles are idle or tasks are not assigned are marked, and the system waits for the next round of vehicle scheduling scheme updates. A dual dynamic response mechanism is set up, which includes periodically refreshing the vehicle scheduling scheme and forcibly regenerating the vehicle scheduling scheme when abnormal events are triggered. The attribute information of each freight point and the status information of the transport vehicles are updated in real time to ensure the accuracy of the vehicle scheduling scheme.

[0014] Secondly, the present invention provides a transportation vehicle scheduling system for smart ports, used to implement the aforementioned transportation vehicle scheduling method for smart ports, comprising: a processor, a memory, a communication interface, a data acquisition module, a port business system, and a vehicle Internet of Things (IoT) device. The processor stores computer program instructions for implementing the aforementioned transportation vehicle scheduling method for smart ports. The communication interface is communicatively connected to the data acquisition module and the vehicle IoT device. The data acquisition module is communicatively connected to the port business system and the vehicle IoT device.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0016] This invention, by introducing a grid map and the A* path search algorithm, accurately quantifies the actual travel cost from vehicles to the loading point, making subsequent clustering and scheduling decisions more closely aligned with the port's actual physical layout and overcoming the bias of simply calculating linear distance. Simultaneously, this invention innovatively proposes a scheduling tendency calculation model based on attraction and repulsion forces, dynamically playing a game and balancing the cargo demand and vehicle status within clustered regions, achieving a deep integration of vehicle and task characteristics. Subsequently, global optimization allocation is performed using the KM algorithm, supplemented by a virtual node balancing mechanism and a dual dynamic response mechanism, ensuring that the scheduling scheme not only... The global optimization also significantly improves the system's real-time response capability and robustness in the face of various emergencies, ultimately achieving optimal matching between transport vehicles and transport tasks, and effectively improving the overall scheduling efficiency of port transport vehicles. This effectively solves the problems of static planning, response lag, and uneven resource utilization in existing smart port vehicle scheduling systems, comprehensively improving the efficiency of port material flow, reducing vehicle empty runs and waiting time, ensuring priority handling of emergency tasks, and enhancing the real-time response capability of the entire scheduling system to emergencies, providing core decision support for the automated and intelligent operation of modern smart ports. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating an implementation method for a transportation vehicle scheduling method in a smart port according to an embodiment of the present invention.

[0018] Figure 2 This is a structural block diagram of a transportation vehicle scheduling system for smart ports according to an embodiment of the present invention. Detailed Implementation

[0019] The technical features of the present invention will be further described in detail below with reference to the accompanying drawings so that those skilled in the art can understand them.

[0020] A method for dispatching transport vehicles in smart ports, the implementation process of which is as follows: Figure 1 As shown, the specific implementation steps are as follows:

[0021] Step S1: Obtain the attribute information of each cargo point and the status information of available vehicles at the port in real time.

[0022] Specifically, the system collects real-time attribute information of cargo points within the port from the port's business system. This includes the number and number of cargo points, the center coordinates of each cargo point, the weight of goods to be transported at each cargo point, and the waiting time for the goods to be transported. It also collects real-time status information of each transport vehicle from IoT devices such as GPS sensors and vehicle terminals. The status information of each transport vehicle includes its current operating status, current location coordinates, current remaining battery percentage, and rated load capacity. The current operating status of a transport vehicle can be divided into transport status and idle status. Vehicles with a current remaining battery percentage below a threshold are removed, and transport vehicles currently in an idle state are marked as available vehicles. The remaining battery threshold can be set to 20%. The number of available vehicles is counted and numbered, and the status information data of available vehicles is extracted to obtain the current location coordinates, current remaining battery percentage, and rated load capacity of each available vehicle.

[0023] The coordinates of the center location of the cargo point and the current location coordinates of the available vehicles are both within a coordinate system established based on the port map. The waiting time of the cargo to be transported can be obtained by extracting the arrival date and time of the cargo to be transported in each cargo point, and subtracting the arrival date and time of the cargo to be transported from the current date and time. The unit is hours.

[0024] Step S2: Based on the attribute information of the freight points, calculate the urgency of the goods to be transported within each freight point.

