Multi-source data-based traffic event road network traffic capacity attenuation prediction method

Abstract

The invention discloses a traffic event road network traffic capacity attenuation prediction method based on multi-source data. The method comprises the following steps: collecting and preprocessing multi-source data; decomposing multi-dimensional influence factors; constructing a space-time attenuation model; analyzing a network propagation effect; and generating and optimizing a prediction result. Compared with the prior art, the method is characterized in that a multi-dimensional influence factor coupling mechanism is established; a spatio-temporal evolution prediction mechanism is established; the method has the following beneficial effects: the prediction precision is improved; the calculation efficiency is improved; and the method has good application value.

Description

Technical Field

[0001] The invention relates to a method for predicting the attenuation of road network capacity during traffic incidents, and belongs to the technical field of intelligent transportation. Background Art

[0002] The reduction in road network capacity caused by traffic incidents has the characteristics of spatiotemporal propagation. The existing methods for predicting the attenuation of road network capacity caused by traffic incidents include:

[0003] A Chinese invention patent application with publication number CN109661692A, published on April 19, 2019, discloses a device and a terminal device, including: collecting traffic data, the traffic data including basic road network data and floating vehicle data; fusing the basic road network data and floating vehicle data to obtain corresponding event features; comparing the event features with event features of predefined traffic events; if the event features are the same as the event features of the predefined traffic events, then taking the traffic event corresponding to the event features as the predicted traffic event.

[0004] The Chinese invention patent application with publication number CN113065691A, published on July 2, 2021, discloses a traffic behavior prediction method, the method comprising: receiving first data in the current road scene, the first data representing the road data from the perspective of the first user in the current road scene; obtaining second data in the current road scene, the second data representing the road data from the perspective of at least one second user in the current road scene; establishing a road prediction model in the current road scene based on the mapping relationship between the first data and the second data; using third data as the input of the road prediction model to predict road traffic events, and outputting it to the terminal of the first user for display, the third data representing the road data from the perspective of the first user in the next road scene. At the same time, the invention patent application also discloses a traffic behavior prediction system.

[0005] The above-mentioned traffic event road network capacity attenuation prediction methods have the following main shortcomings:

[0006] 1. It is difficult to accurately quantify the coupling effect of multiple sources of influencing factors;

[0007] 2. Unable to dynamically capture the spatiotemporal evolution of capacity attenuation;

[0008] 3. The prediction model has poor adaptability and the parameter optimization mechanism is imperfect. Summary of the invention

[0009] The technical problem to be solved by the present invention is that the existing methods for predicting the attenuation of road network capacity due to traffic events mainly have the following shortcomings: it is difficult to accurately quantify the coupling effect of multi-source influencing factors; it is impossible to dynamically capture the spatiotemporal evolution law of capacity attenuation; the prediction model has poor adaptability and the parameter optimization mechanism is imperfect.

[0010] In order to solve the above technical problems, the technical solution of the present invention is to disclose a method for predicting the attenuation of road network capacity due to traffic events based on multi-source data, which is characterized by comprising the following steps:

[0011] Step 1: Multi-source data collection and preprocessing, including the following steps:

[0012] Step 101: multi-source data collection, where the collected multi-source data includes event data, road network data, and environmental data;

[0013] Step 102, preprocessing the multi-source data obtained in step 101, including outlier processing, missing value processing and data standardization, to obtain a standardized feature vector;

[0014] Step 2: Decompose multi-dimensional impact factors to achieve multi-dimensional impact quantification, including:

[0015] Step 201, calculate the event impact We:

[0016] We=α1T+α2S+α3D

[0017] Where: T is the event type impact coefficient, which is the basic impact degree determined based on the event type and level, T = f_type (event type, level);

[0018] S is the degree of space occupancy, which is the ratio of the number of occupied lanes to the total number of lanes, that is, S = number of occupied lanes / total number of lanes;

[0019] D is the time continuity, which is an exponential decay function based on the event duration, D = 1-exp(-λt);

[0020] α1, α2, and α3 are weight coefficients;

[0021] Step 202: Evaluate the road network status Rn:

[0022] Rn=β1V+β2Q+β3K

[0023] Where: V is the speed effect, which is the ratio of the current speed to the free flow speed, that is, V = current speed / free flow speed;

[0024] Q is the flow impact, which is the ratio of current flow to basic capacity, i.e., Q = current flow / basic capacity;

[0025] K is the density effect, which is the ratio of the current density to the critical density, that is, K = current density / critical density;

[0026] β1, β2, and β3 are weight coefficients;

[0027] Step 203: Quantify the environmental conditions to obtain the environmental impact coefficient In:

[0028] In=γ1Ft+γ2Fw+γ3Fr

[0029] Where: Ft is the time periodicity, which is a periodic function based on the time period and day type, Ft = f_time (time period, day type);

[0030] Fw is the weather impact, which is a comprehensive assessment based on weather type and visibility, Fw = f_weather (weather, visibility);

[0031] Fr is the road condition, which is an assessment based on the road friction coefficient and the degree of water accumulation, Fr = f_road (road condition);

[0032] γ1, γ2, and γ3 are weight coefficients;

[0033] Step 204, calculate the comprehensive impact factor I, I = {We, Rn, In};

[0034] Step 3: Constructing a spatiotemporal attenuation model, including the following steps:

[0035] Step 301: Establish a time dimension attenuation function:

[0036]

[0037] In the formula, α is the basic attenuation coefficient, β is the time attenuation rate, γ is the periodic fluctuation amplitude, and ω is the angular frequency. is the phase difference;

[0038] Step 302: Establish a spatial dimension attenuation function:

[0039] S(d)=1 / [1+λ(d / d0) k ]

[0040] In the formula, d is the distance from the event point, d0 is the characteristic distance, λ is the spatial attenuation coefficient, and k is the propagation attenuation index;

[0041] Step 303: Calculate the comprehensive attenuation to obtain the attenuation characteristic matrix M

[0042] C(t,d)=C0·[I.We·I.Rn·I.In]·T(t)·S(d)

[0043]

[0044] In the formula, C0 is the initial traffic capacity;

[0045] Step 4: Network communication effects, including:

[0046] Establish a propagation mechanism and divide the influence range to obtain the network influence matrix N. The obtained network influence matrix N is expressed as:

[0047]

[0048]

[0049] Step 5: Generate prediction results, including the following steps:

[0050] Step 501: Time dimension prediction:

[0051] Prediction sequence = {t:C(t)|t∈[0,T]}

[0052] Recovery time = min{t|C(t)≥0.9C0}

[0053] Step 502: spatial dimension prediction:

[0054]

[0055] Step 503: Obtain the final prediction result R, expressed as:

[0056]

[0057] Preferably, in step 101, the event data includes features such as location coordinates (x, y), number of occupied lanes lanes_blocked, occurrence time t_start, estimated duration duration, type type and severity level, expressed as E, then:

[0058]

[0059]

[0060] Among them, road_id is the lane ID.

[0061] Preferably, in step 101, the road network data includes parameters such as road class road_class, lane number lanes_total, designed traffic capacity capacity_base, current flow flow_current, speed speed_current and density density_current, expressed as R, then:

[0062]

[0063] Preferably, in step 101, the environmental data includes information such as time period, weather type, visibility, and road surface condition, expressed as W, then:

[0064]

[0065] Among them, day_type is the day type.

[0066] Preferably, in step 4, a propagation mechanism is established taking into account direct impact, cascading effect and network regulation:

[0067] The direct impact is the change in the capacity of the road section where the event point is located. In the embodiment of the present invention, the directly affected road section j is calculated using the following formula:

[0068] C j (t) = C j0 ·[1-μ ij ·(1-C i (t) / C i0 )]

[0069] In the formula, μ ij is the impact transfer coefficient between road sections, C i0 , C j0 is the initial capacity of each;

[0070] The cascade effect is a chain reaction of adjacent sections. In the embodiment of the present invention, for k-hop adjacent sections, there are:

[0071] C_k(t)=C_k0·[1-μ k ·Π(1-C i (t) / C i0 )]

[0072] In the formula, μ k is the k-hop propagation attenuation coefficient;

[0073] Network regulation refers to the adaptive regulation capability of the entire road network.

