Cargo transportation real-time path planning method based on artificial intelligence

Abstract

The invention discloses a cargo transportation real-time path planning method based on artificial intelligence, and relates to the technical field of artificial intelligence and intelligent traffic, and the method comprises the following steps: collecting real-time data from a traffic monitoring system, a vehicle-mounted sensor, a road event platform and a weather information system in a transportation process, forming an original data set containing multi-source information; time synchronization and space alignment are carried out on source data in an original data set, and a standardized data frame is constructed to ensure the consistency of different source information in a space-time dimension. According to the method, the accuracy, the real-time performance and the intelligence of path planning are effectively improved through multi-source data consistency identification, weighted fusion and a dynamic re-planning mechanism. The system can timely identify data conflicts, dynamically adjust paths and enhance transportation safety and efficiency, has a self-learning capability, continuously optimizes path strategies and adapts to complex traffic environments.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence and intelligent transportation technology, and in particular to a real-time cargo transportation path planning method based on artificial intelligence. Background Art

[0002] "AI-based real-time route planning for freight transportation" refers to the use of artificial intelligence technologies, particularly algorithms such as machine learning, deep learning, and reinforcement learning, to dynamically calculate the optimal transportation route based on constantly changing real-time information during the freight transportation process (such as road traffic conditions, weather changes, vehicle location, and the urgency of the transportation task). This method breaks through the limitations of traditional static route planning and no longer relies on preset routes or manual intervention. Instead, it autonomously optimizes route selection through real-time analysis of multi-source data to maximize freight transportation efficiency, minimize costs, and enhance the ability to respond to emergencies. This technology is widely used in scenarios such as smart logistics, autonomous freight vehicles, and urban distribution.

[0003] Existing technologies have the following shortcomings: In existing AI-based real-time freight transport routing technology, information conflicts between data sources are not promptly identified and coordinated, which can have serious consequences. For example, when the road traffic status provided by the traffic monitoring system conflicts with the on-site conditions detected by on-board sensors (such as sudden road closures or traffic accidents), if the system lacks an effective multi-source data conflict identification and fusion mechanism, it may misjudge traffic conditions, leading to routing errors, causing vehicles to enter traffic jams or dangerous sections of road, causing transportation delays, and even traffic accidents. This problem is often hidden in the data processing process and difficult to detect in a timely manner. However, once it occurs, it poses a significant threat to the stability and security of the entire logistics system.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0005] The purpose of this invention is to provide an artificial intelligence-based real-time freight transport route planning method. This method effectively improves the accuracy, real-time performance, and intelligence of route planning through multi-source data consistency identification, weighted fusion, and dynamic replanning mechanisms. The system can promptly identify data conflicts and dynamically adjust routes, enhancing transport safety and efficiency. It also possesses self-learning capabilities, continuously optimizing routing strategies and adapting to complex traffic environments, thereby addressing the aforementioned issues in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for real-time cargo transportation path planning based on artificial intelligence, comprising the following steps:

[0007] Collect real-time data from traffic monitoring systems, vehicle sensors, road event platforms, and weather information systems during transportation to form a raw data set containing multi-source information;

[0008] Perform time synchronization and spatial alignment on the data from each source in the original dataset and construct a standardized data frame to ensure the consistency of information from different sources in the temporal and spatial dimensions;

[0009] Set conflict detection rules and identify logical conflicts in standardized data frames based on the rules, including traffic state contradictions, inconsistent road traffic information, and differences in environmental perception;

[0010] The multi-source data fusion module is activated to perform a weighted assessment of conflicting data based on the credibility, real-time nature, and historical error rate of the data sources, generating the most consistent road traffic status judgment results.

[0011] Based on the consistency judgment results, the deep reinforcement learning path planning model is input to dynamically generate the optimal transportation path at the current moment;

[0012] During the path execution process, the system continuously monitors changes in multi-source data, evaluates path reliability in real time, and triggers a dynamic replanning mechanism to deal with path interruptions or sudden changes in the traffic environment.

[0013] Preferably, the collected traffic monitoring system data includes traffic light cycles, road congestion index, and road closure status; vehicle-mounted sensor data includes distance to obstacles ahead, lane status, and real-time speed; road event platform data includes traffic accidents, construction records, and regulatory notices that occurred in the current area within the past 48 hours; and weather information system data includes real-time rainfall, visibility level, surface temperature, and wind speed and direction.

[0014] All data is updated at a frequency of no more than 60 seconds and stored in the local cache for subsequent standardized processing and calls.

[0015] Preferably, time synchronization uses a unified timestamp interpolation algorithm, and spatial alignment relies on a high-precision electronic map to uniformly map all data into a polar coordinate system with the current transport vehicle as the reference point;

[0016] Timestamp interpolation uses spline fitting to estimate the state changes of different systems in the same time period; the spatial position error is controlled within 1 meter, and the Kalman filter is used to dynamically correct the positioning error of the on-board sensors to ensure that all types of information can be analyzed and compared under a unified coordinate and time reference.

[0017] Preferably, the conflict detection rules include the following categories:

[0018] Data inconsistency conflict rules to determine the contradiction between the traffic monitoring clear sign and the vehicle-mounted obstacle distance less than the threshold;

[0019] Logical contradiction and conflict rules to determine when the road event platform shows a closure but the vehicle's sensors do not detect a closure sign;

[0020] Early warning conflict rules issue a warning when any emergency indication exists in multi-source data and the system has not entered the avoidance state.

