A space-time path planning method and system for port cargo flow matching

Abstract

This invention provides a spatiotemporal path planning method and system for port cargo flow matching, belonging to the field of port intelligent transportation and logistics scheduling technology. It is applicable to the closed, high-density operating environment of automated container terminals, solving problems such as road resource competition and mismatched loading and unloading windows. The method sequentially constructs a spatiotemporal model of the port road network, collects and integrates multi-source data from vehicles, loading and unloading equipment, and roads, collaboratively allocates spatiotemporal right-of-way, plans spatiotemporal paths for cargo flow matching, and determines vehicle speed and lanes to avoid head-on conflicts. The system consists of vehicle-mounted sensing and communication, roadside monitoring, loading and unloading equipment status monitoring, and central dispatch control, achieving low-latency interaction via a 5G network. This invention achieves dynamic allocation of spatiotemporal right-of-way and coordinated loading and unloading operations, proactively prevents congestion and avoids head-on conflicts, utilizes road resources, and can quickly respond to dynamic port conditions. Furthermore, it is based on existing hardware, easy to deploy, and has shown significant effectiveness in verification at Yangshan Port, greatly improving port logistics operation efficiency.

Description

Technical Field

[0001] This invention belongs to the field of port intelligent transportation and logistics scheduling technology, specifically involving the organization of transport vehicle traffic and congestion prevention technology in the closed road environment of automated container terminals, and particularly involving a spatiotemporal path planning method that integrates road spatiotemporal resource modeling and loading and unloading operation coordination for the closed high-density operation environment of port container logistics. Background Technology

[0002] Automated container terminals are characterized by enclosed road structures and high-concurrency operations. Transport vehicles need to frequently travel back and forth in areas such as quay crane fronts, yard access roads, and intersections, and must coordinate with the operating rhythm of loading and unloading equipment such as quay cranes and yard cranes. When ships berth in large numbers, loading and unloading tasks are dense, or local access is restricted, vehicles in the port are prone to queuing, idling, and road segment conflicts, which can lead to mismatch of loading and unloading windows, discontinuous operating rhythms, and serious waste of transport capacity.

[0003] Due to the inherent limitations of narrow roads, numerous intersections, and limited detour capacity within the port, localized congestion can spread rapidly throughout the road network, leading to cascading blockages. Existing vehicle scheduling methods within the port primarily rely on task assignment or shortest path planning at a two-dimensional spatial level, treating road resources as continuous, shareable channels. This lacks explicit descriptions of road segment occupancy conflicts within specific time windows, making it impossible to proactively constrain vehicle traffic under high-density concurrent operation conditions. While some solutions incorporate traffic monitoring or congestion prediction, they only participate in decision-making through penalties or detour rules, failing to abstract road resources into allocable "road segment-time slot" access permissions, thus hindering the achievement of spatiotemporal collaborative organization for congestion prevention.

[0004] Furthermore, the operation of transport vehicles is highly dependent on the timing of loading and unloading equipment. Existing methods generally treat loading and unloading as fixed service times or independent constraints, lacking a collaborative control mechanism based on the dynamic time consumption of equipment. This leads to a mismatch between vehicle arrival at the handover point and the loading and unloading window, resulting in additional waiting or even equipment waiting idly, and the transportation plan cannot be continuously connected with the three-dimensional loading and unloading operations. At the same time, scenarios such as narrow passing on closed roads in port areas and competition for one-way lanes can easily lead to head-on collisions between vehicles. Existing collision avoidance strategies mostly focus on braking or following control at the single vehicle level, which is insufficient for speed-lane collaborative decision-making in multi-channel port scenarios with sensitive loading and unloading windows. Local conflicts can easily be amplified into large-scale queuing and congestion.

[0005] Therefore, there is an urgent need for a vehicle traffic organization method for the closed, high-density operating environment of ports, which can uniformly model and dynamically allocate traffic resources in the dimensions of roads and time, realize accurate cargo flow matching between vehicles arriving at loading and unloading handover points and loading and unloading windows, and at the execution level, have the ability to predict head-on conflicts and control vehicle speed-lane joint control, prevent congestion from the source, reduce vehicle idling and queuing time, and improve overall transport capacity efficiency. Summary of the Invention

[0006] To address the aforementioned shortcomings of existing port vehicle scheduling technologies, the present invention aims to provide a spatiotemporal path planning method and system for port cargo flow matching. This method solves problems such as road resource competition, mismatched loading and unloading windows, frequent head-on conflicts, and easy spread of congestion in the closed, high-density operation environment of ports. It enables dynamic allocation and collaborative optimization of vehicle spatiotemporal right-of-way, seamlessly connecting transportation links with the timing of three-dimensional loading and unloading operations, improving the continuity of cargo flow handover, reducing capacity waste, and improving the overall operational efficiency of the port logistics system.

[0007] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0008] A spatiotemporal path planning method for port cargo flow matching, specifically designed for the closed, high-density operational environment of port container logistics, includes the following steps:

[0009] S1. Abstract the port roads into a directed spatiotemporal graph and establish a spatiotemporal resource library oriented towards road segments and time slots;

[0010] S2. Collect vehicle status, loading and unloading equipment operation status, and road traffic status, and fuse the data to form a related dataset;

[0011] S3. Define the passage rights of vehicles in the time slot of the road segment as spatiotemporal right-of-way, and solve the right-of-way sequence of each vehicle based on the spatiotemporal resource library and the associated dataset.

[0012] S4. Generate vehicle spatiotemporal paths under the constraints of the right-of-way sequence, and trigger dynamic replanning for staggered peak hours when it is predicted that loading and unloading windows are unreachable, the risk of queuing at handover points increases, or the conflict of channel occupation intensifies.

[0013] S5. During the path execution phase, predict head-on conflicts and make joint adjustments to vehicle speed and lanes to ensure that the adjustment actions are consistent with the right-of-way allocated in steps S3 and S4 and the planned path.

[0014] On the other hand, the present invention also provides a spatiotemporal path planning system for port cargo flow matching, used to execute the aforementioned spatiotemporal path planning method for port cargo flow matching. The system includes: an onboard sensing and communication unit, a roadside monitoring unit, a loading and unloading equipment status monitoring unit, and a central dispatch and control unit; wherein,

[0015] The vehicle-mounted sensing and communication unit is used to upload vehicle status in real time, receive dispatch instructions, complete vehicle speed adjustment and smooth lane switching, and perform conflict avoidance actions.

[0016] The roadside monitoring unit is used to monitor road occupancy status, traffic flow and potential conflict points in real time. At the same time, it can identify the container ID to prioritize high-value goods. The data update frequency is no less than 2Hz and is uploaded to the central dispatch terminal in real time.