[0074] Preferably, in step 4, when dividing the impact range, the area where the capacity is reduced by more than 40% is divided into the direct impact area, the area where the capacity is reduced by 20%-40% is divided into the secondary impact area, and the area where the capacity is reduced by 10%-20% is divided into the slightly impact area.

[0075] Preferably, a feedback optimization mechanism is also included to optimize parameters through measured data to update the model.

[0076] The invention discloses a method for predicting the attenuation of road network capacity due to traffic events based on multi-source data, including: multi-source data collection and preprocessing; multi-dimensional influencing factor decomposition; spatiotemporal attenuation model construction; network propagation effect analysis; prediction result generation and optimization. Compared with the existing technical solutions, the invention is innovative in that:

[0077] 1) A multi-dimensional influencing factor coupling mechanism was established, including: the first three-dimensional (event-road network-environment) influencing factor decomposition method; the establishment of a dynamic weight adaptive allocation mechanism; and the realization of unified quantitative processing of multi-source heterogeneous data;

[0078] 2) A prediction mechanism for spatiotemporal evolution was established, including: proposing a family of attenuation functions based on spatiotemporal coupling; innovatively introducing joint modeling of periodic fluctuations and spatial propagation; and realizing automatic correction of multi-scale prediction results;

[0079] 3) Conduct network cascade effect analysis, including: establishing a propagation model that takes into account the road network topology; proposing a hierarchical impact range definition method; and achieving accurate prediction of large-scale impacts.

[0080] Due to the adoption of the above innovative technical means, the technical solution disclosed in the present invention has the following beneficial effects compared with the prior art:

[0081] Improved prediction accuracy, including a 40% reduction in capacity prediction error, a 35% increase in impact area definition accuracy, and a 45% reduction in recovery time prediction deviation;

[0082] Improved computing efficiency, including single-point prediction response time <1s, network-level prediction response time <5s, and parameter optimization cycle <5min;

[0083] It has good application value, including: supporting accurate decision-making in traffic management; providing a basis for zoning and grading control; and realizing active traffic induction. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 The figure is a specific algorithm flow chart of an embodiment of the present invention. DETAILED DESCRIPTION

[0085] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the appended claims of the application equally.

[0086] like Figure 1 As shown, a method for predicting the attenuation of road network capacity due to traffic events based on multi-source data disclosed in an embodiment of the present invention specifically includes the following steps:

[0087] Step 1: Multi-source data collection and preprocessing, including the following steps:

[0088] Step 101: Multi-source data collection

[0089] The multi-source data collected in the present invention includes event data, road network data and environmental data, among which:

[0090] The event data further includes features such as location coordinates (x, y), number of occupied lanes lanes_blocked, occurrence time t_start, expected duration duration, type type and severity level, expressed as E, then:

[0091]

[0092] Among them, road_id is the lane ID.

[0093] The road network data further includes parameters such as road class road_class, lane number lanes_total, designed traffic capacity capacity_base, current flow flow_current, speed speed_current and density density_current, expressed as R, then:

[0094]

[0095] Environmental data further includes information such as time period, weather type, visibility, and road surface condition, expressed as W, then:

[0096]

[0097] Among them, day_type is the day type.

[0098] Step 102: Data preprocessing

[0099] The multi-source data obtained in step 101 is preprocessed, including outlier processing, missing value processing and data standardization. In the embodiment of the present invention, the 3σ rule is used to identify and process outliers in the multi-source data obtained in step 101, and the spatiotemporal interpolation method is used to supplement the missing data in the multi-source data obtained in step 101. Finally, data standardization is used to unify the format and scale of data from different sources to obtain a standardized feature vector

[0100]

[0101] Step 2: Decompose multi-dimensional impact factors to achieve multi-dimensional impact quantification, including:

[0102] Step 201: Calculate the event impact We:

[0103] We=α1T+α2S+α3D

[0104] Where: T is the event type impact coefficient, which is the basic impact degree determined based on the event type and level, T = f_type (event type, level);

[0105] S is the degree of space occupancy, which is the ratio of the number of occupied lanes to the total number of lanes, that is, S = number of occupied lanes / total number of lanes;