[0021] Preferably, the multi-source data fusion module constructs a Bayesian scoring model based on the credibility, real-time performance and historical error rate of the data sources;

[0022] Credibility is based on historical accuracy and source stability; real-time performance is determined by the time difference between current data and the standard update time; and the historical error rate is derived from the frequency of misjudgments generated by the system comparing annotated data over the past 30 days.

[0023] After each conflict information is updated through Bayesian probability, the maximum a posteriori estimation mechanism is used to select the optimal passage judgment, and a unified traffic state label is output as one of the input conditions for path planning.

[0024] Preferably, the weighted fusion of multi-source data includes confidence calculation and conflict judgment mechanism, and the specific steps are as follows:

[0025] For each data source, calculate its confidence adjustment value based on its current transmitted data quality and historical performance. The calculation expression is as follows:

[0026]

[0027] Where C i is the initial confidence of the i-th data source, A i is the historical average accuracy of the i-th data source, T i is the current data time deviation, α is the time attenuation coefficient, W i is the weighted confidence of the i-th data source;

[0028] The identified conflict types are divided into several categories, and the information intensity provided by each data source under each type of conflict is defined as I i，k , calculate the weighted total score of all data sources involved in conflict judgment. The calculation expression is as follows:

[0029]

[0030] Where, I i，k is the indication strength provided by the i-th data source for the k-th type of conflict, n is the total number of data sources, S k is the total confidence value of the k-th conflict judgment;

[0031] Then for all confidence sum values ​​S k After comparison, the judgment corresponding to the conflict type k with the highest score is selected as the final road pass judgment result. The judgment logic is as follows:

[0032]

[0033] Where D is the final conflict type label, is a mathematical function.

[0034] Preferably, the deep reinforcement learning path planning model includes a state input layer, a convolutional feature extraction layer, a policy network and a value network dual-branch structure;

[0035] The state input includes the current vehicle position, unified traffic state label, destination coordinates and path history trajectory. The convolution layer extracts significant features in the input space, the policy network outputs the probability distribution of the path set, and the value network evaluates the long-term cumulative benefits of each path.

[0036] Preferably, the path optimization module is based on a multi-factor evaluation function, comprehensively considering traffic efficiency, safety risk and time matching. In the path generation process, the optimal path is represented as a set consisting of several discrete path nodes {p a}={p1,p2,……,p m}, p a is the a-th path node. To achieve multi-objective trade-offs, the comprehensive cost score of a single node is calculated. The calculation expression is as follows:

[0037]

[0038] Where c a is at path node p a The cost of passage at a is at path node p a Safety risk value at t a is the path node p a The deviation from the preset time window, ω1 is the travel cost weight, ω2 is the safety risk weight, ω3 is the time window matching weight, f a is the comprehensive cost score of a single node;

[0039] Calculate the cumulative total score of all path nodes. The calculation expression is as follows:

[0040]

[0041] Where m is the total number of path nodes, and F is the total cost score of the entire path;

[0042] Because paths can jump dramatically in real-time, leading to execution cost and route continuity issues, a path stability control mechanism is introduced to constrain path mutation behavior by penalizing the change amplitude between consecutive nodes. The formula is as follows:

[0043]

[0044] Where λ is the path stability regularization term weight, p j is the jth path node, p j-1 is the j-1th path node, ||p j -p j-1 || 2 is the square of the distance between path nodes, and R is the path instability penalty term;

[0045] When finally optimizing the path, a comprehensive objective function is used to simultaneously minimize the travel cost, safety risk, time deviation, and path discontinuity. The formula is as follows: min(F+R), so that the sum of the total path score F and the stability penalty term R reaches the minimum.

[0046] Preferably, during the path execution process, the traffic monitoring system includes a path stability monitoring module and a dynamic replanning module;

[0047] The path stability monitoring module receives the latest data from onboard sensors, traffic monitoring systems, and road event platforms in real time, and calculates the reliability score of the current path based on the following indicators: path node matching, which measures whether the deviation between the current vehicle's actual position and the preset path exceeds a dynamic threshold; path sudden risk index, which is a weighted sum of the number of new abnormal events occurring on the road section and the risk level; path target time offset, which calculates the time difference between the vehicle and the estimated arrival time;

[0048] When any indicator exceeds its defined threshold, the dynamic replanning module is triggered. The current target destination and the previously confirmed passed segments of the path are retained, and the current vehicle position is used as the new starting point. The deep reinforcement learning path planning model is called to recalculate the remaining path.

[0049] The re-planned path must meet the three evaluation dimensions of stability, traffic efficiency and emergency avoidance capability, and automatically replace the original path to ensure that the transportation task is carried out continuously, reliably and efficiently.

[0050] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0051] The present invention effectively improves the accuracy of information perception of the cargo transportation route planning system in the face of emergencies by introducing a multi-source data consistency perception mechanism. Traditional route planning methods usually rely on a single data source and are unable to promptly detect changes in road conditions caused by sudden road closures, traffic accidents, severe weather, etc., which can easily lead to vehicles entering risky sections. This invention collects heterogeneous data from traffic monitoring, vehicle-mounted sensors, road event platforms, and weather systems, and identifies logical conflicts between data through unified spatiotemporal standardization processing and rule systems, thereby fundamentally improving the comprehensiveness and accuracy of information perception. In particular, when there are contradictions in multi-source data, the system can quickly identify the problem and respond to it, avoiding path deviations caused by data distortion, and significantly enhancing the system's environmental adaptability and safety.