[0017] The loading and unloading equipment status monitoring unit is deployed on the loading and unloading equipment to collect real-time dynamic operation time, calibrate loading and unloading operation time windows, and periodically report operation progress.

[0018] The central dispatch control unit is deployed in the port dispatch center. It is built on an edge computing architecture, equipped with a multi-core processor and a graphics processing acceleration module, and runs customized optimization algorithms and machine learning models.

[0019] Compared with the prior art, the beneficial effects of the present invention are:

[0020] 1. Achieve proactive congestion prevention and significantly reduce idling time: This invention abstracts port area road resources into spatiotemporal resource units of "road segment-time slot". Through unified modeling and collaborative allocation of spatiotemporal right-of-way, traffic constraints and staggered peak arrangements are completed before vehicles enter bottleneck road segments. This mechanism avoids multiple vehicles competing for the same road segment. Experimental verification shows that it can reduce vehicle idling time by 28%-38%, effectively suppressing the generation and spread of local congestion.

[0021] 2. Strengthen the coordination of transportation and loading / unloading operations to improve the matching degree of cargo flow: Introduce the dynamic operation time of loading and unloading equipment and the loading / unloading time window as core constraints in the allocation of time and space right-of-way and route planning. This ensures that the arrival time of vehicles at the handover point is precisely matched with the loading and unloading operation rhythm, reducing vehicle waiting and equipment idleness. On average, the loading and unloading time gap is reduced by 15 seconds per task, improving the timeliness and continuity of cargo flow handover.

[0022] 3. Effectively avoid head-on conflicts and improve operational safety: For the problem of head-on conflicts in the bottleneck section of the port area, the vehicle status is obtained through V2V / roadside perception, the head-on risk is predicted in advance based on the extended collision time index, and the speed-lane joint decision-making is adopted to achieve active avoidance, controlling the conflict handling time within 8 seconds, preventing local conflicts from evolving into long-term channel blockages, and significantly reducing the frequency of conflicts.

[0023] 4. Refined utilization of road resources to improve transport capacity efficiency: By explicitly characterizing the available capacity of road segments in different time slots and dynamically updating the remaining capacity in combination with the heavy load factor, the limited road resources in the port are maximized under the premise of ensuring safety constraints, avoiding the waste of traffic capacity caused by conservative release or disorderly competition, thereby improving the port's transport capacity efficiency by 18%–25%.

[0024] 5. Adaptable to dynamic operating conditions and highly robust: Supported by multi-source real-time data, the channel occupancy rate is predicted through gradient boosting decision trees. When operational disturbances or congestion risks occur, dynamic replanning is triggered at the second to ten-second level. It can quickly respond to typical dynamic operating conditions of ports such as shipping schedule fluctuations and local channel restrictions, and continuously maintain the accessibility of loading and unloading windows and the matching constraints of cargo flow.

[0025] 6. High feasibility and easy deployment: The method of this invention is based on the existing GIS data, IoT sensor network and 5G communication system of the port. It can be directly embedded into the port's automated scheduling software system without the need for large-scale transformation of the port's existing hardware facilities. It has good feasibility and industry promotion value.

[0026] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the overall process of the spatiotemporal path planning method for port cargo flow matching according to the present invention;

[0028] Figure 2 This is a schematic diagram of the directed spatiotemporal graph structure of the port road network spatiotemporal model according to the present invention;

[0029] Figure 3 This is a schematic diagram of the road segment-time slot occupancy status and remaining capacity update in the spatiotemporal resource library according to the present invention;

[0030] Figure 4 This is a schematic diagram of the spatiotemporal right-of-way collaborative allocation and peak-shifting dynamic replanning according to the present invention;

[0031] Figure 5 This is a schematic diagram of head-on conflict and speed-lane joint adjustment according to the present invention. Detailed Implementation

[0032] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings, so as to more clearly understand the purpose, features and advantages of this invention. It should be understood that the embodiments shown in the drawings are not intended to limit the scope of this invention, but are only for illustrating the essential spirit of the technical solutions of this invention. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.

[0033] Unless the context requires otherwise, throughout the specification and claims, the word “comprising” and its variations, such as “including” and “having”, shall be understood to have an open, inclusive meaning, that is, to be interpreted as “including, but not limited to”.

[0034] Throughout this specification, references to "an embodiment" or "an embodiment" indicate that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. Therefore, the appearance of "in an embodiment" or "an embodiment" in various places throughout the specification does not necessarily refer to the same embodiment. Furthermore, a particular feature, structure, or characteristic may be combined in any manner in one or more embodiments.

[0035] The singular forms “a” and “the” used in this specification and the appended claims include plural references unless otherwise expressly stated herein. It should be noted that the term “or” is generally used to mean “and / or” unless otherwise expressly stated herein.

[0036] In the following description, in order to clearly demonstrate the structure and working method of the present invention, a number of directional terms will be used. However, terms such as "front", "back", "left", "right", "outside", "inside", "outward", "inward", "up", and "down" should be understood as convenient terms and not as limiting terms.

[0037] The implementation details of the embodiments of the present invention will be described in detail below with reference to the accompanying drawings. The following content is only for the convenience of understanding the implementation details and is not necessary for implementing this solution.

[0038] Example 1

[0039] This invention provides a spatiotemporal path planning method for port cargo flow matching. This method is specifically designed for the closed, high-density operating environment of port container logistics. The process of this method is as follows: Figure 1As shown, the core of this invention revolves around the collaborative matching needs of "cargo flow - loading and unloading - roads - vehicles". It constructs a full-process collaborative system from spatiotemporal modeling, multi-source perception, right-of-way allocation, path planning to execution control, so as to realize the dynamic allocation and collaborative optimization of vehicle spatiotemporal right-of-way. The following describes the invention in further detail with specific embodiments.

[0040] S1. Construct a spatiotemporal model of the port road network;

[0041] Using dedicated port geographic information system (GIS) software, the closed road network inside the port is abstracted into a directed spatiotemporal graph G=(V,E,T), such as... Figure 2 The spatiotemporal network model diagram shown employs a combined "spatial node—road segment—time progression" approach to represent the coupling relationship between the port road network and time resources. Vehicles do not simply select roads on a two-dimensional plane; instead, they generate executable spatiotemporal paths according to a combination of "node—road segment—time slot," while satisfying departure time, road capacity, and loading / unloading coordination constraints. Right-of-way conditions limit the authorization status of vehicles entering corresponding road segments at a specified initial time, while capacity constraints limit the maximum number of vehicles a road segment can accommodate within a given time slot. When a vehicle is assigned to road segments such as E1 or E2, the system synchronously verifies whether its occupied time falls within the allowed time slot range and whether it overlaps with other vehicles. If overlapping occurs or the deviation from the target loading / unloading window exceeds the limit, the vehicle's entry time is readjusted or an alternative road segment is selected.

