Intelligent traffic event pushing method and device

Abstract

This invention discloses an intelligent traffic event push method and apparatus. The method includes: acquiring alarm data from multiple sources of traffic events and preprocessing it to form a standardized alarm set; wherein the preprocessed traffic event alarm data includes event number, event type, spatiotemporal coordinates, impact range, hazard level, and original confidence level; constructing a MIP model; the MIP model is constructed based on designed decision variables, with the optimization objectives of maximizing the benefits of effective event push, minimizing the losses from low-confidence false alarms, and minimizing redundant pushes of duplicate events, and setting constraints; solving the MIP model based on the standardized alarm set to obtain a traffic event push scheme that maximizes the objective function while satisfying the constraints.

Description

Technical Field

[0001] This invention relates to the field of traffic event push technology, and in particular to an intelligent traffic event push method and device. Background Technology

[0002] With the widespread adoption of intelligent transportation systems, multi-source sensing devices such as video AI terminals, radar sensing terminals, roadside IoT terminals, environmental sensing terminals, and vehicle-mounted sensing terminals are widely used for the automatic identification and alerting of high-level traffic incidents (incident types include, but are not limited to, traffic accidents, wrong-way driving, lane obstruction, sudden congestion, road defects, and violations by hazardous chemical vehicles). Because high-level traffic incidents are typically characterized by high urgency and a wide impact, the timeliness requirements for handling them are extremely stringent, usually requiring a response within minutes. In this context, allowing for an optimization delay of 5 to 10 seconds is perfectly acceptable; the key is to ensure that incident information is accurately pushed to traffic management departments and maintenance personnel, thereby guaranteeing timely handling of the incident and effectively reducing traffic congestion and safety hazards.

[0003] Currently, traffic incident alert technology mainly relies on the independent detection results of a single device or algorithm, and directly pushes alerts after filtering through simple threshold judgments or preset rules. Existing related research and technological improvements mostly focus on improving the accuracy of single event recognition, deduplication for a specific stage, or simply optimizing the real-time performance of the alerts. Summary of the Invention

[0004] To improve the quality and efficiency of traffic event notifications and increase the selection of high-level traffic event alarm technologies, this invention provides an intelligent traffic event notification method and apparatus.

[0005] In a first aspect, embodiments of the present invention provide an intelligent traffic event push method, including:

[0006] Acquire alarm data from multiple sources of traffic events and preprocess them to form a standardized alarm set; the preprocessed alarm data of traffic events includes event number, event type, spatiotemporal coordinates, scope of impact, hazard level and original confidence level;

[0007] A MIP model is constructed based on the designed decision variables, with the optimization objectives of maximizing the benefits of effective event push, minimizing the loss of low-confidence false alarms, and minimizing redundant push of duplicate events, and constraints are set.

[0008] The MIP model is solved based on the standardized alarm set to obtain a traffic event push scheme that maximizes the objective function under the constraints.

[0009] In one or more optional embodiments, the MIP decision variables include: discrete decision variables such as whether the traffic event is valid, repeated events in the same time and space, and the channel through which the traffic event is pushed; and continuous decision variables such as the confidence level of the valid decision and the event priority.

[0010] The objective function is:

[0011] ,

[0012] Where I represents the standardized alarm set, For the i-th traffic event in the standardized alarm set, Let α be the priority score for the i-th traffic incident, β be the false alarm penalty coefficient, and β be the repeat incident penalty coefficient. Let be a set of repeating event pairs, where, For traffic incidents and A pair of repeated events in the same space-time.

[0013] In one or more optional embodiments, the false alarm penalty coefficient is calculated in the following manner:

[0014] ,

[0015] Where L represents the current load rate of all channels. The base false alarm penalty coefficient, For high load sensitivity parameters, For low load sensitivity parameters, The high load trigger threshold, The low load trigger threshold, This is the threshold for full load conditions.

[0016] In one or more optional embodiments, the load rate of all current push channels is calculated in the following manner:

[0017] ,

[0018] in, Let be the weight coefficient of the k-th push channel, and the sum of the weight coefficients of all k push channels is 1; Let be the real-time utilization rate of the k-th push channel, and let be the proportion of traffic event tasks currently in active push status in the k-th push channel to the total channel capacity.

[0019] In one or more optional embodiments, the constraints include event confidence admission constraints constructed based on the Big M method according to a preset confidence admission threshold and the original confidence, mutual exclusion constraints for repeated events in the same space-time, capacity constraints for each push channel, constraints that only valid events can be assigned to push channels, and domain constraints for discrete and continuous decision variables.

[0020] In one or more optional embodiments, the event confidence admission constraint constructed based on the Big M method according to a preset confidence admission threshold and the original confidence level includes:

[0021] Based on the preset confidence threshold and the original confidence level, the following constraint expression is constructed using the Big M method to obtain the event confidence threshold constraint:

[0022] ,

[0023] in, This represents the confidence level of an effective decision regarding a traffic incident, where τ is a pre-set confidence threshold. The original confidence level of the traffic event is M, where M is a large M constant with values ​​in the interval [0,1].

[0024] In one or more optional embodiments, the constraint expression for the mutual exclusion constraint of repeated events in the same space-time is:

[0025]

[0026] Indicates a traffic incident and For repeated events in the same space-time, Indicates a traffic incident and They cannot both be valid events simultaneously;

[0027] In one or more optional embodiments, the constraint expression for the capacity constraint of each push channel is:

[0028]

[0029] in, Let J be the maximum concurrent capacity of the j-th push channel, and J be the total number of push channels. For traffic incidents Push to the j-th push channel.