[0052] The weighted fusion and confidence calculation strategy of conflicting data proposed in the present invention solves the problem of difficulty in quantifying the reliability differences of multi-source data, and provides a more reliable input basis for the path planning system. In actual applications, there are obvious differences in the update frequency, transmission delay, and historical accuracy of various data sources. If they are not effectively distinguished and processed, it is very easy to misjudge or ignore key events. The present invention establishes a weight evaluation mechanism based on the historical accuracy, current timeliness and stability of the data source, and adopts a Bayesian probability model for data fusion, so that the system can dynamically adjust the trust level of different data and output the traffic status judgment that best conforms to the actual situation. This mechanism not only improves the judgment robustness of the path planning model in complex scenarios, but also significantly reduces the economic losses and safety risks caused by incorrect path decisions.

[0053] The path planning and dynamic re-planning coordination mechanism constructed by the present invention significantly improves the real-time response capability and intelligent level of path selection of the cargo transportation system. Unlike traditional static path planning solutions, the present invention continuously monitors traffic and environmental changes during path execution. When abnormal information that affects traffic efficiency or safety is detected, it can quickly start the local re-planning mechanism to achieve instant optimization of the path. This mechanism not only takes into account the path switching cost, but also introduces historical trajectory and risk density parameters to ensure that the planning results minimize driving risks while ensuring transportation efficiency. The system also has self-learning capabilities and can feed back historical path execution results to the path strategy library for optimizing model parameters and gradually building an optimal path library that adapts to different traffic areas, comprehensively improving the intelligence level and long-term operational stability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0055] Figure 1 This is a flow chart of a method for real-time cargo transportation path planning based on artificial intelligence of the present invention. DETAILED DESCRIPTION

[0056] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0057] The present invention provides Figure 1 The method for real-time cargo transportation path planning based on artificial intelligence includes the following steps:

[0058] Collect real-time data from traffic monitoring systems, vehicle sensors, road event platforms, and weather information systems during transportation to form a raw data set containing multi-source information;

[0059] The traffic monitoring system data collected includes traffic light cycles, road congestion index, and road closure status; vehicle sensor data includes the distance to obstacles ahead, lane status, and real-time speed; road event platform data includes traffic accidents, construction records, and regulatory notices that have occurred in the current area within the past 48 hours; and weather information system data includes real-time rainfall, visibility level, surface temperature, and wind speed and direction.

[0060] All data is updated at a frequency of no more than 60 seconds and stored in the local cache for subsequent standardized processing and calls.

[0061] Perform time synchronization and spatial alignment on the data from each source in the original dataset and construct a standardized data frame to ensure the consistency of information from different sources in the temporal and spatial dimensions;

[0062] Time synchronization uses a unified timestamp interpolation algorithm, while spatial alignment relies on high-precision electronic maps, mapping all data uniformly into a polar coordinate system with the current transport vehicle as the reference point;

[0063] Timestamp interpolation uses spline fitting to estimate the state changes of different systems in the same time period; the spatial position error is controlled within 1 meter, and the Kalman filter is used to dynamically correct the positioning error of the on-board sensors to ensure that all types of information can be analyzed and compared under a unified coordinate and time reference.

[0064] Set conflict detection rules and identify logical conflicts in standardized data frames based on the rules, including traffic state contradictions, inconsistent road traffic information, and differences in environmental perception;

[0065] Conflict detection rules include the following categories:

[0066] Data inconsistency conflict rules to determine the contradiction between the traffic monitoring clear sign and the vehicle-mounted obstacle distance less than the threshold;

[0067] Logical contradiction and conflict rules to determine when the road event platform shows a closure but the vehicle's sensors do not detect a closure sign;

[0068] Early warning conflict rules issue a warning when any emergency indication exists in multi-source data and the system has not entered the avoidance state.

[0069] Each type of rule has an independent priority and response delay threshold to control the timeliness and accuracy of conflict judgment.

[0070] The multi-source data fusion module is activated to perform a weighted assessment of conflicting data based on the credibility, real-time nature, and historical error rate of the data sources, generating the most consistent road traffic status judgment results.

[0071] The multi-source data fusion module builds a Bayesian scoring model based on the credibility, real-timeness and historical error rate of the data sources;

[0072] Credibility is based on historical accuracy and source stability; real-time performance is determined by the deviation between the current data and the standard update time; the historical error rate is derived from the frequency of misjudgments generated by the system comparing annotated data over the past 30 days; after each conflict information is updated through Bayesian probability, the maximum a posteriori (MAP) estimation mechanism is used to select the optimal passage judgment, and a unified traffic status label is output as one of the input conditions for path planning.