[0042] V represents the node set, which includes road intersection nodes, container loading and unloading points, and temporary buffer zone nodes. The node coordinates are represented using the port's local coordinate system, with a positioning accuracy of 0.05m.

[0043] E represents an edge set, where each edge corresponds to a dedicated vehicle lane within the port, specifying the lane length, width, and capacity limits. Lane length is in meters; lane width is also in meters, supporting 1-3 lanes, but taking into account the 2.5-meter width limit for container vehicles; capacity limits... This refers to the maximum number of container vehicles that can pass at the same time. The default value is 1.2 times the width of the passage, and it can be dynamically adjusted according to the heavy-duty vehicle factor with a container weight exceeding 20 tons.

[0044] T represents the time dimension, discretizing continuous time into a time slot Δt, which is 3 seconds. This parameter is specifically optimized for high-density port operations and can be dynamically adjusted according to real-time traffic flow, with an adjustment range of 2 to 6 seconds.

[0045] Based on historical operation databases and real-time IoT (Internet of Things) sensor data on occupancy status, the occupancy rate, remaining capacity, and loading / unloading coordination factors of each spatiotemporal unit (v, e, t) are dynamically updated to form a spatiotemporal resource database. The historical operation database stores vehicle trajectories, loading / unloading records, and traffic flow data over a past period, using a relational database. The occupancy status collected by real-time IoT sensors uses LiDAR point cloud processing algorithms to detect vehicle density and container stacking interference, considering an occupancy rate exceeding 70% as saturation, and incorporating the influencing factors of the loading / unloading window. The spatiotemporal resource database is stored in the server's memory cache and refreshed at a frequency of 2Hz. The relationship between the occupancy status of a road segment and time slot and the update of remaining capacity is as follows: Figure 3 As shown. Figure 3 The spatiotemporal resource database update relationship is represented by a "road segment - time slot - state variable" structure. Horizontally, different road segments correspond to different locations, and vertically, consecutive time slots correspond to different locations. Each spatiotemporal unit characterizes the resource status of a specific road segment within a specific time period. Each road segment records its occupancy status, occupancy rate, and remaining capacity within its corresponding time slot, and can incorporate heavy-load vehicle factors to reflect changes in channel resource occupancy. When a vehicle enters a road segment, the central dispatch server refreshes the corresponding spatiotemporal unit based on real-time sensor data and historical operation data. When a vehicle leaves the road segment or enters the next time slot, the state variable of the corresponding spatiotemporal unit continues to be updated to reflect the dynamic changes in road resources over time. Through this update process, the system can intuitively determine whether a road segment still has allocable capacity in subsequent time slots and update the remaining capacity of that road segment in the corresponding time slot accordingly, serving as the basis for subsequent spatiotemporal right-of-way allocation and route planning.

[0046] The remaining capacity Update as follows: ;

[0047] In the formula, The basic maximum traffic capacity of port road section e; For vehicles The amount of channel capacity resources occupied within time slot t of road segment e; For heavy-duty vehicles, the heavy-duty vehicle factor is applied when the container weight exceeds 20 tons. .

[0048] S2. Real-time acquisition and fusion of multi-source data;

[0049] On the vehicle side: Autonomous vehicles collect current location, speed, and cargo status via an enhanced GPS module with an integrated IMU. Current location uses local port coordinates, updated at a frequency of 2Hz; speed is measured in km / h with an accuracy of 0.05 km / h; cargo status is categorized as empty, fully loaded, and heavily loaded, detected by an integrated onboard weighing sensor. A weight exceeding 10 tons is considered heavy load, while also taking into account container types such as 20ft and 40ft. The estimated time of arrival (ETA) is predicted using an extended Kalman filter, with a prediction error controlled within 5 seconds. All data is transmitted to the central dispatch server via the onboard 5G communication unit (latency <10ms); the ETA prediction formula is as follows: ,in For Kalman gain, For measured values, for k Time-based observation matrix.

[0050] On the loading and unloading equipment side: Quay cranes and yard cranes collect dynamic time-consuming data through the OPC UA interface of the loading and unloading equipment monitoring system, mainly including container lifting time and container turning time. The average container lifting time is 10 to 25 seconds, which is affected by the container weight and the load on the robotic arm, and is estimated using a support vector machine (SVM) model; the average container turning time is 15 to 35 seconds, and the impact of yard density needs to be considered.

[0051] Roadside: Roadside IoT sensor network collects traffic flow data and potential conflict points. Traffic flow data is acquired through heavy-load vehicle counters and updated every 30 seconds; potential conflict points include scenarios such as loading and unloading lane intersections, which are detected by integrating LiDAR and RFID scanning, with a detection range of 80 meters and a resolution of 0.05 meters. It can also identify container IDs to prioritize the passage of high-value goods.

[0052] All data is transmitted to the central dispatch server via a dedicated 5G network at the port. The central dispatch server uses an extended particle filter to fuse the multi-source data to suppress noise and anomalies in the complex port environment, forming a vehicle-task-road-loading / unloading associated dataset. The extended particle filter has 2000 particles, specifically optimized for the multi-interference environment of the port. The associated dataset is stored in a binary serialization format and includes information such as vehicle ID, task ID, road segment ID, and loading / unloading window. Simultaneously, the central dispatch server writes the target loading / unloading point and loading / unloading window corresponding to the task as cargo flow matching constraint information into this associated dataset.

[0053] S3, Spatiotemporal right-of-way coordination allocation;

[0054] The passage permission of autonomous vehicles in a time slot t of a certain road segment e is defined as spatiotemporal right-of-way, which is divided into exclusive mode and shared mode. In exclusive mode, the permission value is 1, which is applicable to heavy-load vehicles. In shared mode, the permission value is 1 / N, where N is the number of vehicles at the same time, N ≤ channel capacity, and the adjustment is based on loading and unloading priority. The spatiotemporal right-of-way is represented by a passage authorization variable. Road capacity constraints must be met. .