[0030] In one or more optional embodiments, the constraint expression for assigning only valid events to the push channel constraint is:

[0031] ,

[0032]

[0033] In one or more optional embodiments, the constraint expressions for the domain constraints of the discrete and continuous decision variables are as follows:

[0034] ,

[0035] in, This indicates that the i-th traffic incident is determined to be a valid incident; This indicates that the i-th traffic incident is determined to be an invalid incident; Indicates a traffic incident and As an independent event; This indicates that the i-th traffic incident is assigned to the j-th push channel; This indicates that the i-th traffic incident is not assigned to the j-th push channel.

[0036] In one or more optional embodiments, the intelligent traffic event push method further includes:

[0037] Traffic events in the standardized alarm set are pre-filtered according to preset lightweight rules, and duplicate event pairs in the same time and space are generated for the pre-filtered traffic events to obtain the duplicate event pair set.

[0038] Secondly, embodiments of the present invention provide an intelligent traffic event push device, comprising:

[0039] The first acquisition module is used to acquire alarm data of multi-source traffic events and preprocess it to form a standardized alarm set; the preprocessed alarm data of traffic events includes event number, event type, spatiotemporal coordinates, scope of impact, hazard level and original confidence level;

[0040] The second acquisition module is used to construct the MIP model. The MIP model is constructed based on the designed decision variables, with the optimization objectives of maximizing the benefits of effective event push, minimizing the loss of low-confidence false alarms, and minimizing redundant push of duplicate events, and with constraints set.

[0041] The solution module is used to solve the MIP model based on the standardized alarm set to obtain a traffic event push scheme that maximizes the objective function under the constraints.

[0042] Thirdly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, characterized in that, when the program is executed by a processor, it implements the intelligent traffic event push method as described in the first aspect.

[0043] Fourthly, embodiments of the present invention provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the program to implement the intelligent traffic event push method as described in the first aspect.

[0044] Fifthly, embodiments of the present invention provide a computer program product containing instructions that, when the computer program product is run on a computer device, cause the computer device to execute the intelligent traffic event push method as described in the first aspect.

[0045] The beneficial effects of the above-described technical solutions provided in the embodiments of the present invention include at least the following:

[0046] The intelligent traffic event push method provided in this invention processes the acquired multi-source traffic event raw data into standardized alarm data, introduces a mixed integer programming (MIP) model, and unifies the modeling and solving of the four major stages of event deduplication, authenticity confirmation, priority sorting, and channel allocation. This breaks through the limitations of the traditional serial processing of "deduplication, confirmation, sorting, and allocation," transforms the local optima of multiple stages into global optimal decisions, and obtains a traffic event push scheme that maximizes the objective function under the constraints, significantly reducing the rate of duplicate alarms and false alarm pushes.

[0047] The intelligent traffic event push method provided in this invention innovatively constructs event confidence admission constraints based on the Big M method, strongly coupling confidence verification with event authenticity determination. When the original confidence is lower than the admission threshold, the constraint becomes mathematically infeasible, forcing the MIP solver to automatically determine the event as invalid, achieving hard threshold truncation without writing condition judgment code.

[0048] The intelligent traffic event push method provided in this invention adaptively adjusts the false alarm penalty intensity based on the real-time load rate of the push channel through a dynamic false alarm penalty coefficient α. When the channel is ample, the false alarm penalty is reduced to ensure event coverage; when the channel is congested, the false alarm penalty is increased to ensure push quality, achieving robust decision-making based on dynamic trade-offs within the scenario context.

[0049] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0050] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0051] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0052] Figure 1 This is a flowchart illustrating the intelligent traffic event push method provided in an embodiment of the present invention;

[0053] Figure 2 This is a structural block diagram of the intelligent traffic event push device provided in the embodiments of the present invention;

[0054] Figure 3 This is a structural block diagram of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0055] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0056] The inventors discovered that existing intelligent transportation high-level event confirmation and push technologies have significant shortcomings:

[0057] The proliferation of duplicate alarms: Multiple sensing devices repeatedly report the same event in the same time and space, resulting in redundant push messages. This not only occupies limited push channel resources, but also seriously interferes with the judgment and response efficiency of operation and maintenance personnel.

[0058] High false alarm rate: Using a fixed confidence threshold to determine the authenticity of traffic events cannot balance "not missing high-level events" and "suppressing false alarms". Either the threshold is set too high, resulting in the omission of high-level events, or the threshold is set too low, leading to a flood of false alarm messages and reducing the quality of push notifications.

[0059] Disordered priority ranking: The lack of a global priority optimization mechanism means that high-urgency, high-impact events may be overwhelmed by low-priority events, causing delays in emergency response.

[0060] Unreasonable channel allocation: The selection of push channels lacks load balancing constraints. Some channels are overloaded due to message backlog, while others are idle, making it impossible to achieve hierarchical and accurate push based on event level.

[0061] Lack of global optimization: Existing technologies mostly adopt a local decision-making approach of "simple threshold judgment + preset rule filtering", which cannot coordinate the entire process of event deduplication, authenticity confirmation, priority ranking, and channel allocation, and cannot achieve global optimal decision-making.

[0062] Overall, when faced with complex and ever-changing real-world traffic scenarios, existing technologies, lacking a global optimization approach, cannot simultaneously address multiple optimization goals such as event deduplication, authenticity verification, priority ranking, and channel allocation throughout the entire process of high-level traffic event confirmation and push notification. This leads to problems such as serial decision-making in multi-source alarm processing, resource waste, false positives and false negatives due to hard thresholds, and difficulties in channel load balancing. Therefore, there is still significant room for improvement in alarm efficiency and quality when dealing with complex and ever-changing real-world traffic scenarios.