[0025] Specifically, based on the waiting time of goods to be transported at each freight point, the Sigmoid function is used to quantify the urgency of goods to be transported at each freight point, and the priority of goods to be transported at each freight point is obtained. The priority of goods to be transported at each freight point is normalized, and the normalized value is used as the urgency of goods to be transported at each freight point. The urgency of goods to be transported at each freight point is calculated in turn.

[0026] The urgency level of the goods awaiting transport at the j-th freight point is: The calculation formula is as follows:

[0027]

[0028] In the formula, This represents the waiting time for goods to be transported at the j-th shipping point. This represents the natural exponential function. express The normalization function is expressed as follows: , This indicates the urgency of goods awaiting transport at a specific shipping point. This represents the minimum urgency level of goods awaiting transport across all shipping points. This represents the maximum urgency level among all cargo points awaiting transport; the higher the urgency value, the higher the urgency level, and the more urgent the transport task of the cargo awaiting transport at the corresponding cargo point.

[0029] Step S3: Construct a grid map of the port and generate the basic path length from each available vehicle to each cargo point based on the path search algorithm.

[0030] Specifically, by scanning the entire port area using cameras or LiDAR, map data, building data, and obstacle data can be obtained. Based on the actual layout of the port, the entire port area is divided into several identical grid units to construct a rasterized map containing obstacle information. The size of each grid unit in the map is set according to the scene's accuracy requirements, and can be set to 1m×1m. Areas where transport vehicles can pass normally are marked on the map as passable grids, corresponding to areas such as passageways and work areas where transport vehicles can pass normally. Areas where transport vehicles cannot pass are marked on the map as obstacle grids, corresponding to areas such as steps, pillars, walls, and large equipment where transport vehicles cannot pass. The available vehicles for each task are then extracted. The current location coordinates of each vehicle and the center location coordinates of each delivery point are used. The current location coordinates of each available vehicle are taken as the starting point of the path, and the center location coordinates of each delivery point are taken as the ending point of the path. The A* algorithm (A-Star, a path finding and graph traversal algorithm) is used to generate the shortest passage path from the starting point to each ending point of each path, which is used as the shortest collision-free path from the corresponding available vehicle to each delivery point. Based on the raster map, the passage grid sequence of each path is obtained. Based on the set grid cell size, the product of the number of passage grids and the grid cell size of each path is taken as the basic path length of the corresponding path. The basic path length from each available vehicle to each delivery point is calculated in turn.

[0031] Step S4: Based on the attribute information, urgency and basic path length of the delivery point, construct a multi-dimensional feature vector for each delivery point and perform clustering to obtain several clusters, and calculate the scheduling path length from each available vehicle to each cluster.

[0032] Specifically, the basic path length from each available vehicle to each freight point is extracted. The average of the basic path lengths corresponding to each freight point is used as the path feature value of the corresponding freight point. The center location coordinates, weight of the goods to be transported, urgency of the goods to be transported, and path feature values ​​of each freight point are merged to obtain the feature vector of each freight point. The DBSCAN clustering algorithm (Density-Based Spatial Clustering of Applications with Noise, a density-based unsupervised learning clustering algorithm) is used to cluster the feature vectors of all freight points to obtain several clusters. The center location coordinates of each freight point in each cluster are extracted. The average of the center location coordinates of each freight point in each cluster is used as the center coordinates of the corresponding cluster. Based on the current location coordinates of each available vehicle, the A* algorithm is used to generate the shortest travel path from each available vehicle to the center coordinates of each cluster. Then, based on the port's grid map, the scheduling path length from each available vehicle to each cluster is calculated, consistent with the calculation process of the basic path length in step S3.

[0033] Within the same cluster, each cargo point has similar actual travel costs and time constraints, meaning they have similar urgency and path characteristics of the goods to be transported. When scheduling transport vehicles, they can be considered as a whole. For example, if the same cargo ship unloads its cargo at several adjacent cargo points, and these cargo points have basically the same travel costs and time constraints, they can be treated as a whole area for scheduling transport vehicles, thereby improving the efficiency of subsequent matching calculations and enhancing the real-time performance of transport vehicle scheduling.