[0073] Multi-source data weighted fusion includes confidence calculation and conflict judgment mechanism. The specific steps are as follows:

[0074] For each data source, its confidence adjustment value is calculated based on its current transmitted data quality and historical performance. Each data source i initially has a set confidence C i , represents the credibility level of its source data, such as vehicle-mounted radar, traffic cloud platform or meteorological system. Taking into account the past accuracy of the data source and the acquisition delay of the current data, a confidence correction model with time decay term is introduced. The calculation expression is as follows:

[0075]

[0076] Where C i It is the initial confidence of the i-th data source, reflecting the basic credibility level assigned to the data source in the design phase or empirical analysis, with a value range of 0-1. iis the historical average accuracy of the i-th data source, indicating the degree of match between the predicted information of the data source and the actual results in the past historical records. The value range is 0-1. The higher the value, the more credible the data source is. i is the current data time deviation, which indicates the time interval between the generation time of the currently collected data and the current system time (in seconds). It is a non-negative real number. α is the time decay coefficient, which is used to control the impact of data time delay on confidence. Typical values ​​are 0.01, 0.05, 0.1, etc. The larger the value, the smaller the impact of old data (i.e., the confidence of expired data decreases faster). W i is the weighted confidence of the i-th data source, which is the final confidence score obtained by combining the initial confidence, historical accuracy and time decay. It is used to weight conflicting information. The value range is: real number, usually less than or equal to 1;

[0077] The identified conflict types are divided into several categories (such as road access conflict, weather consistency conflict, etc.), and the information intensity provided by each data source under each type of conflict is defined as I i，k , calculate the weighted total score of all data sources involved in conflict judgment. The calculation expression is as follows:

[0078]

[0079] Where, I i，k It is the indication strength provided by the i-th data source for the k-th type of conflict, indicating the strength of the data source's support for conflict type k, such as 1 for strong support, 0 for no support, and between for partial support. The value range is 0-1. n is the total number of data sources. S k is the total confidence value of the k-th conflict judgment, which is the weighted sum of the judgment strengths of all data sources for this conflict type. It is used to determine the priority of the final conflict type. A higher value indicates that the conflict type is more likely to be true.

[0080] Then for all confidence sum values ​​S k After comparison, the judgment corresponding to the conflict type k with the highest score is selected as the final road pass judgment result. The judgment logic is as follows:

[0081]

[0082] Where D is the final conflict type label, which means selecting the total confidence value S from all conflict types k k The largest, most likely real, conflict category, It is a mathematical function that returns the parameter that maximizes the function value (that is, the maximum index).

[0083] Argmax is a commonly used function in mathematics, known as the "argument of the maximum" function. It is a type of optimization function and is widely used in machine learning, statistics, operations research, artificial intelligence and other fields.

[0084] Specifically, The meaning is: among all possible x values, find the x that makes the function f(x) reach the maximum value. For example, for the value of function f(x) in a certain set, What is returned is the optimal input, not the output result itself (that is, not the maximum value, but the point where the maximum value occurs).

[0085] In path planning or data decision systems, argmax is used to select the optimal solution from multiple candidate solutions. The system calculates the comprehensive confidence score S for each conflict type. k Then use Finding which type of conflict is most likely to exist in the current situation, and thus triggering the corresponding avoidance strategy, not only implements the logic of "selecting the largest", but also provides the mathematical basis for decision-making in intelligent systems.

[0086] This judgment will be converted into a digitally coded pass identifier for direct call and analysis by the path planning module in the next stage, ensuring that the data state based on the path optimization logic is unified, reliable, and conflict-resolved.

[0087] Based on the consistency judgment results, the deep reinforcement learning path planning model is input to dynamically generate the optimal transportation path at the current moment;

[0088] The deep reinforcement learning path planning model includes a state input layer, a convolutional feature extraction layer, a policy network, and a value network dual-branch structure;

[0089] The state input includes the current vehicle position, unified traffic state label, destination coordinates and path history trajectory. The convolution layer extracts significant features in the input space, the policy network outputs the probability distribution of the path set, and the value network evaluates the long-term cumulative benefits of each path.

[0090] The model uses the Proximal Policy Optimization algorithm to update parameters, and a penalty function is introduced during the training process to avoid historically high accident frequency bands.

[0091] The Proximal Policy Optimization (PPO) algorithm is a commonly used reinforcement learning policy optimization method and an improved version of the policy gradient algorithm. The core idea of ​​PPO is to limit the step size of policy updates to avoid drastic policy changes during training, thereby improving learning stability and sample efficiency. It introduces a "clipped surrogate" objective, which limits the ratio between the new and old policies during policy updates. This ensures that the policy is gradually optimized and does not deviate too far from the existing policy, thereby improving convergence speed and robustness.

[0092] The penalty function introduced in path planning models is typically a regularized function with a weighted penalty term. This function significantly increases the loss of a planned path when it contains high-accident-prone areas, automatically avoiding these areas during policy optimization. By embedding this penalty mechanism into the PPO objective function, the model is guided to prioritize safer, lower-risk paths during learning, enhancing its reliability in real-world transportation scenarios.

[0093] The path optimization module is based on a multi-factor evaluation function, comprehensively considering traffic efficiency, safety risk and time matching. In the path generation process, the optimal path is represented as a set of several discrete path nodes {p a}={p1,p2,……,p m}, p a is the a-th path node. To achieve multi-objective trade-offs, the comprehensive cost score of a single node is calculated. The calculation expression is as follows:

[0094]