[0063] To at least partially solve the technical problems existing in the prior art, the inventors propose to introduce mixed integer programming (MIP) technology to break through the limitations of the traditional serial processing of "deduplication, confirmation, sorting, and allocation". A joint optimization model covering "deduplication, authenticity verification, sorting, and allocation" is constructed to execute a full-process technical solution of "multi-source traffic event aggregation, MIP model construction, model solving, and traffic event push". Under the premise of meeting business constraints, it realizes the joint optimization of multiple dimensions such as event deduplication, authenticity verification, priority sorting, and channel allocation, achieves the global optimum of traffic event push decision, and significantly improves the quality and efficiency of high-level event push.

[0064] Based on this, the inventors conducted further research and development and developed this invention, which provides a method and device for intelligent traffic event push.

[0065] Example 1

[0066] This invention provides an intelligent traffic event push method, referring to... Figure 1 As shown, the method includes:

[0067] S101: Acquire alarm data from multiple sources of traffic events and preprocess it to form a standardized alarm set; the preprocessed alarm data of traffic events includes event number, event type, spatiotemporal coordinates, scope of impact, hazard level and original confidence level;

[0068] S102: Construct the MIP model; the MIP model is constructed based on the designed decision variables, with the optimization objectives of maximizing the benefits of effective event push, minimizing the loss of low-confidence false alarms, and minimizing redundant push of duplicate events, and with constraints set.

[0069] S103: Solve the MIP model based on the standardized alarm set to obtain a traffic event push scheme that maximizes the objective function under the constraints.

[0070] In this embodiment of the invention, the original alarm data for the multi-source traffic event in step S101 above can come from multi-source alarm devices and manual channels. The multi-source alarm devices include, but are not limited to, the following alarm devices:

[0071] Video AI terminals: high-definition cameras, intelligent PTZ cameras, and intelligent video equipment for tunnels, used to identify accidents, congestion, spills, illegal parking, and road conditions;

[0072] Radar sensing terminals: millimeter-wave radar and lidar, adapted to low visibility scenarios such as nighttime and rain / fog, used to accurately identify road condition events such as low speed, stagnation, collision with ice, and lane occupancy.

[0073] Roadside IoT terminals: RSU roadside units, geomagnetic detectors, and road condition sensors, used to collect data on traffic congestion, road damage, and water accumulation;

[0074] Environmental sensing terminals: meteorological monitoring equipment and visibility detectors, used to collect data on abnormal traffic events caused by weather conditions such as fog, heavy rain, heavy snow, and strong winds;

[0075] Vehicle-mounted sensing terminal: OBU vehicle-mounted unit, operating vehicle monitoring terminal, used to report dynamic events such as vehicle malfunctions, driving abnormalities, and sudden road hazards.

[0076] Raw alarm data for traffic incidents received through manual channels can come from data reported by the following types of devices:

[0077] Frontline operation and maintenance: road patrol, highway operation and maintenance, road maintenance, handheld / vehicle terminals for traffic police on duty;

[0078] Public reporting channels: 122, 110, 12328, government mini-programs, map traffic reporting, and public hotline reporting;

[0079] Industry regulatory end: Road emergencies reported by safety supervisors of public transportation, passenger transport, and freight transport companies.

[0080] The raw alarm data for each traffic incident can be standardized and preprocessed to cleanse unstructured data into standardized alarm objects. Ensure that each preprocessed traffic incident alarm data contains the following core attributes: incident number (unique identifier), incident type, spatiotemporal coordinates (latitude and longitude of the incident, timestamp (accurate to the second)), impact range, raw confidence level (characterizing the reliability of the incident), and hazard level.

[0081] Specifically, spatiotemporal alignment can unify data from different sources into the same spatiotemporal coordinate system, and satellite time synchronization can be used to correct the timestamps of different source data. The raw confidence level of traffic events from multi-source alarm devices can be obtained through AI detection algorithms, such as using a multimodal fusion model to directly output the raw confidence level. The raw confidence level of traffic events from manual channels can be a preset value, such as 1. The impact range of a traffic event can include the event handling time, the extent of congestion spread, and the duration, which can be obtained either from manual annotation or through AI detection algorithms. For specific methods, please refer to the detailed description in existing technologies.

[0082] The hazard level of a traffic incident can be determined by combining the incident type and the scope of its impact. For example, based on the scope of impact, the hazard level can be divided into four levels, as shown in Table 1:

[0083]

[0084] Table 1

[0085] The preprocessed traffic event alarm data are formed into a standardized alarm set, denoted as I={1,2,...,i,...,N}, where N is the total number of traffic event alarm data, and i represents the i-th alarm data. Let i be the i-th traffic event in the standardized alarm set.

[0086] In this embodiment of the invention, there can be multiple channels for pushing traffic events, including but not limited to SMS channels, traffic management platforms, operation and maintenance dashboards, smart terminal apps, inspection terminals, and Webhook callbacks.

[0087] In this embodiment of the invention, in step S102 above, when constructing the MIP model, the decision variables of the MIP model are first designed in conjunction with business requirements, and are divided into 0-1 integer variables (for discrete decision-making) and continuous variables (for quantitative representation). The variables are defined as follows to ensure that the variables are strongly bound to the business scenario and to avoid abstraction:

[0088] (1) Discrete decision variables (0-1 integer variables), including event authenticity judgment variables that characterize whether a traffic event is valid, repeated event correlation variables that characterize repeated events in the same time and space, and push channel allocation variables that characterize the push channels of traffic events:

[0089] Variables for determining the truth or falsity of an event: , This indicates that the i-th event is determined to be a valid event and needs to be included in the push notification; This indicates that the i-th event is an invalid event (false alarm or duplicate event) and will not be pushed.

[0090] Recurring event associated variables: , This indicates that the i-th event and the k-th event are duplicate events in the same spacetime (e.g., judging criteria: spacetime distance ≤ 50 meters, time difference ≤ 30 seconds, and event type is the same). This indicates that the two events are independent events.