[0106] D is the time continuity, which is an exponential decay function based on the event duration, D = 1-exp(-λt);

[0107] α1, α2, and α3 are weight coefficients;

[0108] Step 202: Evaluate the road network status Rn:

[0109] Rn=β1V+β2Q+β3K

[0110] Where: V is the speed effect, which is the ratio of the current speed to the free flow speed, that is, V = current speed / free flow speed;

[0111] Q is the flow impact, which is the ratio of current flow to basic capacity, i.e., Q = current flow / basic capacity;

[0112] K is the density effect, which is the ratio of the current density to the critical density, that is, K = current density / critical density;

[0113] β1, β2, and β3 are weight coefficients;

[0114] Step 203: Quantify the environmental conditions to obtain the environmental impact coefficient In:

[0115] In=γ1Ft+γ2Fw+γ3Fr

[0116] Where: Ft is the time periodicity, which is a periodic function based on the time period and day type, Ft = f_time (time period, day type);

[0117] Fw is the weather impact, which is a comprehensive assessment based on weather type and visibility, Fw = f_weather (weather, visibility);

[0118] Fr is the road condition, which is an assessment based on the road friction coefficient and the degree of water accumulation, Fr = f_road (road condition);

[0119] γ1, γ2, and γ3 are weight coefficients;

[0120] Step 204: Calculate the comprehensive impact factor I, where I = {We, Rn, In}.

[0121] Step 3: Constructing a spatiotemporal attenuation model, including the following steps:

[0122] Step 301: Establish a time dimension attenuation function:

[0123]

[0124] In the formula, α is the basic attenuation coefficient, β is the time attenuation rate, γ is the periodic fluctuation amplitude, and ω is the angular frequency. is the phase difference;

[0125] In the above time dimension attenuation function, the exponential attenuation function is used to describe the weakening of the impact of the event over time, which is the basic attenuation term, the sine function is used to describe the daily cycle change characteristics, which is the periodic fluctuation term, and the combined weight of the multi-dimensional influencing factors is used as the adjustment coefficient;

[0126] Step 302: Establish a spatial dimension attenuation function:

[0127] S(d)=1 / [1+λ(d / d0) k ]

[0128] In the formula, d is the distance from the event point, d0 is the characteristic distance, λ is the spatial attenuation coefficient, and k is the propagation attenuation index;

[0129] In the above spatial dimension attenuation function, the distance attenuation is represented by the power function attenuation based on the distance from the event point, and the spatial propagation characteristics of the road network connectivity are considered based on the topological relationship. At the same time, the asymmetric attenuation of the upstream and downstream impact differences is considered based on the directionality;

[0130] Step 303: Calculate the comprehensive attenuation to obtain the attenuation characteristic matrix M

[0131] C(t,d)=C0·[I.We·I.Rn·I.In]·T(t)·S(d)

[0132]

[0133] In the formula, C0 is the initial traffic capacity;

[0134] In the above-mentioned comprehensive attenuation model, time-space coupling is realized through the product form of time attenuation and space attenuation, and the combined adjustment mechanism of multi-dimensional influencing factors is adopted to realize impact adjustment. At the same time, the constraint conditions of the actual capacity limit are considered as boundary constraints.

[0135] Step 4: Network communication effects, including:

[0136] Establish a propagation mechanism and divide the scope of influence to obtain the network influence matrix N, where:

[0137] Establish a communication mechanism that takes into account direct impact, cascading effects, and network regulation:

[0138] The direct impact is the change in the capacity of the road section where the event point is located. In the embodiment of the present invention, the directly affected road section j is calculated using the following formula:

[0139] C j (t) = C j0 ·[1-μ ij ·(1-C i (t) / C i0 )]

[0140] In the formula, μ ij is the impact transfer coefficient between road sections, C i0 , C j0 is the initial capacity of each;

[0141] The cascade effect is a chain reaction of adjacent sections. In the embodiment of the present invention, for k-hop adjacent sections, there are:

[0142] C_k(t)=C_k0·[1-μ k ·Π(1-C i (t) / C i0 )]

[0143] In the formula, μ k is the k-hop propagation attenuation coefficient;

[0144] Network regulation refers to the adaptive regulation capability of the entire road network;