[0095] Where c a is at path node p a The travel cost at path node p a The cost that a vehicle needs to consume to pass through the area usually includes the weighted result of factors such as traffic congestion delay, signal light waiting time, speed limit constraint, etc., with a value range of 0-1. a is at path node p a The safety risk value at the i-th path node represents the safety risk level of the road at the i-th path node, taking into account the historical accident frequency, current environmental factors (such as rainfall, visibility), and the temporary control information obtained by the system from the event platform. The value range is 0 / 1 (such as 0 means safe and 1 means extremely high risk). a is the path node p a The deviation from the preset time window indicates the system's expected arrival time at path node p. aThe deviation between the time of arrival and the actual predicted time; if the node is constrained within a certain time window (such as the customer's delivery time), the smaller the deviation, the more it meets the transportation timeliness requirements. ω1 is the travel cost weight, which controls the importance of travel cost in path optimization. The larger the value, the more likely the planning system is to choose roads with low congestion and smooth traffic. ω2 is the safety risk weight, which indicates the priority of the system for safety considerations when selecting a path. If the transport object is high-value or dangerous goods, this parameter can be increased to avoid high-risk areas. ω3 is the time window matching weight, which indicates the importance of on-time delivery in path planning. If the transportation business has strict requirements on delivery time, this weight should be increased to make the model pay more attention to on-time arrival. a is the comprehensive cost score of a single node, indicating the path node p a In the comprehensive evaluation of traffic efficiency, safety and time accuracy, the three are weighted and combined, with the smaller the better, as the optimization goal;

[0096] Calculate the cumulative total score of all path nodes. The calculation expression is as follows:

[0097]

[0098] Where m is the total number of path nodes, F is the total cost score of the entire path, and the sum of the scores of all path nodes reflects the overall performance of the entire path in multiple dimensions;

[0099] Because paths can jump dramatically in real-time, leading to execution cost and route continuity issues, a path stability control mechanism is introduced to constrain path mutation behavior by penalizing the change amplitude between consecutive nodes. The formula is as follows:

[0100]

[0101] Where p is the weight of the path stability regularization term, which is used to balance the stability requirements in the path score, prevent the path from jittering or jumping significantly during each real-time update, and ensure the continuity and physical feasibility of the path execution. The larger the value, the more likely the system is to keep the path stable and unchanged. It is suitable for highly dynamic scenarios with frequent path updates. j is the jth path node, p j-1 is the j-1th path node, ||p j -p j-1 || 2 is the square of the distance between path nodes, indicating that path node p j With the previous node p j-1 The square of the geometric distance is used to measure the severity of the change in the path segment. R is the path instability penalty term, which is used to evaluate the discontinuity or mutation between path segments to prevent the execution inconvenience caused by excessive jumps. It also acts as a regular term to prevent frequent path changes.

[0102] When the system finally optimizes the path, it uses a comprehensive objective function to simultaneously minimize the travel cost, safety risk, time deviation, and path discontinuity. The formula is as follows: min(F+R), so that the sum of the total path score F and the stability penalty term R reaches the minimum.

[0103] The purpose of min(F+R) is to find the path among all possible paths that is efficient, safe, on time, and has a stable, continuous, and smooth overall path.

[0104] A composite objective function (COPF) is a commonly used function in optimization problems, used to combine multiple evaluation metrics into a unified optimization objective. It is not a specific "off-the-shelf function" but rather a custom weighted sum function, often used in multi-objective optimization scenarios.

[0105] In the present invention's "real-time freight transportation path planning method based on artificial intelligence," the comprehensive objective function is in the form of:

[0106] min(F+R)

[0107] It combines the path pass cost (FF) with the path stability penalty (RR), allowing the system to simultaneously optimize the following two objectives:

[0108] Path quality indicators: such as travel cost, risk level, and time error (expressed as FF);

[0109] Path execution stability: whether the path is continuous and whether it is adjusted too frequently (reflected by RR).

[0110] Its purpose is to balance efficiency, safety, timeliness, and feasibility when generating the optimal transportation route. The resulting route is not only theoretically optimal but also stable and reliable in practice. This design is commonly used in systems such as logistics scheduling, robotic path planning, and energy distribution, and is a typical "weighted multi-objective optimization model."

[0111] During route execution, it continuously monitors changes in multi-source data, assesses route reliability in real time, and triggers dynamic replanning mechanisms to respond to route interruptions or sudden changes in the traffic environment.

[0112] During the path execution process, the traffic monitoring system includes a path stability monitoring module and a dynamic replanning module;

[0113] The path stability monitoring module receives the latest data from onboard sensors, traffic monitoring systems, and road event platforms in real time, and calculates the reliability score of the current path based on the following indicators: path node matching, which measures whether the deviation between the current vehicle's actual position and the preset path exceeds a dynamic threshold; path sudden risk index, which is a weighted sum of the number of new abnormal events occurring on the road section and the risk level; path target time offset, which calculates the time difference between the vehicle and the estimated arrival time;

[0114] When any indicator exceeds its defined threshold, the dynamic replanning module is triggered. The current target destination and the previously confirmed passed segments of the path are retained, and the current vehicle position is used as the new starting point. The deep reinforcement learning path planning model is called to recalculate the remaining path.

[0115] The re-planned path must meet the three evaluation dimensions of stability, traffic efficiency and emergency avoidance capability, and automatically replace the original path to ensure that the transportation task is carried out continuously, reliably and efficiently.