[0091] Push channel allocation variables: , Indicates the i-th valid event ( ) is assigned to the j-th push channel; This indicates that the i-th traffic incident is not assigned to the j-th push channel. The set of push channels is denoted as J={1,2,...,K}, where K is the total number of channels (e.g., J1=traffic management platform, J2=maintenance screen, J3=smart terminal APP, J4=inspection terminal).

[0092] (2) Continuous variables, including effective decision confidence and event priority:

[0093] Confidence level of effective decision: This is a decision variable, not simply an input parameter, and represents the final confidence assessment after validation by the MIP model.

[0094] Event Priority: Event priority is determined by a priority score for traffic events. (i.e., comprehensive event priority score) indicates that, , The larger the value, the higher the event priority, indicating that it should be pushed out first.

[0095] In one specific embodiment, the priority scoring of the aforementioned traffic incidents It can be calculated by weighting the risk level, impact range, and confidence level of effective decision-making in a traffic incident, such as design priority scoring. The formula is: =0.5 × Hazard Level Weight + 0.3 × Impact Area Weight + 0.2 × .

[0096] In one specific embodiment, the objective function of the MIP model constructed above, with the optimization objectives of maximizing the benefits of effective event pushes, minimizing the losses from low-confidence false alarms, and minimizing redundant pushes of duplicate events, is as follows:

[0097] ,

[0098] Where I represents the standardized alarm set, For the i-th traffic event in the standardized alarm set, Let α be the priority score for the i-th traffic incident, β be the false alarm penalty coefficient, and β be the repeat incident penalty coefficient. Let be a set of repeating event pairs, where, For traffic incidents and A pair of repeated events in the same space-time.

[0099] In this embodiment of the invention, the objective function is centered on "globally optimal push," taking into account four main requirements: maximizing the benefits of pushing genuine high-level events, minimizing the losses from false alarms with low confidence, and minimizing redundant pushes for duplicate events. Among these:

[0100] This means maximizing the benefits of pushing high-priority valid events, ensuring that events with high urgency and high reliability are pushed first, and achieving priority sorting optimization.

[0101] Minimize the push loss due to low-confidence false alarms (convert low confidence into a penalty term instead of filtering), false alarm penalty coefficient. It can be adjusted according to actual operation and maintenance needs, (1- The larger the value, the higher the probability of false alarms and the stronger the penalty, thus avoiding false alarm pushes and optimizing channel allocation, event deduplication, and authenticity verification.

[0102] Minimize redundant pushes of duplicate events; duplicate event penalty coefficient. This can be adjusted based on experience and actual conditions. The more duplicate events there are, the stronger the penalty, encouraging merging and deduplication to optimize event deduplication. This is achieved through a global perspective on duplicate event variables. The MIP model can identify and eliminate low-value duplicate alarms that meet the deduplication criteria but occupy important resources (such as traffic management screens), thereby significantly reducing operation and maintenance noise and significantly improving the duplicate alarm rate and false alarm push rate.

[0103] In this embodiment of the invention, the false alarm penalty coefficient α can adopt a dynamic mechanism. Preferably, when the push channel load rate is > 80%, α is automatically increased (e.g., α is adjusted from 3 to 5) to increase the false alarm penalty and guide the MIP model to prioritize the elimination of low confidence events; while when the load rate is < 30%, α is decreased (e.g., α is reduced from 3 to 1.5) to reduce the false alarm penalty and guide the model to retain more events to reduce false negatives.

[0104] By introducing a dynamic false alarm penalty coefficient, the model can tolerate low-confidence events to ensure coverage (reduce false alarms) when channels are plentiful, while automatically prioritizing the elimination of low-confidence events to ensure quality (reduce false alarms) when channels are congested. This dynamic trade-off capability based on scenario context significantly improves decision robustness.

[0105] In one specific embodiment, the false alarm penalty coefficient can be calculated in the following manner:

[0106] ,

[0107] Where L represents the current load rate of all channels. The base false alarm penalty coefficient, For high load sensitivity parameters, For low load sensitivity parameters, The high load trigger threshold, The low load trigger threshold, This is the threshold for full load conditions.

[0108] In this embodiment of the invention, based on the method for calculating the false alarm penalty coefficient, when the push channel is in a high-load range, i.e., L > 0.8, =0, It works; α increases with increasing load rate; high load sensitivity parameter. By controlling the increase in α, the false alarm penalty is increased, guiding the MIP model to prioritize the elimination of low-confidence events; while in the low-load region, i.e., L < 0.3, =0, It takes effect; α decreases as the load rate decreases, indicating a low load sensitivity parameter. By controlling the magnitude of the decrease in α, the penalty for false positives is reduced, guiding the model to retain more events to reduce false negatives. However, when 0.3 ≤ L ≤ 0.8, and Both are 0, α = .

[0109] In this embodiment of the invention, The value of is in the range (0,1]. The value can take the range (0, 1). For example, =3, =0.5, =0.5, =0.8, =0.3, =1.

[0110] In one specific embodiment, the current load rate L for all channels ranges from 0 to 1. In a concrete implementation, L can be calculated using a weighted average based on the real-time usage rate of each channel.

[0111] Suppose there are K push channels (such as SMS gateway, APP Push channel, Webhook callback, etc.), and the real-time utilization rate of each channel is... (k=1,2,...,K, defined as the proportion of traffic events currently in active push status in the k-th push channel to the total channel capacity). The load rate of all current push channels can then be calculated as follows:

[0112] ,

[0113] in, Let be the weight coefficient of the k-th push channel, and the sum of the weight coefficients of all k push channels is 1; Let be the real-time utilization rate of the k-th push channel, and let be the proportion of traffic event tasks currently in active push status in the k-th push channel to the total channel capacity.