[0034] Step S5: Based on the attribute information and urgency of the freight points within each cluster, as well as the status information of available vehicles and the length of the scheduling path, calculate the scheduling preference between each cluster and each available vehicle.

[0035] Specifically, the weight and urgency of the goods to be transported corresponding to each shipping point are extracted. Based on the clustering results, the total weight of the goods to be transported in each cluster is taken as the load of the corresponding cluster, and the average urgency of the goods to be transported in each cluster is taken as the average time urgency of the corresponding cluster. The maximum value among all cluster loads is extracted, and the ratio between the cluster load and the maximum value is taken as the load feature value of the cluster. The product of the load feature value of each cluster and the corresponding average time urgency is taken as the attraction of the corresponding cluster.

[0036] Wherein, the attractiveness of the k-th cluster is The calculation formula is as follows:

[0037]

[0038] In the formula, This represents the total weight of goods to be transported within the k-th cluster at the current time, i.e., the load capacity of the k-th cluster at the current time. This represents the maximum value among all cluster loads. This represents the average urgency of goods to be transported within the k-th cluster, i.e., the average time urgency of the k-th cluster; The range of values ​​is , The range of values ​​is ,therefore, The range of values ​​is also The higher the attractiveness value, the higher the demand for vehicle scheduling for the goods to be transported within the corresponding cluster. When scheduling transport vehicles, it tends to prioritize matching available vehicles to clusters with a large number of goods to be transported and tight schedules.

[0039] Furthermore, since the location of each available vehicle at the port is random, and the travel distance from each available vehicle to each cluster is also different, vehicle scheduling should prioritize the available vehicles with shorter travel distances for cargo transportation. Moreover, the remaining battery power of each available vehicle is also a factor to consider when matching transportation tasks. If the same transportation vehicle is frequently scheduled, its battery power will drop rapidly, and transportation tasks may not be completed due to insufficient power. Therefore, when scheduling vehicles, the battery power of each transportation vehicle should be balanced, and available vehicles with high remaining battery power at the current moment should be prioritized for cargo transportation. This allows transportation vehicles to be scheduled in batches and alternate between charging and transportation states to improve the overall carrying efficiency of transportation vehicles. Therefore, a repulsion force is defined by combining the scheduling path length and the current remaining battery power percentage data to reflect and quantify the scheduling cost of available vehicles to clusters, so as to guide each cluster to select the optimal available vehicle to perform the transportation task.

[0040] Extract the current remaining battery percentage data and the scheduling path length data from each available vehicle to each cluster. Calculate the ratio between the scheduling path length from an available vehicle to a specific cluster and the current remaining battery percentage data of that available vehicle. Normalize the ratio and use the normalized value as the repulsive force between that available vehicle and that cluster. Similarly, obtain the repulsive force between each available vehicle and each cluster. Based on the attractive force of each cluster and the repulsive force between each cluster and each available vehicle...

[0041] Wherein, the repulsive force between the k-th cluster and the i-th available vehicle is The calculation formula is as follows:

[0042]

[0043] In the formula, This represents the length of the scheduling path from the i-th available vehicle to the k-th cluster. This represents the current remaining battery percentage of the i-th available vehicle. express The normalization function is expressed as follows: , This represents the ratio between the scheduling path length from an available vehicle to a specific cluster and the current remaining battery percentage of that available vehicle. This represents the minimum value among all ratios. This represents the maximum value among all ratios; the larger the repulsive force value, the higher the cost of performing transportation tasks between the corresponding available vehicles and the cluster, and the worse the matching degree.

[0044] Subsequently, based on the attractiveness of each cluster and the repulsion between each cluster and each available vehicle, the difference between the attractiveness of the cluster and the repulsion between each available vehicle is calculated sequentially. The natural exponential function is then used to map the difference to obtain the scheduling tendency between each cluster and each available vehicle. The attractiveness of the cluster is used to guide available vehicles to prioritize serving clusters with high demand, while the repulsion between the cluster and available vehicles is used to optimize vehicle selection. The larger the value of the scheduling tendency between the cluster and available vehicles, the stronger the suitability of scheduling the corresponding available vehicle to the corresponding cluster, and the higher the matching degree between the two.