[0116] Implementation 1: This implementation addresses the issue of traditional route planning systems experiencing slow response or even inattention to real-time information conflicts by deploying a multi-source data consistency awareness module within the freight transportation system. The module's core purpose is to integrate multiple heterogeneous data sources, including traffic monitoring, on-board sensors, road event platforms, and weather systems, and implement standardized conflict identification. During the data collection phase, the system first standardizes the sampling frequency, data item content, and spatial identifier format for each data source. For example, raw data from traffic monitoring includes information such as road ID, congestion level, camera location, and acquisition time. On-board sensors provide information such as vehicle speed, distance and type of obstacles ahead, lane status, and environmental perception. The road event platform provides announcements pushed by the government or traffic management center, such as construction, road closures, and accidents. The weather system provides real-time local meteorological information, including precipitation, temperature, wind speed, and visibility.

[0117] After the module receives the raw data, the first step is time synchronization, which is to calibrate the timestamps from different sources. Taking into account the clock drift between different devices, the system will introduce a time correction mechanism, using the main server to unify the standard time, combined with the device return delay to calculate the actual time of the event, eliminating the error caused by asynchrony. The second step is spatial alignment, which is to convert the spatial coordinate systems of different data sources into a unified local reference coordinate system. The on-board equipment uses a combination of high-precision GPS and IMU to achieve precise vehicle positioning. At the same time, the coordinates of other information sources (such as the location of traffic incidents and road congestion points) are converted through electronic maps so that they can be compared and analyzed in the same spatial dimension.

[0118] After completing spatiotemporal normalization, the module uses a rule system to check data consistency. For example, the rule base includes: if the traffic monitoring status of a certain road section is clear, but the vehicle's front radar detects an immovable obstacle for three consecutive frames, it will be considered a data conflict; or if the weather system indicates low visibility in the area, but the surveillance video is clear, it will also be considered a possible false alarm. Each rule has a priority and scope to prevent false triggering.

[0119] Once a conflict is detected, the system identifies the conflict type, level, and suspected cause, and pushes this information to the upper-level decision-making module, invoking subsequent fusion mechanisms or directly triggering a route reassessment process. This module possesses both recognition and learning capabilities, dynamically adjusting rule thresholds to adapt to varying traffic conditions and seasonal changes, further enhancing the system's practicality and stability.

[0120] Implementation Method 2: This implementation method addresses situations where logical conflicts exist among multiple data sources. A weighted fusion of conflicting data and a confidence calculation strategy are introduced to quantitatively assess the reliability of information from each source, ultimately leading to a unified traffic assessment. This strategy is based on the following principles: different data sources offer distinct advantages. For example, onboard sensors offer high real-time performance and high resolution, making them suitable for short-range obstacle detection. Meanwhile, traffic management systems, while updating less frequently, offer extensive coverage and global information. Therefore, relying solely on a single data source can easily lead to route planning errors, necessitating a weighted fusion mechanism to improve decision accuracy and robustness.

[0121] The specific process consists of three phases: evaluation, integration, and decision-making. During the evaluation phase, the system first calculates the historical performance of each data source, assessing its consistency with actual conditions in similar scenarios. For example, if a road event platform reported 10 road closures in the past month and 7 of them were verified as true by on-board equipment, its credibility is rated at 0.7. In addition, the system also calculates the delay between each data update and the standard update time. For example, the delay of on-board equipment is usually in seconds, while some external platforms may exceed 10 minutes.

[0122] After the evaluation is complete, the system enters the integration phase, where it calculates a weight for each piece of conflicting data based on the various evaluation results. This weighting depends not only on the current data quality but also on the data's location and logical consistency with other information. For example, when the on-board equipment and traffic monitoring provide the same traffic status, the credibility of both is weighted and enhanced. If there is a conflict, adjustments are made based on factors such as historical accuracy and time delay. During this process, the system incorporates a dynamic trust allocation model, enabling different handling strategies for the same type of conflict in different time periods or regions.

[0123] Finally, in the decision-making phase, the integrated confidence level is input into the traffic state judgment logic, which outputs the final road passability judgment. This judgment is not only used for current path selection but also for establishing system logs and feedback mechanisms to learn patterns for future scenarios. For example, if the system discovers that a certain type of conflict frequently occurs near construction areas and ultimately determines that most roads are impassable, it can automatically adjust the rule base and trust model parameters to prioritize the corresponding source information in subsequent similar incidents. This approach not only improves data consistency processing capabilities but also provides a high-quality input data source for subsequent AI path planning models.

[0124] Implementation Method 3: This implementation primarily achieves dynamic replanning capabilities by building a reinforcement learning-driven path planning model and tightly integrating it with a multi-source information monitoring mechanism. Throughout the cargo transportation process, environmental conditions are extremely complex and volatile, including numerous disruptive factors such as traffic control, inclement weather, and unexpected accidents. Traditional pre-set routing methods often struggle to adapt flexibly. However, a dynamic path planning mechanism allows for timely adjustments based on actual traffic conditions, ensuring efficient and safe delivery of cargo.

[0125] During the initial path planning phase, the system first receives traffic status information from the multi-source consistency judgment module, including integrated parameters such as road accessibility, risk level, and event density. Combining the current vehicle position, destination coordinates, time requirements, and cargo priority, a state space is constructed and input into the reinforcement learning model. The model structure utilizes a dual-branch architecture consisting of a policy network and a value network. The policy network outputs multiple candidate paths, while the value network evaluates the potential benefits of each path. The path selection process not only considers traditional factors such as distance and time, but also incorporates risk cost, traffic stability, and historical accident density into the decision-making mechanism, making path selection more intelligent and comprehensive.