[0114] In this embodiment of the invention, the weighting coefficient of the push channel This can be determined based on the priority of each channel or historical usage frequency. For example, in actual deployment, The weighting can be set according to business priorities. For example, if the push channels are SMS gateway, APP Push channel, and Webhook callback, then SMS channel has the highest weight (w_SMS=0.5), followed by APP Push (w_APP=0.3), and Webhook callback has the lowest weight (w_Webhook=0.2). Of course, when the priorities of each channel are the same, it can be simplified to an equal-weighted average. = 1 / K).

[0115] In one specific embodiment, the aforementioned real-time utilization rate It can be calculated in the following way:

[0116] Obtain the usage rate of the kth push channel at each moment within a preset sliding time window, and calculate the moving average of the usage rate of the kth push channel at each moment within the preset sliding time window to obtain the real-time usage rate of the kth push channel.

[0117] Determining the real-time utilization rate of the k-th push channel using this calculation method can smooth out the impact of instantaneous fluctuations on the false alarm penalty coefficient α. The sliding window mechanism ensures that the load rate estimate strikes a balance between response speed and anti-interference capability.

[0118] In this embodiment of the invention, when constructing the MIP model, five types of core constraints are set in conjunction with the actual business rules of intelligent transportation operation and maintenance: event confidence admission constraints based on the Big M method according to the preset confidence admission threshold and the original confidence level; mutual exclusion constraints for repeated events in the same time and space; capacity constraints for each push channel; constraints that only valid events can be assigned to push channels; and domain constraints for discrete and continuous decision variables, wherein:

[0119] (1) The event confidence admission constraint constructed based on the Big M method according to the preset confidence admission threshold and the original confidence level includes:

[0120] Based on the preset confidence threshold and the original confidence level, the following constraint expression is constructed using the Big M method to obtain the event confidence threshold constraint:

[0121] ,

[0122] in, This represents the confidence level of an effective decision regarding a traffic incident, where τ is a pre-set confidence threshold. The original confidence level of the traffic incident is M, where M is a large M constant with values ​​in the interval [0,1].

[0123] The aforementioned event confidence threshold constraint is a two-way coupled constraint; only events meeting the confidence threshold are considered valid events, preventing false alarms with low confidence from being pushed. The preferred value for M is 1, because the decision confidence level... The domain is [0, 1]. That is, a valid event is a traffic event whose valid decision confidence is greater than or equal to the preset confidence threshold.

[0124] When traffic incidents The original confidence level is insufficient. When < τ, the right side of the inequality exist Time degenerates into Smaller than what is needed on the left ≥ τ. This is mathematically infeasible, and the geometric property forces the solver to be explicitly set. .

[0125] When traffic incidents The original confidence level is sufficient. When ≥ τ, the solver can be freely selected. At this point, the MIP model will be driven by the "maximizing profit" objective function, making... as close as possible This enables dynamic confidence calibration.

[0126] This constraint defines a value that varies with the original confidence level. The dynamically changing feasible region is also known as the dynamic contraction of the feasible region. Effective decision confidence. As a decision variable, it must take values ​​between the upper and lower limits. When When < τ, the lower limit line τ moves up to the upper limit line. Above, leading to The feasible region at the point completely disappears (Empty Set), mathematically guaranteeing that invalid events cannot enter the push notification. Meanwhile, in solving the MIP model, when the value of a certain integer variable (such as...) When a continuous variable constraint leads to a contradiction (infeasibility), the solver will automatically back off and try other branches, a process known as infeasibility-driven switching. ,when When it is not feasible, the only remaining possibility in the solution space is Therefore, without designing any conditional judgment logic, the MIP model achieves hard threshold truncation through the self-consistency of mathematical constraints. Furthermore, In the middle, when When, the constraint becomes ,because It has an upper bound of 1, so setting M=1 guarantees that when When the constraint does not impose any restrictions (i.e., a relaxed constraint), the solver can safely classify traffic events as invalid events even with low confidence.

[0127] In this embodiment of the invention, for example, the preset confidence threshold τ = 0.65. In this embodiment, the event confidence threshold constraint constructed based on the Big M method, compared to a simple "hard threshold truncation," strongly couples confidence verification with event selection. Subsequent business logic can adjust the value of the preset confidence threshold τ as needed, and the MIP model can automatically reassess feasibility without rewriting the constraints of the MIP model or designing any conditional judgment logic. Forced filtering of low-confidence events is achieved solely through the self-consistency of mathematical constraints.

[0128] (2) The constraint expression for the mutual exclusion constraint of repeated events in the same space-time is:

[0129]

[0130] Indicates a traffic incident and For repeated events in the same space-time, Indicates a traffic incident and They cannot both be valid events simultaneously;

[0131] In this embodiment of the invention, through the mutual exclusion constraint of repeated events in the same space-time, for events determined to be repeated ( Only one valid event can be pushed to avoid duplicate pushes. This is because if two events are duplicates (…), then only one valid event can be pushed. ),but and It cannot be 1 at the same time, meaning only one of them will be pushed. During the solver's solution process, the MIP model will automatically select the event with higher priority and higher confidence.

[0132] (3) The constraint expression for the capacity constraint of each push channel is:

[0133]

[0134] in, Let J be the maximum concurrent capacity of the j-th push channel, and J be the total number of push channels. For traffic incidents Push to the j-th push channel.