[0055] An optimization algorithm is run on the central scheduling server to construct a mixed-integer programming model with the objective function being: ,in Let be the idling time of vehicle i. For conflict penalties, For the goods flow matching deviation, among which For vehicle arrival time, The loading and unloading time window is defined to minimize the weighted sum of vehicle idling time, conflict penalties, and cargo flow matching deviations. Constraints include time window matching, capacity constraints, and priority constraints. Time window matching requires that the deviation between vehicle arrival time and the loading and unloading time window be less than Δt, and the time consumption estimated by support vector machine (SVM) must be incorporated. Capacity constraints require that the total passage permissions of all vehicles do not exceed the channel capacity, while also considering the occupancy of heavy-load vehicles at a factor of 1.5. Priority constraints stipulate that high-priority tasks (such as emergency exit transfers) have priority passage, with a weight coefficient set to 2.0.

[0056] An improved spatiotemporal graph-based approach was adopted. The algorithm generates candidate right-of-way / path sequences, then combines a linear programming solver to refine and improve the feasibility of these sequences, ultimately assigning a right-of-way sequence to each vehicle. The improvements The algorithm is specifically optimized for the closed, high-density operation scenario of ports. Based on traditional heuristic search, it incorporates Manhattan distance, time offset, loading / unloading coordination costs, and the impact of road congestion into the heuristic function to improve the adaptability of candidate paths to loading / unloading cycle time constraints, bottleneck road occupancy status, and cargo flow matching requirements, thereby improving the generation accuracy of candidate right-of-way / path sequences. The linear programming solver is used to perform feasibility and refinement processing on candidate sequences under capacity constraints, time window constraints, and priority constraints, with a single-vehicle solution time preferably not exceeding 0.5 seconds. For easily congested sections such as the leading main thoroughfare, the maximum number of vehicles entering within the same time slot is limited to 1-2 vehicles to achieve advance flow control; and loading / unloading operation sequences are prioritized for integration. For example, if the loading / unloading window is […]. , For low-priority vehicles, a dynamic buffer is added to ensure seamless connection between transportation and loading / unloading, while outputting the right-of-way sequence for each vehicle; the duration of this dynamic buffer is 3 to 10 seconds, determined based on support vector machine (SVM) prediction. The spatiotemporal right-of-way collaborative allocation and peak-shifting dynamic replanning relationship are as follows: Figure 4 As shown, specifically, candidate right-of-way / path sequences are first formed based on data such as vehicle status, task information, road status, and loading / unloading windows. Then, under the action of a linear programming solver, feasibility and refinement processes are completed under capacity constraints, time window constraints, and priority constraints, outputting the right-of-way sequence corresponding to each vehicle. The subsequent branches in the diagram represent the triggering logic for dynamic replanning. That is, when an increased congestion risk, unreachable loading / unloading windows, or rising queuing risks at handover points are predicted, the overall control objective is not directly changed. Instead, based on the existing right-of-way and path results, adjustments are made to the timing of vehicles entering key channels or handover points, and the passage routes of bottleneck sections, through time or spatial offsets, thereby achieving staggered and off-peak dynamic replanning.

[0057] S4. Spatiotemporal path planning for cargo flow matching;

[0058] Under the aforementioned right-of-way sequence constraints, a variant of the Bellman-Ford algorithm is used to generate the spatiotemporal paths of vehicles, and the path weights can be expressed as follows: ,in, Let e ​​be the length of road segment e. As the reference distance, As a congestion factor, The current weight of the container being transported by vehicle i. Based on the standard box weight, This is a normalized loading and unloading coordination cost term. By introducing this normalized loading and unloading coordination cost term and a congestion factor during the Bellman-Ford relaxation process, iterative updates of path costs for port scenarios are achieved, thereby improving the accuracy of characterizing loading and unloading cycle time constraints and bottleneck section occupancy status. To facilitate engineering implementation and enhance authorization stability, the weighting coefficients... , , , All are non-negative coefficients, among which The value used to characterize the importance of the distance factor is preferably in the range of 0.20 to 0.35; The value used to characterize the importance of congestion factors is preferably in the range of 0.25 to 0.40; The value used to characterize the importance of the load-bearing factor is preferably in the range of 0.10 to 0.20; The value used to characterize the importance of loading and unloading coordination costs is preferably in the range of 0.20 to 0.30. When the congestion factor... A replanning is triggered when the value is greater than 0.6. The weight coefficients are calibrated offline using gradient descent based on historical port operation data. Specifically, a calibration loss function is constructed using the total actual vehicle running time, congestion penalty, and loading / unloading window deviation. The weight coefficients are iteratively updated on the training set, and the set of parameters that minimizes the overall loss is selected on the validation set as the final calibration result.

[0059] A staggered peak and off-peak mechanism is introduced: This mechanism includes time offset and spatial offset. The time offset is used to adjust the start time of vehicles entering key channels or loading and unloading handover points. The offset range is controlled within 2-15 seconds and incorporates loading and unloading prediction. The spatial offset is used to select alternative channels to bypass bottleneck sections. The core is to use the original planned channel as the benchmark and select alternative channels that meet the requirement of an increase in length of less than 15% and priority to avoid the loading and unloading area as the offset path. The value of the spatial offset is the difference in path length between the alternative channel and the original planned channel. At the same time, it is necessary to combine the channel attributes and real-time occupancy status in the port road network spatiotemporal model to ensure the feasibility and efficiency of alternative channels.

[0060] A gradient boosting decision tree (GBDT) machine learning model was used to predict peak traffic flow. The training dataset consisted of historical port traffic and loading / unloading records, with features including time, number of vehicles, container weight, and weather. The model had 150 trees and a depth of 8, with a prediction duration of 45 seconds and a prediction accuracy greater than 90%. If the prediction showed a channel density greater than 0.7 vehicles / channel, indicating congestion, or if the prediction would result in unreachable loading / unloading windows or a significantly increased risk of queuing at handover points, dynamic route replanning would be initiated. The replanning iterations would not exceed 3 times, prioritizing the protection of high-value container tasks. After replanning, route instructions were issued via the vehicle communication unit. The instructions used a custom XML file format, containing information about path points, timestamps, and loading / unloading synchronization points.

[0061] S5. Speed ​​and lane decision-making to avoid head-on conflicts.