[0126] Once path planning is complete and execution begins, the system continuously receives real-time data input from external and onboard devices. If a significant change in road conditions is detected during execution, such as an accident ahead, a road closure, or rapidly deteriorating weather, the system immediately triggers a replanning mechanism. This mechanism does not recalculate the entire path, but instead uses the current location as a starting point to make the optimal choice on a local path, thereby reducing replanning latency. The system also analyzes the switching costs between the current path and candidate paths to avoid operational instability caused by frequent switching.

[0127] The system also features a route planning feedback mechanism. After each route execution, the system records data such as the actual driving trajectory, arrival time, and number of route changes into a database, which is used to train models and optimize parameters. Through continuous iteration, the system gradually develops a diverse library of route strategies that adapt to the traffic characteristics of different cities. In long-term operation, this mechanism has significantly improved the overall stability, responsiveness, and operational efficiency of the transportation system, significantly reducing delays caused by unexpected events.

[0128] The present invention effectively improves the accuracy of information perception of the cargo transportation route planning system in the face of emergencies by introducing a multi-source data consistency perception mechanism. Traditional route planning methods usually rely on a single data source and are unable to promptly detect changes in road conditions caused by sudden road closures, traffic accidents, severe weather, etc., which can easily lead to vehicles entering risky sections. This invention collects heterogeneous data from traffic monitoring, vehicle-mounted sensors, road event platforms, and weather systems, and identifies logical conflicts between data through unified spatiotemporal standardization processing and rule systems, thereby fundamentally improving the comprehensiveness and accuracy of information perception. In particular, when there are contradictions in multi-source data, the system can quickly identify the problem and respond to it, avoiding path deviations caused by data distortion, and significantly enhancing the system's environmental adaptability and safety.

[0129] The weighted fusion and confidence calculation strategy of conflicting data proposed in the present invention solves the problem of difficulty in quantifying the reliability differences of multi-source data, and provides a more reliable input basis for the path planning system. In actual applications, there are obvious differences in the update frequency, transmission delay, and historical accuracy of various data sources. If they are not effectively distinguished and processed, it is very easy to misjudge or ignore key events. The present invention establishes a weight evaluation mechanism based on the historical accuracy, current timeliness and stability of the data source, and adopts a Bayesian probability model for data fusion, so that the system can dynamically adjust the trust level of different data and output the traffic status judgment that best conforms to the actual situation. This mechanism not only improves the judgment robustness of the path planning model in complex scenarios, but also significantly reduces the economic losses and safety risks caused by incorrect path decisions.

[0130] The path planning and dynamic re-planning coordination mechanism constructed by the present invention significantly improves the real-time response capability and intelligent level of path selection of the cargo transportation system. Unlike traditional static path planning solutions, the present invention continuously monitors traffic and environmental changes during path execution. When abnormal information that affects traffic efficiency or safety is detected, it can quickly start the local re-planning mechanism to achieve instant optimization of the path. This mechanism not only takes into account the path switching cost, but also introduces historical trajectory and risk density parameters to ensure that the planning results minimize driving risks while ensuring transportation efficiency. The system also has self-learning capabilities and can feed back historical path execution results to the path strategy library for optimizing model parameters and gradually building an optimal path library that adapts to different traffic areas, comprehensively improving the intelligence level and long-term operational stability of the system.

[0131] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0132] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

[0133] It should be noted that, in this document, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device that includes the element.

[0134] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0135] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0136] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0137] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0138] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0139] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0140] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

Claims

1. A real-time cargo transportation path planning method based on artificial intelligence, characterized in that: The following steps are involved: Collect real-time data from traffic monitoring systems, vehicle sensors, road event platforms, and weather information systems during transportation to form a raw data set containing multi-source information; Perform time synchronization and spatial alignment on the data from each source in the original dataset and construct a standardized data frame to ensure the consistency of information from different sources in the temporal and spatial dimensions; Set conflict detection rules and identify logical conflicts in standardized data frames based on the rules, including traffic state contradictions, inconsistent road traffic information, and differences in environmental perception; The multi-source data fusion module is activated to perform a weighted assessment of conflicting data based on the credibility, real-time nature, and historical error rate of the data sources, generating the most consistent road traffic status judgment results. Based on the consistency judgment results, the deep reinforcement learning path planning model is input to dynamically generate the optimal transportation path at the current moment; During the path execution process, the system continuously monitors changes in multi-source data, evaluates path reliability in real time, and triggers a dynamic replanning mechanism to deal with path interruptions or sudden changes in the traffic environment.

2. The method for real-time cargo transportation path planning based on artificial intelligence according to claim 1, characterized in that: The traffic monitoring system data collected includes traffic light cycles, road congestion index, and road closure status; vehicle sensor data includes the distance to obstacles ahead, lane status, and real-time speed; road event platform data includes traffic accidents, construction records, and regulatory notices that have occurred in the current area within the past 48 hours; and weather information system data includes real-time rainfall, visibility level, surface temperature, and wind speed and direction. All data is updated at a frequency of no more than 60 seconds and stored in the local cache for subsequent standardized processing and calls.

3. The method for real-time cargo transportation path planning based on artificial intelligence according to claim 1, characterized in that: Time synchronization uses a unified timestamp interpolation algorithm, while spatial alignment relies on high-precision electronic maps, mapping all data uniformly into a polar coordinate system with the current transport vehicle as the reference point; Timestamp interpolation uses spline fitting to estimate the state changes of different systems in the same time period; the spatial position error is controlled within 1 meter, and the Kalman filter is used to dynamically correct the positioning error of the on-board sensors to ensure that all types of information can be analyzed and compared under a unified coordinate and time reference.