[0135] In this embodiment of the invention, each push channel has an upper limit on the number of concurrent pushes, and channel overload is avoided by limiting the capacity of the push channels (e.g., a mobile app can push a maximum of 10 events at a time). The maximum concurrent capacity of the j-th push channel is defined as follows. The maximum concurrent capacity can be preset according to the channel performance, such as setting the maximum concurrent capacity of the traffic management platform to 20, the maximum concurrent capacity of the operation and maintenance screen to 15, and the maximum concurrent capacity of the mobile APP to 10.

[0136] (4) The constraint expression for the constraint that only valid events can be assigned to the push channel is:

[0137] ,

[0138]

[0139] In this embodiment of the invention, through the above-mentioned push association constraints, only valid events ( ) can be assigned to push channels, invalid events ( Channels cannot be assigned; ensure no invalid events are pushed to the channels.

[0140] (5) The constraint expressions for the domain constraints of the discrete and continuous decision variables are as follows:

[0141] ,

[0142] in, This indicates that the i-th traffic incident is determined to be a valid incident; This indicates that the i-th traffic incident is determined to be an invalid incident; Indicates a traffic incident and As an independent event; This indicates that the i-th traffic incident is assigned to the j-th push channel; This indicates that the i-th traffic incident is not assigned to the j-th push channel.

[0143] In this embodiment of the invention, the domain constraints of the push variables described above clarify the value range of all variables, ensuring that the model conforms to the MIP definition. Through the above five core constraints, the event authenticity judgment variable, the repeated event association variable, the push channel allocation variable, and the effective decision confidence are tightly coupled together through complex mathematical constraints (especially the Big M method and mutual exclusion constraints). Specifically:

[0144] (1) Variables for determining the authenticity of an event Associated variables with recurring events Deduplication and mutual exclusion constraints: The implementation logic is to associate variables with repeated events. It is a variable for determining the truth or falsity of an event. The pre-filter. If (If determined to be a duplicate), then the constraint is... It takes effect. This means the variable... and The value cannot be 1 in both repeated event pairs. The solver must choose the variable with the greater benefit and set it to 1, while forcing the other to be 0.

[0145] (2) Variables for determining the authenticity of events With confidence in effective decision-making Event confidence threshold constraint: When a traffic event The original confidence level is insufficient. When < τ, the right side of the inequality , exist Time degenerates into Smaller than what is needed on the left ≥ τ. This is mathematically infeasible, therefore the solver must be forced to set it. When traffic incidents The original confidence level is sufficient. When ≥ τ, the solver can be freely selected. At this point, the MIP model will be driven by the "maximizing benefit" objective function, thus increasing the confidence level of effective decisions. As close as possible to the original confidence level This enables dynamic confidence level calibration. Event authenticity determination variables. The value directly determines the confidence level of effective decisions. The value space of the effective decision confidence. The rationality of this, in turn, limits the variables used to determine the truth or falsehood of an event. The feasibility of this.

[0146] (3) Variables for determining the authenticity of events Variables allocated to push channels Logical association constraints: Implementing variable allocation for push channels It is a variable for determining the truth or falsity of an event. Subordinate variables. Constraints Ensure that only events deemed valid are considered valid. Only then can channels be allocated. If Then all related It must be forced to be 0.

[0147] (4) Compared with the existing technology that treats event deduplication and push channel determination as two independent steps in a sequential manner, which may lead to overload of a certain push channel even after deduplication, the embodiments of the present invention use repeated event association variables. Assigning variables to push channels Due to the constraint coupling, the MIP model can proactively eliminate low-value repetitive events when resources are limited, thus avoiding overload of the push channel.

[0148] The intelligent traffic event push method provided in this embodiment of the invention uses recurring event association variables. Simultaneously participating in the objective function and capacity constraints couples deduplication decisions with channel capacity, something that sequential rule-based methods cannot achieve. When resources are limited, the MIP model can proactively select to remove certain high-priority duplicate events, freeing up capacity for higher-value events. Compared to the existing technology's "sort first, then allocate" approach (determining the order first, then filling channels according to priority), it can obtain more satisfactory push results.

[0149] In this embodiment of the invention, the process of solving the MIP model based on the standardized alarm set in step S103 to obtain a traffic event push scheme that maximizes the objective function under the constraints can be implemented using a mixed integer programming solver (MIP Solver). Specifically, a mature and practically applicable mixed integer programming optimizer can be used to solve the MIP model. In this embodiment, the global optimal solution is obtained through the solver.

[0150] In this embodiment of the invention, by constructing the MIP model, compared with the LP relaxation rounding method in the prior art, the integer constraint of MIP avoids the problem of infeasible solutions caused by violating capacity constraints after rounding of LP relaxation solutions during the global optimization process.

[0151] In one specific embodiment, in the above-described intelligent traffic event push method, to control computational complexity, after obtaining the standardized alarm set, the above-described recurring event association variables are constructed. The method may further include:

[0152] Traffic events in the standardized alarm set are pre-filtered according to preset lightweight rules, and duplicate event pairs in the same time and space are generated for the pre-filtered traffic events to obtain the duplicate event pair set.

[0153] The aforementioned preset lightweight rules can be set according to actual conditions. They can be constraints based on spatiotemporal distance and time difference, such as a spatiotemporal distance ≤ 50 meters and a time difference ≤ 30 seconds. Pre-screening is performed according to the preset lightweight rules to pre-filter traffic events in the standardized alarm set. Only events in the same spatiotemporal space that meet the rules are selected as candidate duplicate event pairs, generating duplicate event association variables. Instead of the full combination, the size of candidate duplicate event pairs is reduced from O(N²) to O(kN²) (typically k≈0.05, meaning only 5% of duplicate event pairs enter the MIP solution), ensuring that the MIP model solver converges within an acceptable time. It's important to note that because the duplicate event association variable itself is a decision variable output by the MIP model solver, rather than a predefined "decision result," the pre-screening stage yields candidate duplicate event pairs, and the duplicate event association variable z is assigned a value of 0 or 1 in the solver.