[0045] Wherein, the scheduling preference between the k-th cluster and the i-th available vehicle is The calculation formula is as follows:

[0046]

[0047] In the formula, This represents the attractiveness of the k-th cluster. This represents the repulsive force between the k-th cluster and the i-th available vehicle. This represents the natural exponential function.

[0048] Step S6: Based on the scheduling preference, find the optimal matching scheme between available vehicles and clusters at the current moment, and generate corresponding task instructions to schedule available vehicles.

[0049] Specifically, the rated load data of port transport vehicles is obtained. Based on the total weight of the cargo to be transported corresponding to each cluster, the total weight of the cargo to be transported within each cluster is decomposed into several standardized task units. Each unit represents the rated load of a transport vehicle when fully loaded. When the weight of the remaining cargo to be transported is less than the rated load of the transport vehicle, it is still processed as a standardized task unit. A bipartite graph model is constructed, where the left nodes represent available vehicles, with each available vehicle corresponding to a left node, and the right nodes represent task units, with each task unit corresponding to a right node. The edge weights are set based on the scheduling preference between each cluster and each available vehicle. When the number of available vehicles is not equal to the number of task units, a virtual node mechanism is used to add virtual vehicles or task units and set their edge weights to zero to balance the bipartite graph. The KM algorithm (Kuhn-Munkres, the maximum weight matching algorithm for bipartite graphs) is used to solve the maximum weight matching of the balanced bipartite graph to obtain the optimal matching scheme between available vehicles and clusters at the current time.

[0050] In addition, the optimal matching scheme between available vehicles and clusters at the current moment is transformed into an executable scheduling task instruction and sent to each available vehicle. Virtual matching scenarios where vehicles are idle or tasks are not assigned are marked and await the next round of vehicle scheduling scheme updates. A dual dynamic response mechanism is set up to periodically refresh the vehicle scheduling scheme and force the regeneration of the vehicle scheduling scheme when abnormal events are triggered. The attribute information of each freight point and the status information of the transport vehicles are updated in real time to ensure the accuracy of the vehicle scheduling scheme.

[0051] In virtual matching scenarios where vehicles are idle or tasks are not assigned, when the number of available vehicles is greater than the number of task units, some available vehicles are idle and enter a standby state, such as waiting in place or parking in the charging area; when the number of available vehicles is less than the number of task units, the transportation demand in the corresponding area is not completed, and the uncompleted amount of goods to be transported is re-included in the data collected in the next round of vehicle dispatching process, waiting for the update of vehicle dispatching task instructions.

[0052] For the dual dynamic response mechanism, when the transport vehicles are operating normally and no abnormal events are triggered, the scheduling plan is automatically updated and optimized based on a preset time period. When an abnormal event is triggered, the current vehicle scheduling process is immediately interrupted, and a new scheduling plan is generated by re-collecting information data based on the changes caused by the abnormal situation. Abnormal events include: the arrival of new transport tasks, such as a new ship docking at the port or receiving a task order that requires urgent transport; sudden changes in resource status, such as a transport vehicle suddenly malfunctioning, the remaining power of a transport vehicle falling below the safety threshold during operation, or a transport vehicle remaining stationary for a long time due to traffic congestion; and task completion or resource release, such as when a loading and unloading operation is completed and idle transport vehicles and site resources are released.

[0053] This invention also discloses a transportation vehicle scheduling system for smart ports, used to implement the aforementioned transportation vehicle scheduling method for smart ports. The system structure is as follows: Figure 2 As shown, it includes: a processor, a memory, a communication interface, a data acquisition module, a port business system, and vehicle IoT devices. The vehicle IoT devices include GPS sensors and vehicle-mounted terminals. The processor stores computer program instructions for implementing the aforementioned method for scheduling transport vehicles in a smart port. The communication interface is connected to the data acquisition module and the vehicle IoT devices, and is used to receive information data acquired by the data acquisition module and to issue scheduling instructions to the vehicle IoT devices of each transport vehicle to execute the scheduling operation of the transport vehicles. The data acquisition module is connected to the port business system and the vehicle IoT devices, and is used to collect attribute information of each cargo point in the port and status information of transport vehicles in real time.