[0062] In the path execution phase, head-on collisions are predicted and speed-lane joint adjustments are made to ensure that the execution-level adjustments are consistent with the right-of-way and path in steps S3 / S4. In this invention, head-on collisions specifically refer to traffic congestion caused by two or more transport vehicles encountering each other in bottleneck sections such as narrow passages, one-way passages, and loading / unloading area intersections within the closed-loop environment of an automated container terminal. Due to the limited bypass capacity and space of the port roads, passing or yielding is impossible, resulting in traffic jams. This is distinct from conventional following or rear-end collisions. Such collisions easily lead to prolonged vehicle idling and, if not handled promptly, can quickly escalate into localized traffic jams or even chain reactions within the road network, affecting the pace of port loading / unloading operations and overall logistics efficiency. The relationship between head-on collisions and speed-lane joint adjustments is as follows: Figure 5As shown, the closed-loop relationship of head-on collision handling at the execution level is as follows: First, the relative distance and relative speed of the oncoming vehicle are obtained; then, based on the extended collision time index, it is determined whether there is a head-on collision risk. If there is a head-on collision risk, vehicle speed adjustment or lane switching is executed sequentially based on lane availability, and the conflict handling result is fed back to the vehicle path execution process. This diagram mainly corresponds to the sequential relationship between V2V / roadside perception, TTC threshold judgment, adaptive PID speed adjustment, and lane switching decision in this embodiment. Specifically, as follows:

[0063] During the path execution phase, the vehicle's onboard unit acquires the relative distance d and relative speed Δv with oncoming vehicles based on V2V communication or roadside perception, and extends the collision time index. Predicting head-on conflicts, a head-on risk is determined when TTC < 3 seconds. The coverage range of V2V communication and roadside sensing is 150 meters, with a latency of no more than 20ms.

[0064] For scenarios with a risk of head-on collisions, adaptive PID control is used to fine-tune the target vehicle speed. The target speed is calculated as current speed - k × Δv, where k can be dynamically adjusted between 0.3 and 0.8, and the vehicle speed limit is 3-25 km / h. Simultaneously, lane-switching decisions are performed when lane availability constraints are met, with a switching time of <2 seconds. Cubic spline curves are used to smooth the trajectory, avoiding interference with loading and unloading operations and ensuring conflict resolution time is <8 seconds. Conflict logs are recorded in real-time to the central dispatch server, including time, location, vehicle ID, and loading / unloading association information, for subsequent optimization of the Gradient Boosting Decision Tree (GBDT) model. The PID control is... In the formula .

[0065] Through the above steps, the dynamic allocation and coordinated optimization of vehicle time and space right-of-way are achieved, while emphasizing seamless connection with the timing of three-dimensional loading and unloading operations to prevent congestion and reduce transportation capacity waste.

[0066] In some implementations, to realize the above-mentioned spatiotemporal path planning method for port cargo flow matching, and to ensure the efficient implementation of steps such as spatiotemporal modeling, right-of-way allocation, path planning, and conflict avoidance in the method, and to adapt to the actual working conditions of closed roads and high-concurrency operations in automated container terminals, this spatiotemporal path planning system is specially designed. This system can carry the functions of each link of the method, and through the collaborative work of multiple modules, the technical ideas of the method are transformed into a feasible hardware and software combined execution solution, realizing intelligent management and control of the entire process of port vehicle passage.

[0067] This system is applicable to the traffic organization and congestion prevention of transport vehicles on closed roads in automated container terminals. It consists of four main modules: an onboard sensing and communication unit, a roadside monitoring unit, a loading and unloading equipment status monitoring unit, and a central dispatch and control unit. Each unit achieves low-latency data interaction and command issuance via a dedicated 5G network at the port, completing the entire process of port vehicle spatiotemporal right-of-way allocation, route planning, and conflict avoidance management. It is adapted to the high-density, enclosed operational conditions of ports and can be seamlessly integrated into existing port automated dispatch systems. This spatiotemporal route planning system for port cargo flow matching includes:

[0068] I. Vehicle-mounted sensing and communication unit

[0069] Equipped in port container transport vehicles, this system includes an enhanced GPS+IMU, a 5G communication module, status acquisition sensors, a V2V communication module, and an adaptive PID controller. The enhanced GPS+IMU has a positioning accuracy of no more than 0.5 meters and an update frequency of 2Hz; the 5G communication module has a latency of no more than 10ms; the status acquisition sensors collect vehicle speed and cargo status; and the V2V communication module has a monitoring range of 150 meters and a latency of no more than 20ms. This onboard sensing and communication unit can upload vehicle status in real time, receive dispatch instructions, and perform fine-tuning of vehicle speed and smooth lane switching (lane switching time no more than 2 seconds). It also executes conflict avoidance actions, corresponding to the vehicle status acquisition, path execution, and conflict avoidance stages in the corresponding method.

[0070] II. Roadside Monitoring Unit

[0071] Deployed at key locations such as port intersections and bottleneck sections, the system comprises a LiDAR + camera fusion module, an RFID identification module, and a traffic flow statistics module. The LiDAR + camera fusion module has a detection range of 100 meters and a resolution of 0.05 meters. This roadside monitoring unit can monitor road occupancy, vehicle flow, and potential conflict points in real time. It also prioritizes high-value goods by identifying container IDs, updating data at a frequency of no less than 2Hz and uploading it to the central dispatch center in real time. This provides core roadside data support for the spatiotemporal modeling and congestion prediction methods employed in the methodology.

[0072] III. Loading and Unloading Equipment Status Monitoring Unit

[0073] Deployed on loading and unloading equipment such as quay cranes and yard cranes, it includes operation status sensors, OPC UA standardized communication interfaces, and dynamic time consumption calculation modules; it collects the dynamic time consumption of operations such as container lifting / turning in real time, calibrates the loading and unloading operation time window, and provides operation progress feedback in 1-2 second cycles, providing loading and unloading collaborative constraint data for the allocation of spatiotemporal right-of-way and cargo flow matching in the method, and ensuring seamless connection between transportation and loading and unloading rhythms.

[0074] IV. Central Dispatch and Control Unit

[0075] This is the core processing unit of the device, deployed in the port dispatch center. Built on an edge computing architecture, it is equipped with a multi-core processor and a graphics processing acceleration module, running customized optimization algorithms and machine learning models. Its core functions correspond one-to-one with the core steps of the planning method described above, mainly including multi-source data fusion and noise reduction, port road directed spatiotemporal graph modeling, spatiotemporal right-of-way collaborative allocation, cargo flow matching-oriented spatiotemporal path planning and peak-shifting replanning, head-on conflict prediction and avoidance decision-making, and dynamic adjustment throughout the entire process. It can respond to sudden congestion or operational disturbances in the port area within seconds, prioritizing high-value container transportation tasks and ensuring the efficient implementation of the planning method. The time slot for port road directed spatiotemporal graph modeling is 2 to 6 seconds, with a default setting of 3 seconds.