[0154] In this embodiment of the invention, a pre-screening mechanism is used to control the size of candidate duplicate pairs to the level of O(kN²) (k≈0.05), ensuring that the solver converges within the time required to meet the event push requirements, such as within 5 to 10 seconds.

[0155] In one specific embodiment, step S103 above, which involves solving the MIP model based on the standardized alarm set to obtain a traffic event push scheme that maximizes the objective function under the constraints, includes:

[0156] Traffic events in the standardized alarm set are obtained according to a preset time sliding window and added to the buffer queue;

[0157] The MIP model is solved based on the traffic events in the buffer queue to obtain a traffic event push scheme that maximizes the objective function under the constraints.

[0158] In this embodiment of the invention, a periodic batch processing method is used by setting a time sliding window. During the window period, newly arrived standardized alarms are continuously stored in a buffer queue. When the time limit is reached or the buffer reaches a certain size (e.g., N > 10), a "trigger signal" is sent to the solver. The size of the preset time sliding window can be set according to actual conditions, for example, 5-10 seconds. Then, a solution is triggered every 5-10 seconds, collecting alarm data from all traffic events within the time sliding window for global calculation. This allows the MIP model to perform global optimization, obtaining a traffic event push scheme that maximizes the objective function under constraints, rather than blindly pushing alarms one by one. Periodic batch processing not only improves the quality of globally optimal decisions but also meets the timeliness requirements of high-level traffic events.

[0159] In this embodiment of the invention, the traffic event push scheme that maximizes the objective function under the given constraints includes all valid events, their priority ranking, and push channels; specifically, it can be:

[0160] List of valid events: All The event;

[0161] Event Priority Sorting: Sort by priority score from highest to lowest to clarify the push order of each valid event;

[0162] Channel allocation scheme: Each valid event corresponds to a push channel to ensure balanced channel load.

[0163] In one specific embodiment, after obtaining the allocation scheme that maximizes the objective function under the constraints, the method further includes:

[0164] All valid events are pushed to the corresponding push channels in priority order and recorded.

[0165] In this embodiment of the invention, traffic events with high priority scores (such as traffic accidents, violations of hazardous chemicals, etc., which are the highest priority events and are pushed to all channels) are prioritized by sorting them by priority. Furthermore, events are pushed to the corresponding channels according to the channel allocation scheme to ensure that the content pushed by different channels is adapted to their usage scenarios (such as the traffic management platform pushing all valid events, and the mobile APP pushing high-priority emergency events). In addition, after the push is completed, the push time, channel, and reception status are recorded to form a push log for subsequent operation and maintenance review and parameter optimization.

[0166] The intelligent traffic event push method provided in this invention processes the acquired multi-source traffic event raw data into standardized alarm data. It uses the validity of the traffic event, duplicate events in the same time and space, and the push channel of the traffic event as discrete decision variables, and the confidence level of the valid decision and the event priority as continuous decision variables. The optimization objectives are to maximize the benefits of valid event push, minimize the loss of low-confidence false alarms, and minimize redundant push of duplicate events. Multiple constraints are coupled together to construct a MIP model. The method solves the traffic event push scheme that maximizes the objective function under the constraints. Through the global optimization capability of MIP, it solves four core problems: event deduplication, authenticity verification, priority ranking, and channel allocation. This effectively avoids duplicate and false alarms of traffic events, enables push of traffic events according to priority, and achieves hierarchical and precise push based on event priority. It ensures load balancing of the push channels, achieving a dual improvement in the quality and efficiency of traffic event push, and meeting the accuracy and real-time requirements of traffic event push.

[0167] The intelligent traffic event push method provided in this invention breaks through the limitations of the traditional serial processing of "deduplication, confirmation, sorting, and allocation," and introduces Mixed Integer Programming (MIP) into the entire process of intelligent traffic event confirmation and push. By transforming event authenticity judgment, duplicate event association, and channel allocation into discrete decision variables, and confidence and priority into continuous variables, an objective function that takes into account "effective event retention, false alarm suppression, duplicate penalty, and channel balance" is constructed. An MIP model is built in conjunction with multiple business constraints, and the four stages are processed simultaneously within a single solution framework. The local optima of multiple stages are transformed into global optimal decisions, and the globally optimal solution is obtained. This achieves joint optimization of multiple dimensions such as event deduplication, authenticity confirmation, priority sorting, and channel allocation, thus achieving global optimization of traffic event push decisions.

[0168] The intelligent traffic event push method provided in this embodiment of the invention can automatically prioritize events based on their value in scenarios where push channel capacity is limited (such as when the push slots of a mobile APP are full). This ensures that each push message is the most valuable traffic event at the current moment, thus avoiding low-value events from occupying the push channel.

[0169] Example 2

[0170] Based on the same inventive concept, embodiments of the present invention also provide an intelligent traffic event push device, referring to... Figure 2 As shown, it includes:

[0171] The first acquisition module 101 is used to acquire alarm data of multi-source traffic events and preprocess them to form a standardized alarm set; wherein the preprocessed alarm data of traffic events includes event number, event type, spatiotemporal coordinates, impact range, hazard level and original confidence level;

[0172] The second acquisition module 102 is used to construct a MIP model; the MIP model is constructed based on the designed decision variables, with the optimization objectives of maximizing the benefits of effective event push, minimizing the loss of low-confidence false alarms and minimizing redundant push of duplicate events, and setting constraints.

[0173] The solution module 103 is used to solve the MIP model based on the standardized alarm set to obtain a traffic event push scheme that maximizes the objective function under the constraints.

[0174] Example 3

[0175] Based on the same inventive concept, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the intelligent traffic event push method as described in Embodiment 1.