[0145] When dividing the impact range, the area where the traffic capacity is reduced by more than 40% is classified as the direct impact area, the area where the traffic capacity is reduced by 20%-40% is classified as the secondary impact area, and the area where the traffic capacity is reduced by 10%-20% is classified as the slightly impact area;

[0146] The obtained network influence matrix N is expressed as:

[0147]

[0148] Step 5: Generate prediction results, including the following steps:

[0149] Step 501: Time dimension prediction:

[0150] Prediction sequence = {t:C(t)|t∈[0,T]}

[0151] Recovery time = min{t|C(t)≥0.9C0}

[0152] Step 502: spatial dimension prediction:

[0153]

[0154] Step 503: Obtain the final prediction result R

[0155]

[0156] In the above steps, the data flow depends on:

[0157] -Step 1 → Step 2: The eigenvector F is the basis for factor decomposition

[0158] -Step 2 → Step 3: Impact factor I determines the attenuation model parameters

[0159] -Step 3 → Step 4: Attenuation feature M guides network propagation calculation

[0160] -Step 4 → Step 5: The network influence N forms the basis for the final prediction.

[0161] In addition, the method for predicting the attenuation of road network capacity due to traffic events based on multi-source data disclosed in the present invention also includes a feedback optimization mechanism, and its steps can be roughly described as follows:

[0162] Measured data → parameter optimization → model update

[0163] Further including:

[0164] Error calculation: E = Σ(C i -C i ′) 2 ;

[0165] Parameter update:

[0166] Model correction: regular / real-time updates.

Claims

1. A method for predicting the attenuation of road network capacity due to traffic events based on multi-source data, characterized in that: The following steps are involved: Step 1: Multi-source data collection and preprocessing, including the following steps: Step 101: multi-source data collection, where the collected multi-source data includes event data, road network data, and environmental data; Step 102, preprocessing the multi-source data obtained in step 101, including outlier processing, missing value processing and data standardization, to obtain a standardized feature vector; Step 2: Decompose multi-dimensional impact factors to achieve multi-dimensional impact quantification, including: Step 201: Calculate the event impact We: We=α1T+α2S+α3D Where: T is the event type impact coefficient, which is the basic impact degree determined based on the event type and level, T = f_type (event type, level); S is the degree of space occupancy, which is the ratio of the number of occupied lanes to the total number of lanes, that is, S = number of occupied lanes / total number of lanes; D is the time continuity, which is an exponential decay function based on the event duration, D = 1-exp(-λt); α1, α2, and α3 are weight coefficients; Step 202: Evaluate the road network status Rn: Rn=β1V+β2Q+β3K Where: V is the speed effect, which is the ratio of the current speed to the free flow speed, that is, V = current speed / free flow speed; Q is the flow impact, which is the ratio of current flow to basic capacity, i.e., Q = current flow / basic capacity; K is the density effect, which is the ratio of the current density to the critical density, that is, K = current density / critical density; β1, β2, β3 are weight coefficients; Step 203: Quantify the environmental conditions to obtain the environmental impact coefficient In: In=γ1Ft+γ2Fw+γ3Fr Where: Ft is the time periodicity, which is a periodic function based on the time period and day type, Ft = f_time (time period, day type); Fw is the weather impact, which is a comprehensive assessment based on weather type and visibility, Fw = f_weather (weather, visibility); Fr is the road condition, which is an assessment based on the road friction coefficient and the degree of water accumulation, Fr = f_road (road condition); γ1, γ2, and γ3 are weight coefficients; Step 204, calculate the comprehensive impact factor I, I = {We, Rn, In}; Step 3: Constructing a spatiotemporal attenuation model, including the following steps: Step 301: Establish a time dimension attenuation function: In the formula, α is the basic attenuation coefficient, β is the time attenuation rate, γ is the periodic fluctuation amplitude, and ω is the angular frequency. is the phase difference; Step 302: Establish a spatial dimension attenuation function: S(d)=1 / [1+λ(d / d0) k ] In the formula, d is the distance from the event point, d0 is the characteristic distance, λ is the spatial attenuation coefficient, and k is the propagation attenuation index; Step 303: Calculate the comprehensive attenuation to obtain the attenuation characteristic matrix M C(t,d)=C0·[I.We·I.Rn·I.In]·T(t)·S(d) M＝{ Time series:[t1:C1,t2:C2,...,t n :C n ], Spatial distribution: [d1:C1,d2:C2,...,d m :C m ] } In the formula, C0 is the initial traffic capacity; Step 4: Network communication effects, including: Establish a propagation mechanism and divide the influence range to obtain the network influence matrix N. The obtained network influence matrix N is expressed as: N＝{ Direct impact: {road_id:capacity_ratio}, Secondary impact: {road_id:capacity_ratio}, Propagation path:{source_id:[affected_ids]} }。 Step 5: Generate prediction results, including the following steps: Step 501: Time dimension prediction: Prediction sequence = {t:C(t)|t∈[0,T]} Recovery time = min{t|C(t)≥0.9C0} Step 502: spatial dimension prediction: Impact Partition = { Severely affected area: C(t)<0.6C0, Medium impact area: [0.6C0, 0.8C0), Slightly affected area: [0.8C0, 0.9C0) } Step 503: Obtain the final prediction result R, expressed as: R＝{ Time prediction: Attenuation process: [t→capacity], Recovery time: t_recovery }, Spatial prediction: Impact range: [area → degree], Propagation path: [segment → sequence] }, Warning information: Level: level, Suggestions } }。 2. The method for predicting the attenuation of road network capacity due to traffic events based on multi-source data as claimed in claim 1, characterized in that: In step 101, the event data includes the location coordinates (x, y), the number of occupied lanes lanes_blocked, the occurrence time t_start, the estimated duration duration, the type type and the severity level, etc., denoted as E, then: E＝{ Location: (x,y,road_id), Time: (t_start, duration), Features: (type,severity,lanes_blocked) } Among them, road_id is the lane ID.