[0076] This embodiment proposes a spatiotemporal path planning method specifically designed for the closed, high-density operational environment of port container logistics. The core of this method revolves around constructing a full-process collaborative system based on the collaborative matching needs of "cargo flow—loading and unloading—roads—vehicles." This is achieved by sequentially building a spatiotemporal model of the port road network containing node sets, edge sets, and time dimensions, and dynamically updating the spatiotemporal resource library. Real-time collection and fusion of multi-source data from vehicle, loading / unloading equipment, and road sides forms a correlated dataset. Spatiotemporal right-of-way is defined, and a mixed-integer programming model is constructed to complete the collaborative allocation. Under the constraint of right-of-way sequence, a variant of the Bellman-Ford algorithm, a peak-shifting mechanism, and a laddering mechanism are combined. The method involves five steps: improving the decision tree model to perform spatiotemporal path planning and dynamic replanning for cargo flow matching; predicting head-on conflicts based on extended collision time indices; and achieving joint adjustment of vehicle speed and lanes through adaptive PID control and lane switching decisions. This enables the dynamic allocation and collaborative optimization of vehicle spatiotemporal right-of-way. Simultaneously, a spatiotemporal path planning system composed of four modules—vehicle-mounted perception and communication, roadside monitoring, loading and unloading equipment status monitoring, and central dispatch control—was designed. Each module interacts with low latency via a dedicated 5G network at the port, implementing the method as a hardware and software integrated execution solution that is compatible with the port's existing automated dispatch system.

[0077] This method and system, through a full-process design including spatiotemporal resource modeling, multi-source data fusion, dynamic allocation of right-of-way, intelligent path planning, and proactive conflict avoidance, achieves refined utilization of road resources and deep collaboration of loading and unloading operations in the closed, high-density operation environment of ports. It proactively prevents congestion from a mechanism perspective, effectively reduces vehicle idling time and the frequency of head-on conflicts, and improves the timeliness and continuity of cargo flow. At the same time, it can quickly respond to dynamic operating conditions such as port shipping schedule fluctuations and channel restrictions. Moreover, it is based on the existing port hardware system, requiring no large-scale transformation, and has high engineering feasibility and industry promotion value, ultimately significantly improving the overall capacity efficiency of the port logistics system.

[0078] Example 2

[0079] This embodiment applies the spatiotemporal path planning method for port cargo flow matching from Embodiment 1 to the Yangshan Phase IV Automated Container Terminal in Shanghai Yangshan Port. This terminal operates in a closed, automated environment, with transport vehicles primarily consisting of self-driving trucks. These trucks shuttle between the quay crane front operating area and the yard operating area, traversing multiple main roads, intersections, and local bottleneck sections, while maintaining coordinated operational rhythms with quay cranes, yard cranes, and other automated loading and unloading equipment. During peak operating hours, approximately 40-60 transport vehicles operate simultaneously within the port area, with an average task cycle of 18-25 minutes per vehicle. When ships berth in concentrated areas or local access is restricted, vehicles are prone to queuing, idling, and oncoming conflicts in the front operating areas and narrow sections.

[0080] In this embodiment, the system performs a unified modeling of the internal road network of Yangshan Phase IV Terminal based on existing GIS data of the port. Road intersections, quay crane front junctions, yard operation points, and buffer areas are defined as network nodes. Dedicated transport channels within the port are abstracted as directed road segments, and each road segment is labeled with its physical length (80-350 meters), the number of permitted lanes (1-2 lanes), and its designed capacity (20-45 vehicles / hour). Continuous operating time is discretized into fixed-length time slots, with the time slot length set to 3 seconds, to describe the occupancy of road resources by vehicles within a specific time period, thereby forming a computable spatiotemporal network model.

[0081] During system operation, transport vehicles report their real-time location, speed, and cargo status at a frequency of 2Hz via onboard positioning and communication modules; the quay crane and yard crane control systems provide feedback on the progress of container handling operations at a cycle of 1-2 seconds; and roadside data collection points update lane occupancy and vehicle density at a cycle of 5 seconds. This data is then integrated and processed in the central dispatch system to form a unified operational status that includes vehicle location, road load, and loading / unloading rhythm, dynamically reflecting the real-time operational situation of the port area.

[0082] As the core decision-making object of the scheduling system, spatiotemporal right-of-way is calculated uniformly within each scheduling cycle (10-20 seconds) based on the current transportation tasks to be performed and the vehicles en route. Under the premise of meeting road capacity constraints, vehicle safety intervals, and loading and unloading operation time windows, the system allocates available spatiotemporal right-of-way to each vehicle. For congestion-prone sections such as leading main roads, the system limits the maximum number of vehicles entering in the same time slot to 1-2 vehicles to avoid concentrated vehicle influx leading to queuing and idling, thereby achieving early constraint on potential congestion.

[0083] After obtaining the spatiotemporal right-of-way allocation results, the system generates a corresponding spatiotemporal route plan for each transport vehicle. When it is predicted that the occupancy rate of a certain road segment may exceed 70% within the next 30-60 seconds, the system implements staggered passage by delaying vehicle entry time (2-10 seconds) or guiding vehicles to choose alternative lanes, thus keeping the lane load within a controllable range. This method effectively reduces the queuing probability of local road segments while ensuring the basic stability of the task sequence.

[0084] During vehicle operation, the system continuously monitors the relative positions and operational status of vehicles. In high-risk areas such as narrow passages and entrances to one-way road sections, when it predicts that two vehicles are approaching each other and the estimated collision time is less than 3 seconds, the system issues speed adjustment or traffic order adjustment commands in advance, causing one vehicle to slow down or wait briefly, thus avoiding a head-on collision. In this way, the processing time for a single conflict can be controlled within 5 to 8 seconds, preventing the conflict from escalating into a prolonged traffic jam.

[0085] Under simulation and trial operation conditions at Yangshan Phase IV Terminal, the method of this invention reduced the average queue length at the front access road and bottleneck sections from 4-6 vehicles to 1-2 vehicles. Idle waiting time per vehicle task was reduced by approximately 30%. The matching degree between vehicle arrival at the front access road or yard handover point and the operation rhythm of quay cranes and yard cranes was significantly improved, and the frequency of head-on collisions decreased significantly. The results demonstrate that this invention can achieve orderly organization of vehicle traffic resources in a closed, high-density port operation environment, exhibiting good engineering feasibility and application value.