[0054] The embodiments included in this invention are merely descriptions of preferred embodiments of the invention and are not limited to the precise structures described above and shown in the accompanying drawings. Various modifications and changes can be made without departing from the scope of protection. Any variations and improvements made by those skilled in the art to the technical solutions of this invention without departing from the design concept of this invention should fall within the scope of protection of this invention.

Claims

1. A method for scheduling transport vehicles in smart ports, characterized in that: Real-time acquisition of the attribute information of each delivery point and the state information of available vehicles in the port at the current time; Based on the attribute information of the delivery point, the urgency of the goods to be transported in each delivery point is calculated; a grid map of the port is constructed and based on the path search algorithm, the basic path length of each available vehicle to each delivery point is generated; Based on the attribute information of the delivery point, the urgency and the basic path length, a multi-dimensional feature vector of each delivery point is constructed and clustered to obtain a number of clustering clusters and calculate the scheduling path length of each available vehicle to each clustering cluster; based on the attribute information and urgency of the delivery point in each clustering cluster and the state information and scheduling path length of the available vehicle, the scheduling tendency between each clustering cluster and each available vehicle is calculated; Based on the scheduling tendency, the optimal matching scheme between the available vehicles and the clustering clusters at the current time is solved, and the corresponding task instructions are generated to schedule the available vehicles; Wherein, the attribute information of the delivery point in the port is collected in real time from the port business system, including the number and number of delivery points, the center position coordinates of each delivery point, the weight of the goods to be transported in each delivery point and the waiting time of the goods to be transported; the state information of each transport vehicle is collected in real time from the Internet of Things device of the vehicle, including the current working state, the current position coordinates, the current remaining percentage data of the power, the vehicle information with the current remaining percentage data of the power below the threshold value is excluded, and the transport vehicle with the current working state as idle is marked as available vehicle, the number of available vehicles is counted and numbered, the state information data of the available vehicle is extracted, and the current position coordinates and the current remaining percentage data of the power of each available vehicle are obtained; The total weight of the goods to be transported in each clustering cluster is taken as the load of the corresponding clustering cluster, the mean value of the urgency of the goods to be transported in each clustering cluster is taken as the average time urgency of the corresponding clustering cluster, the maximum value in the load of all clustering clusters is extracted, the ratio between the load of the clustering cluster and the maximum value is taken as the load characteristic value of the clustering cluster, and the product between the load characteristic value of each clustering cluster and the corresponding average time urgency is taken as the attraction of the corresponding clustering cluster; The current remaining percentage data of the power of each available vehicle and the scheduling path length data of each available vehicle to each clustering cluster are extracted, the ratio between the scheduling path length of a certain available vehicle to a certain clustering cluster and the current remaining percentage data of the power of the available vehicle is calculated, and the ratio is normalized to obtain the repulsion between the available vehicle and the clustering cluster, and the repulsion between each available vehicle and each clustering cluster is obtained in the same way; The difference between the attraction of the clustering cluster and the repulsion between each available vehicle is calculated in turn, and the difference is mapped by using the natural exponential function to obtain the scheduling tendency between each clustering cluster and each available vehicle. 2.The method for scheduling a transport vehicle in a smart port according to claim 1, wherein, The attribute information based on the freight delivery point is used to calculate the urgency of the goods to be transported in each freight delivery point, including: based on the waiting time of the goods to be transported in each freight delivery point, using a Sigmoid function to quantify the urgency of the goods to be transported in each freight delivery point, obtaining the priority of the goods to be transported in each freight delivery point, and performing normalization processing on the priority of the goods to be transported in each freight delivery point, taking the normalized value as the urgency of the goods to be transported in the freight delivery point, and sequentially calculating the urgency of the goods to be transported in each freight delivery point. 