3. The method for predicting the attenuation of road network capacity due to traffic incidents based on multi-source data as claimed in claim 1, characterized in that: In step 101, the road network data includes parameters such as road class road_class, lane number lanes_total, designed traffic capacity capacity_base, current flow flow_current, speed speed_current and density density_current, expressed as R, then: R＝{ Static features: (road_class, lanes_total, capacity_base), Dynamic state: (flow_current, speed_current, density_current) }。 4. The method for predicting the attenuation of road network capacity due to traffic events based on multi-source data as claimed in claim 1, characterized in that: In step 101, the environmental data includes information such as time period, weather type, visibility, and road surface condition, expressed as W, then: W＝{ Time attributes: (time_period, day_type), Weather status: (weather_type, visibility), Road condition: (surface_condition) } Among them, day_type is the day type.

5. The method for predicting the attenuation of road network capacity due to traffic incidents based on multi-source data as claimed in claim 1, characterized in that: In step 4, establish a propagation mechanism that takes into account direct impact, cascading effects, and network mediation: The direct impact is the change in the capacity of the road section where the event point is located. In the embodiment of the present invention, the directly affected road section j is calculated using the following formula: C j (t)=C j0 ·[1-μ ij ·(1-C i (t) / C i0 )] In the formula, μ ij is the impact transfer coefficient between road sections, C i0 , C j0 is the initial capacity of each; The cascade effect is a chain reaction of adjacent sections. In the embodiment of the present invention, for k-hop adjacent sections, there are: C_k(t)=C_k0·[1-μ k ·Π(1-C i (t) / C i0 )] In the formula, μ k is the k-hop propagation attenuation coefficient; Network regulation refers to the adaptive regulation capability of the entire road network.

6. The method for predicting the attenuation of road network capacity due to traffic incidents based on multi-source data as claimed in claim 1, characterized in that: In step 4, when dividing the impact range, the area where the capacity is reduced by more than 40% is divided into the direct impact area, the area where the capacity is reduced by 20%-40% is divided into the secondary impact area, and the area where the capacity is reduced by 10%-20% is divided into the slightly impact area.

7. The method for predicting the attenuation of road network capacity due to traffic incidents based on multi-source data as claimed in claim 1, characterized in that: It also includes a feedback optimization mechanism to optimize parameters through measured data to update the model.

Images