4. The method for real-time cargo transportation path planning based on artificial intelligence according to claim 1, characterized in that: Conflict detection rules include the following categories: Data inconsistency conflict rules to determine the contradiction between the traffic monitoring clear sign and the vehicle-mounted obstacle distance less than the threshold; Logical contradiction and conflict rules to determine when the road event platform shows a closure but the vehicle's sensors do not detect a closure sign; Early warning conflict rules issue a warning when any emergency indication exists in multi-source data and the system has not entered the avoidance state.

5. The method for real-time cargo transportation path planning based on artificial intelligence according to claim 1, characterized in that: The multi-source data fusion module builds a Bayesian scoring model based on the credibility, real-timeness and historical error rate of the data sources; Credibility is based on historical accuracy and source stability; real-time performance is determined by the time difference between current data and the standard update time; and the historical error rate is derived from the frequency of misjudgments generated by the system comparing annotated data over the past 30 days. After each conflict information is updated through Bayesian probability, the maximum a posteriori estimation mechanism is used to select the optimal passage judgment, and a unified traffic state label is output as one of the input conditions for path planning.

6. The method for real-time cargo transportation path planning based on artificial intelligence according to claim 5, characterized in that: Multi-source data weighted fusion includes confidence calculation and conflict judgment mechanism. The specific steps are as follows: For each data source, calculate its confidence adjustment value based on its current transmitted data quality and historical performance. The calculation expression is as follows: Where C i is the initial confidence of the i-th data source, A i is the historical average accuracy of the i-th data source, T i is the current data time deviation, α is the time attenuation coefficient, W i is the weighted confidence of the i-th data source; The identified conflict types are divided into several categories, and the information intensity provided by each data source under each type of conflict is defined as I i，k , calculate the weighted total score of all data sources involved in conflict judgment. The calculation expression is as follows: Where, I i，k is the indication strength provided by the i-th data source for the k-th type of conflict, n is the total number of data sources, S k is the total confidence value of the k-th conflict judgment; Then for all confidence sum values ​​S k After comparison, the judgment corresponding to the conflict type k with the highest score is selected as the final road pass judgment result. The judgment logic is as follows: Where D is the final conflict type label, is a mathematical function.

7. The method for real-time cargo transportation path planning based on artificial intelligence according to claim 1, characterized in that: The deep reinforcement learning path planning model includes a state input layer, a convolutional feature extraction layer, a policy network, and a value network dual-branch structure; The state input includes the current vehicle position, unified traffic state label, destination coordinates and path history trajectory. The convolution layer extracts significant features in the input space, the policy network outputs the probability distribution of the path set, and the value network evaluates the long-term cumulative benefits of each path.

8. The method for real-time cargo transportation path planning based on artificial intelligence according to claim 1, characterized in that: The path optimization module is based on a multi-factor evaluation function, comprehensively considering traffic efficiency, safety risk and time matching. In the path generation process, the optimal path is represented as a set of several discrete path nodes {p a }={p1,p2,……,p m }, p a is the a-th path node. To achieve multi-objective trade-offs, the comprehensive cost score of a single node is calculated. The calculation expression is as follows: Where c a is at path node p a The cost of passage at a is at path node p a Safety risk value at t a is the path node p a The deviation from the preset time window, ω1 is the travel cost weight, ω2 is the safety risk weight, ω3 is the time window matching weight, f a is the comprehensive cost score of a single node; Calculate the cumulative total score of all path nodes. The calculation expression is as follows: Where m is the total number of path nodes, and F is the total cost score of the entire path; Because paths can jump dramatically in real-time, leading to execution cost and route continuity issues, a path stability control mechanism is introduced to constrain path mutation behavior by penalizing the change amplitude between consecutive nodes. The formula is as follows: Where λ is the path stability regularization term weight, p j is the jth path node, p j-1 is the j-1th path node, ||p j -p j-1 || 2 is the square of the distance between path nodes, and R is the path instability penalty term; When finally optimizing the path, a comprehensive objective function is used to simultaneously minimize the travel cost, safety risk, time deviation, and path discontinuity. The formula is as follows: min(F+R), so that the sum of the total path score F and the stability penalty term R reaches the minimum.

9. The method for real-time cargo transportation path planning based on artificial intelligence according to claim 1, characterized in that: During the path execution process, the traffic monitoring system includes a path stability monitoring module and a dynamic replanning module; The path stability monitoring module receives the latest data from onboard sensors, traffic monitoring systems, and road event platforms in real time, and calculates the reliability score of the current path based on the following indicators: path node matching, which measures whether the deviation between the current vehicle's actual position and the preset path exceeds a dynamic threshold; path sudden risk index, which is a weighted sum of the number of new abnormal events occurring on the road section and the risk level; path target time offset, which calculates the time difference between the vehicle and the estimated arrival time; When any indicator exceeds its defined threshold, the dynamic replanning module is triggered. The current target destination and the previously confirmed passed segments of the path are retained, and the current vehicle position is used as the new starting point. The deep reinforcement learning path planning model is called to recalculate the remaining path. The re-planned path must meet the three evaluation dimensions of stability, traffic efficiency and emergency avoidance capability, and automatically replace the original path to ensure that the transportation task is carried out continuously, reliably and efficiently.

Images