[0086] This embodiment applies a spatiotemporal path planning method and system for port cargo flow matching to the closed-loop automated operation scenario of the Yangshan Phase IV Automated Container Terminal in Shanghai Yangshan Port. During peak hours, 40 to 60 autonomous trucks operate back and forth at this terminal, which is prone to problems such as vehicle queuing, idling, and oncoming conflicts due to concentrated ship berthing. In implementation, a spatiotemporal network model containing nodes and directed road segments is first constructed based on the port's existing GIS data. Time is discretized into 3-second time slots. Then, vehicle data is reported at different frequencies through multi-source acquisition devices on vehicles, loading and unloading equipment, and roadside devices. Data such as status, work progress, and channel load are integrated by the central dispatch system to form a unified operational status. Then, the system allocates time and space rights to vehicles in a dispatch cycle of 10 to 20 seconds. For congested road sections, the system limits the flow of 1 to 2 vehicles per time slot. Based on the right-of-way, a time and space route plan is generated. For road sections with an occupancy rate of more than 70%, staggered passage is achieved through time delays or backup lanes. Finally, during the vehicle execution phase, the system monitors the status of oncoming vehicles in high-risk areas. When the expected collision time is less than 3 seconds, the system issues speed / traffic order adjustment instructions to control the conflict resolution time to 5 to 8 seconds.

[0087] The method and system have achieved significant results in the simulation and trial operation of Yangshan Port Phase IV. The average queue length of vehicles in the front channel and bottleneck section has been reduced from 4-6 vehicles to 1-2 vehicles. The idling waiting time of vehicles for a single task has been reduced by about 30%. The matching degree between the arrival of vehicles at the handover point and the loading and unloading rhythm of the quay crane and yard crane has been greatly improved. At the same time, the frequency of head-on collisions has been significantly reduced. The orderly organization of vehicle traffic resources in the closed high-density operation environment of the port has been effectively realized, verifying that the technical solution has good engineering feasibility and industry promotion and application value.

[0088] This invention belongs to the field of port intelligent transportation and logistics scheduling technology. Addressing the problems of existing scheduling methods in the closed, high-density operation environment of automated container terminals, such as road resource competition, mismatched loading and unloading windows, frequent head-on conflicts, and easy congestion spread, this invention proposes a spatiotemporal path planning method and supporting system for port cargo flow matching. The method revolves around the collaborative matching needs of "cargo flow—loading and unloading—roads—vehicles," and sequentially completes five core steps: constructing a spatiotemporal model of the port road network, collecting and fusing multi-source data from vehicles / loading and unloading equipment / roadside, coordinating spatiotemporal right-of-way allocation, spatiotemporal path planning and dynamic replanning for cargo flow matching, and joint speed and lane decision-making for head-on conflicts. It abstracts port road resources into spatiotemporal resource units of "road segment-time slot," integrates dynamic time constraints of loading and unloading equipment, and achieves dynamic allocation of spatiotemporal right-of-way and intelligent path planning through improved algorithms and machine learning models. Furthermore, it proactively avoids head-on conflicts based on V2V / roadside perception and adaptive PID control. The supporting system consists of vehicle-mounted sensors... Composed of four major modules—information communication, roadside monitoring, loading and unloading equipment status monitoring, and central dispatch control—this invention achieves low-latency interaction via a dedicated 5G network for the port, allowing for seamless integration into existing automated dispatch systems. The method has already undergone simulation and trial operation verification at the Yangshan Phase IV automated container terminal in Shanghai Yangshan Port. This invention proactively prevents congestion by reducing vehicle idling time by 28%–38%, enhancing the coordination of transportation and loading / unloading operations, reducing the average loading / unloading time gap by 15 seconds per task, effectively avoiding head-on conflicts and controlling processing time to within 8 seconds, and improving port capacity efficiency by 18%–25% through refined utilization of road resources. It also enables rapid response to dynamic port conditions such as shipping schedule fluctuations. Furthermore, it is implemented based on existing port hardware, requiring no large-scale modifications, demonstrating high engineering feasibility and industry promotion value. In practical applications, it has significantly reduced the average queue length and idling time by approximately 30% on bottleneck sections of Yangshan Port Phase IV, significantly improving cargo flow matching and reducing the frequency of conflicts.

[0089] Although the present invention has been described in detail with reference to the accompanying drawings and preferred embodiments, the invention is not limited thereto. Various equivalent modifications or substitutions can be made to the embodiments of the invention by those skilled in the art without departing from the spirit and essence of the invention. Such modifications or substitutions should all fall within the scope of the invention, or any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the invention should be covered within the protection scope of the invention. Therefore, the protection scope of the invention should be determined by the scope of the claims.

Claims

1. A space-time path planning method for port cargo flow matching, characterized in that, The method is specifically designed for the enclosed, high-density operating environment of port container logistics and includes the following steps: S1. Abstract the port roads into a directed spatiotemporal graph and establish a spatiotemporal resource library oriented towards road segments and time slots; S2. Collect vehicle status, loading and unloading equipment operation status, and road traffic status, and fuse the data to form a related dataset; S3. Define the passage rights of vehicles in the time slot of the road segment as spatiotemporal right-of-way, and solve the right-of-way sequence of each vehicle based on the spatiotemporal resource library and the associated dataset. S4. Generate vehicle spatiotemporal paths under the constraints of the right-of-way sequence, and trigger dynamic replanning for staggered peak hours when it is predicted that loading and unloading windows are unreachable, the risk of queuing at handover points increases, or the conflict of channel occupation intensifies. S5. During the path execution phase, predict head-on conflicts and make joint adjustments to vehicle speed and lanes to ensure that the adjustment actions are consistent with the right-of-way allocated in steps S3 and S4 and the planned path. Specifically, defining the passage rights of vehicles within a time slot on a road segment as spatiotemporal right-of-way, and solving the right-of-way sequence for each vehicle based on the spatiotemporal resource library and the associated dataset, includes: (1) The spatiotemporal right-of-way is represented as a vehicle In the road section With time slot Pass authorization variable And satisfy capacity constraints ; (2) Construct a mixed-integer programming model with the objective function as follows: ,in Let be the idling time of vehicle i. For conflict penalties, For the goods flow matching deviation, among which For vehicle arrival time, The loading and unloading time window is defined to minimize the weighted sum of vehicle idling time, conflict penalties, and cargo flow matching deviation. Constraints include time window matching, capacity constraints, and priority constraints. Time window matching requires that the deviation between vehicle arrival time and the loading / unloading time window be less than Δt, and the estimated time consumption by Support Vector Machine (SVM) must be incorporated. Capacity constraints require that the total passage permissions of all vehicles not exceed the channel capacity, while also considering the occupancy of heavy-load vehicles at a factor of 1.

5. Priority constraints stipulate that high-priority tasks have priority passage, with a weight coefficient set to 2.

0. (3) An improved spatiotemporal graph-based approach is adopted. The algorithm generates candidate road rights and path sequences, and combines a linear programming solver to make the candidate sequences feasible and refined, outputting a road rights list for each vehicle. .