3.The method for scheduling the transport vehicles in the smart port according to claim 1, wherein, The method for constructing the grid map of the port and generating the basic path length of each available vehicle to each delivery point based on the path search algorithm comprises the following steps: based on the actual layout of the port, the entire port area is divided into a plurality of same grid units, a grid map containing obstacle information is constructed, the size of each grid unit in the map is set according to the scene accuracy requirement, the area where the transport vehicle can normally pass is marked as a passing grid in the map, and the area where the transport vehicle cannot pass is marked as an obstacle grid in the map; the current position coordinates of each available vehicle and the center position coordinates of each delivery point are extracted, the current position coordinates of each available vehicle are taken as the starting point of the path, the center position coordinates of each delivery point are taken as the end point of the path, the A algorithm is used to generate the shortest passing path from each path starting point to each path end point as the shortest collision-free path of each available vehicle to each delivery point; based on the grid map, the passing grid sequence of each path is obtained, based on the set grid unit size, the product of the number of passing grids of each path and the grid unit size is taken as the basic path length of the corresponding path, and the basic path length of each available vehicle to each delivery point is sequentially calculated. 4.The method for scheduling a transport vehicle in a smart port according to claim 1, wherein, The attribute information of the delivery points, the urgency and the basic path length are used to construct a multi-dimensional feature vector of each delivery point and clustering is performed to obtain a plurality of clustering clusters and to calculate a scheduling path length of each available vehicle to each clustering cluster, including: extracting a basic path length of each available vehicle to each delivery point, taking an average of the basic path lengths corresponding to each delivery point as a path characteristic value of the corresponding delivery point, combining the center position coordinates corresponding to each delivery point, the weight of the goods to be transported, the urgency of the goods to be transported and the path characteristic value to obtain a feature vector of each delivery point, performing clustering on the feature vectors of all delivery points by using a DBSCAN clustering algorithm to obtain a plurality of clustering clusters; extracting the center position coordinates of each delivery point in each clustering cluster, taking an average of the center position coordinates of each delivery point in each clustering cluster as a center coordinate of the corresponding clustering cluster, and based on a current position coordinate of each available vehicle, generating a shortest travel path of each available vehicle to the center coordinate of each clustering cluster by using an A algorithm, and then calculating a scheduling path length of each available vehicle to each clustering cluster based on a grid map of the port. 5.The method for scheduling a transport vehicle in a smart port according to any one of claims 1 to 4, characterized in that, The optimal matching scheme between the available vehicles and the clustering clusters at the current time is solved based on the scheduling tendency, including: obtaining rated load data of port transport vehicles, based on the total weight of the goods to be transported corresponding to each clustering cluster, decomposing the total weight of the goods to be transported in each clustering cluster into several standardized task units, each unit representing the rated load of a full load transport vehicle, when the weight of the remaining goods to be transported is less than the rated load of the transport vehicle, it is still treated as a standardized task unit; a bipartite graph model is constructed, the left nodes represent available vehicles, each available vehicle corresponds to a left node, the right nodes represent task units, each task unit corresponds to a right node, and the edge weight is set based on the scheduling tendency between each clustering cluster and each available vehicle; when the number of available vehicles and the number of task units are not equal, a virtual node mechanism is used to add virtual vehicles or task units and set their edge weight to zero, and the bipartite graph is balanced; the KM algorithm is used to solve the maximum weight matching of the balanced bipartite graph, and the optimal matching scheme between the available vehicles and the clustering clusters at the current time is obtained. 6.The method for scheduling a transport vehicle in a smart port according to claim 5, wherein, The corresponding task instructions are generated to schedule the available vehicles, including: converting the optimal matching scheme between the available vehicles and the clustering clusters at the current time into executable scheduling task instructions and issuing them to each available vehicle, marking the virtual matching scenario of idle vehicles or unassigned tasks, and waiting for the update of the next round of vehicle scheduling scheme; a double dynamic response mechanism of vehicle scheduling scheme regular refresh and forced regeneration of vehicle scheduling scheme triggered by abnormal events is superimposed, the attribute information of each freight delivery point and the state information of the transport vehicles are updated in real time, and the accuracy of the vehicle scheduling scheme is ensured.

7. A transport vehicle dispatching system for a smart port, characterized by: It comprises a processor, a memory, a communication interface, a data acquisition module, a port business system and a vehicle Internet of Things device, the processor stores computer program instructions for implementing the transport vehicle scheduling method for a smart port according to any one of claims 1 to 6, the communication interface is communicatively connected with the data acquisition module and the vehicle Internet of Things device, and the data acquisition module is communicatively connected with the port business system and the vehicle Internet of Things device.

Images