[0176] The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires; a portable computer disk drive; a hard disk drive; a random access memory (RAM); a read-only memory (ROM); an erasable programmable read-only memory (EPROM); a register; a hard disk drive; an optical fiber; a portable compact disc read-only memory (CD-ROM); an optical storage device; a magnetic storage device; or any suitable combination thereof; or any other form of computer-readable storage medium known in the art. An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium may also be a component of the processor. The processor and the storage medium may reside in an application-specific integrated circuit (ASIC). In the embodiments of this application, the computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0177] Example 4

[0178] Based on the same inventive concept, this embodiment of the invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the intelligent traffic event push method as described in Embodiment 1.

[0179] Figure 3 A possible structural diagram of the computer device involved in the above embodiments is shown. The computer device includes a processor 1002 and a communication interface 1003. The processor 1002 is used to control and manage the operation of the computer device, for example, executing the alarm association method described above, and / or other processes of the technology described herein. The communication interface 1003 is used to support communication between the computer device and other network entities, for example, executing the steps performed by the communication unit 902 described above. The computer device may also include a memory 1001 and a bus 1004, the memory 1001 being used to store the program code and data of the computer device.

[0180] The memory 1001 may be a memory in a computer device, and the memory may include: volatile memory, such as random access memory; the memory may also include: non-volatile memory, such as read-only memory, flash memory, hard disk or solid-state drive; the memory may also include: a combination of the above types of memory.

[0181] The processor 1002 described above can implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. The processor can be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. The processor can also be a combination that implements computing functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0182] Bus 1004 can be an Extended Industry Standard Architecture (EISA) bus, etc. Bus 1004 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0183] Example 5

[0184] Based on the same inventive concept, this embodiment of the invention also provides a computer program product containing instructions, which, when run on a computer device, causes the computer device to execute the intelligent traffic event push method as described in Embodiment 1.

[0185] The principles by which the above-mentioned devices, media, and related products in the embodiments of the present invention solve the problem are similar to those of the aforementioned methods. Therefore, their implementation can refer to the implementation of the aforementioned methods, and repeated details will not be repeated.

[0186] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0187] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0188] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0189] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0190] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for intelligent traffic event push notification, characterized in that, include: Acquire alarm data from multiple traffic incidents and preprocess it to form a standardized alarm set; The preprocessed traffic incident alarm data includes the incident number, incident type, spatiotemporal coordinates, impact range, hazard level, and original confidence level. A MIP model is constructed based on the designed decision variables, with the optimization objectives of maximizing the benefits of effective event push, minimizing the loss of low-confidence false alarms, and minimizing redundant push of duplicate events, and constraints are set. The MIP model is solved based on the standardized alarm set to obtain a traffic event push scheme that maximizes the objective function under the constraints.

2. The intelligent traffic event push method according to claim 1, characterized in that, The MIP decision variables include: discrete decision variables such as whether the traffic event is effective, repeated events in the same time and space, and the channel through which the traffic event is pushed; and continuous decision variables such as the confidence level of the effective decision and the event priority. The objective function is: ， Where I represents the standardized alarm set, For the i-th traffic event in the standardized alarm set, Let α be the priority score for the i-th traffic incident, β be the false alarm penalty coefficient, and β be the repeat incident penalty coefficient. Let be a set of repeating event pairs, where, For traffic incidents and A pair of repeated events in the same space-time.

3. The intelligent traffic event push method according to claim 2, characterized in that, The false alarm penalty coefficient is calculated as follows: ， Where L represents the current load rate of all channels. The base false alarm penalty coefficient, For high load sensitivity parameters, For low load sensitivity parameters, The high load trigger threshold, The low load trigger threshold, This is the threshold for full load conditions.

4. The intelligent traffic event push method according to claim 3, characterized in that, The load rate of all current push channels is calculated as follows: ， in, Let be the weight coefficient of the k-th push channel, and the sum of the weight coefficients of all K push channels is 1; Let be the real-time utilization rate of the k-th push channel, and let be the proportion of traffic event tasks currently in active push status in the k-th push channel to the total channel capacity.

5. The intelligent traffic event push method according to claim 2, characterized in that, The constraints include event confidence admission constraints constructed based on the Big M method according to the preset confidence admission threshold and the original confidence, mutual exclusion constraints for repeated events in the same space-time, capacity constraints for each push channel, constraints that only valid events can be assigned to the push channel, and domain constraints for discrete and continuous decision variables.

6. The intelligent traffic event push method according to claim 2, characterized in that, Also includes: Traffic events in the standardized alarm set are pre-filtered according to preset lightweight rules, and duplicate event pairs in the same time and space are generated for the pre-filtered traffic events to obtain the duplicate event pair set.

7. An intelligent traffic event push device, characterized in that, include: The first acquisition module is used to acquire alarm data from multi-source traffic events and preprocess it to form a standardized alarm set. The preprocessed traffic incident alarm data includes the incident number, incident type, spatiotemporal coordinates, impact range, hazard level, and original confidence level. The second acquisition module is used to construct the MIP model; The MIP model is constructed based on the designed decision variables, with the optimization objectives of maximizing the benefits of effective event push, minimizing the loss of low-confidence false alarms, and minimizing redundant push of duplicate events, and with constraints set. The solution module is used to solve the MIP model based on the standardized alarm set to obtain a traffic event push scheme that maximizes the objective function under the constraints.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the intelligent traffic event push method as described in any one of claims 1-6.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the intelligent traffic event push method as described in any one of claims 1-6.

10. A computer program product containing instructions, which, when run on a computer device, causes the computer device to perform the intelligent traffic event push method as described in any one of claims 1-6.

Images