2. The method according to claim 1, characterized in that, The abstraction of port roads into a directed spatiotemporal graph and the establishment of a spatiotemporal resource library oriented towards road segments and time slots specifically include: the directed spatiotemporal graph is represented as G=(V,E,T), where: The node set V includes road intersection nodes, container loading and unloading point nodes, and temporary buffer zone nodes; the node coordinates are represented using the port's local coordinate system, with a positioning accuracy of 0.05m. Each edge of edge set E corresponds to a dedicated port vehicle lane, and the lane attributes include at least length, width, and maximum capacity. The passage supports 1–3 lanes and takes into account container vehicle width constraints; The time dimension T discretizes continuous time into time slots. , The value is set to 3 seconds, and can be adaptively adjusted within the range of 2 seconds to 6 seconds based on real-time traffic. Based on the historical job database and real-time sensor occupancy detection results, each spatiotemporal unit in the spatiotemporal resource library is refreshed at an update frequency of 2Hz. The occupancy rate and remaining capacity; where the remaining capacity is updated using the following formula: In the formula, For vehicles exist The amount of space occupied For heavy-duty vehicles, when the container weight exceeds 20 tons... .

3. The method according to claim 1, characterized in that, The process of collecting vehicle status, loading and unloading equipment operation status, and road traffic status, and then fusing the data to form a related dataset, specifically includes: On the vehicle side, an enhanced GPS module and an integrated IMU are used to collect vehicle location, speed, and cargo status. The location update frequency is 2Hz, and the speed resolution is 0.05km / h. Cargo status is categorized into empty, fully loaded, and heavily loaded, with a vehicle weighing more than 10 tons considered heavily loaded. The estimated time of arrival (ETA) is predicted using an extended Kalman filter, with a prediction error controlled within 5 seconds. On the loading / unloading equipment side, dynamic time-consuming data is collected via an OPC UA interface, and a support vector machine model trained on historical samples is used to estimate the dynamic time-consuming data. On the road side, a sensor network is used to collect channel traffic and conflict point information. This sensor network includes at least a combination of LiDAR and cameras for vehicle lane occupancy detection. All data is transmitted to the central dispatch server via the port's 5G network. The central dispatch server uses an extended particle filter to fuse the above multi-source data, forming a vehicle-task-road-loading / unloading associated dataset. The particle number of the extended particle filter is set to 2000.

4. The method according to claim 1, characterized in that, The process involves generating vehicle spatiotemporal paths under the constraints of the right-of-way sequence, and triggering dynamic replanning to stagger peak times and avoid peak-shifting issues when predictions indicate unreachable loading / unloading windows, increased queuing risks at handover points, or intensified lane occupancy conflicts. Specifically, this includes: (1) Under the constraints of the right-of-way sequence, a variant of the Bellman-Ford algorithm is used to generate the spatiotemporal paths of vehicles, and the path weights are expressed as follows: ,in, Let e ​​be the length of road segment e. As the reference distance, As a congestion factor, The current weight of the container being transported by vehicle i. Based on the standard box weight, For the normalized loading and unloading coordination cost item, , , , These are non-negative weighting coefficients, which are calibrated offline using gradient descent based on historical port operation data; whereby... Used to characterize the importance of distance factors Used to characterize the importance of congestion factors Used to characterize the importance of load-bearing factors Used to characterize the importance of loading and unloading coordination costs; when congestion factors A value greater than 0.6 triggers replanning; (2) The staggered peak and staggered positioning mechanism includes time offset and spatial offset, wherein the time offset is used to adjust the time when the vehicle enters the key channel or loading and unloading handover point, and the spatial offset is used to select alternative channels to bypass the bottleneck section. (3) Gradient boosting decision tree (GBDT) is used to predict the channel occupancy rate in the future short window. When the predicted occupancy rate exceeds the threshold, dynamic replanning is triggered, and priority is given to maintaining the accessibility of loading and unloading windows and the continuity of cargo flow.

5. The method according to claim 1, characterized in that, The step of predicting head-on conflicts and making joint adjustments to vehicle speed and lanes during the path execution phase, ensuring that the adjustments are consistent with the right-of-way allocation and planned path in steps S3 and S4, specifically includes: The vehicle's onboard unit acquires the relative distance d and relative speed Δv with oncoming vehicles based on V2V communication or roadside sensing, and extends the collision time index. Predict head-on conflicts; when TTC is less than a preset threshold, it is determined that there is a risk of head-on conflict. To address the risk of head-on collisions, adaptive PID control is used to adjust the target speed of the vehicle. The target speed is calculated as current speed - k × Δv, with the k value dynamically adjusted between 0.3 and 0.

8. Lane switching decisions are executed when the channel availability constraint is met, with a switching time of <2 seconds. Cubic spline curves are used to smooth the trajectory to avoid interfering with loading and unloading operations, ensuring that the conflict resolution time is <8 seconds. The adjustments are consistent with the right-of-way and path in steps S3 and S4.

6. The method according to claim 2, characterized in that, The real-time sensor detects vehicle density and container stacking interference through LiDAR point cloud processing algorithm, considers an occupancy rate of more than 70% as a saturation state, and incorporates the influencing factors of loading and unloading windows.

7. The method according to claim 1, characterized in that, The improvements The algorithm's heuristic function consists of Manhattan distance, time offset, and loading / unloading coordination cost. The linear programming solver solves a single vehicle in no more than 0.5 seconds.

8. The method according to claim 5, characterized in that, The coverage range of the V2V communication and roadside perception is 150 meters, and the delay is no more than 20ms. The conflict log contains time, location, vehicle ID and loading / unloading association information, which is used for subsequent optimization of the gradient boosting decision tree (GBDT) model.

9. A spatiotemporal path planning system for port cargo flow matching, used to execute the spatiotemporal path planning method for port cargo flow matching as described in any one of claims 1 to 8, characterized in that, The system includes: an onboard sensing and communication unit, a roadside monitoring unit, a loading and unloading equipment status monitoring unit, and a central dispatch and control unit; wherein... The vehicle-mounted sensing and communication unit is used to upload vehicle status in real time, receive dispatch instructions, complete vehicle speed adjustment and smooth lane switching, and perform conflict avoidance actions. The roadside monitoring unit is used to monitor road occupancy status, traffic flow and potential conflict points in real time. At the same time, it can identify the container ID to prioritize high-value goods. The data update frequency is no less than 2Hz and is uploaded to the central dispatch terminal in real time. The loading and unloading equipment status monitoring unit is deployed on the loading and unloading equipment to collect real-time dynamic operation time, calibrate loading and unloading operation time windows, and periodically report operation progress. The central dispatch control unit is deployed in the port dispatch center. It is built on an edge computing architecture, equipped with a multi-core processor and a graphics processing acceleration module, and runs customized optimization algorithms and machine learning models.

Images