A macro-micro fusion trajectory reconstruction method applied to emission estimation

Abstract

This invention belongs to the interdisciplinary field of electrical data processing technology and transportation engineering, specifically relating to a macro-micro fusion trajectory reconstruction method applied to emissions estimation. It includes the following steps: S1. Generating a regional macroscopic velocity field reference; S2. Correcting the floating car operating conditions; S3. Determining a candidate trajectory set for non-floating vehicles; S4. Determining the final trajectory of non-floating vehicles that have not changed lanes based on the candidate trajectories; S5. Determining the final trajectory of non-floating vehicles that have changed lanes based on the candidate trajectories; S6. Integrating the two types of trajectories, verifying the operating conditions, and generating complete vehicle trajectories. This method can effectively align trajectory evolution with traffic wave propagation trends, significantly improving the spatiotemporal precision of the reconstructed trajectory and providing accurate input for high-resolution emissions assessment; it also effectively avoids unrealistic speed or acceleration abrupt changes common in traditional deterministic car-following models; and it has advantages such as efficient utilization of multi-source data and optimization of spatiotemporal continuity.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of electrical data processing technology and transportation engineering, and specifically relates to a macro-micro fusion trajectory reconstruction method applied to emission estimation. Background Technology

[0002] As the source of pollution closest to the urban living environment, road traffic is also a significant contributor to carbon emissions, facing the core challenge of achieving sustainable development while reducing energy consumption and emissions.

[0003] Planners and policymakers require comprehensive, high-resolution spatiotemporal emissions assessments to evaluate overall changes and hotspot shifts before and after policy or engineering interventions. Such assessments enable differentiated governance and refined management, supporting broader goals including improved air quality, climate mitigation, public health, and energy security. High-precision spatiotemporal emissions assessments are primarily achieved through road trajectory reconstruction and the subsequent assessment of emissions by linking the reconstructed trajectory to an emissions factor library. For road trajectory reconstruction, current models include Newell, IDM, and GM car-following models. Among these, the open-source Newell car-following model proposed by GF Newell has become a primary tool for high-precision spatiotemporal emissions assessment due to its stability, few parameters, and suitability for sparse data trajectory reconstruction. However, the inherent strong assumptions and insufficient characterization of vehicle dynamics constraints in existing car-following models limit their adaptability. A key challenge in conducting high-precision emissions assessments is overcoming the shortcomings of traditional deterministic car-following models in capturing real-world traffic stochastic factors and addressing the systematic bias of existing model parameter calibration strategies that focus on traffic indicators while neglecting emissions guidance. Furthermore, in terms of modeling lane-changing behavior in complex scenarios, existing models still have shortcomings in dynamic interaction and multi-source information fusion, which may lead to an overestimation of the emissions of vehicles in adjacent lanes during lane-changing behavior.

[0004] Therefore, in order to solve the problem of incomplete trajectories caused by sparse detection data, and to meet the physical rationality, spatiotemporal continuity and cross-scale consistency required for high-precision emission assessment, it is crucial to seek a macro-micro fusion trajectory reconstruction method based on road sensor data to achieve accurate emission estimation. This method can accurately reconstruct the spatiotemporal trajectory of road tracks and generate inputs for high-precision emission assessment. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a macro-micro fusion trajectory reconstruction method for emissions estimation. Based on detection data from real road sensors, it can accurately reconstruct the spatiotemporal trajectory of road tracks and generate inputs for high-precision emissions assessment.

[0006] Technical solution of the present invention:

[0007] A macro-micro fusion trajectory reconstruction method for emissions estimation includes the following steps:

[0008] S1. Generate a regional macroscopic velocity field reference;

[0009] S2. Correct abnormal operating conditions of floating cars;

[0010] S3. Determine the candidate trajectory set for non-floating vehicles;

[0011] S4. Determine the fused trajectory of vehicles that have not changed lanes among non-floating vehicles based on candidate trajectories;

[0012] S5. Determine the fused trajectory of lane-changing vehicles among non-floating vehicles based on candidate trajectories;

[0013] S6. Integrate the two types of trajectories, verify the operating conditions, and generate a complete vehicle trajectory.

[0014] This invention relies on detection data from two sensors to capture the propagation speed of traffic waves through a spatiotemporal velocity interpolation-smoothing method in the macroscopic module. Combined with a bidirectional car-following model with traffic wave-DTW synergy in the microscopic module, it aligns the microscopic trajectory evolution with the macroscopic traffic wave behavior, achieving consistent reconstruction across scales. Another Hidden Markov Model (HMM) in the microscopic module aims to characterize various driving states (including acceleration, deceleration, and cruise modes). Dynamic constraints are set through acceleration-impact boundaries to ensure physically reasonable operating conditions, effectively avoiding unrealistic abrupt changes common in traditional car-following models. In the fusion module, a virtual monitor is used to identify feasible lane-changing points under safety constraints, merging candidate trajectories into continuous paths. Upstream and downstream candidate trajectories are fused using time-varying fusion equations, ensuring spatiotemporal continuity and thus achieving reliable improvement in the accuracy of lane-level state and emission inference.

[0015] Beneficial effects

[0016] Compared with traditional methods, the present invention has the following significant advantages:

[0017] High-precision trajectory output: Through the cross-scale synergy of macroscopic velocity field reference (S1) and microscopic car-following (S2), this method can effectively align trajectory evolution with traffic wave propagation trends, significantly improve the spatiotemporal precision of reconstructed trajectories, and provide accurate input for high-resolution emission assessment.

[0018] Physical rationality guarantee: The Hidden Markov Model (HMM) is introduced and combined with the acceleration-jerk dynamic constraint (S2) to make random corrections to the floating car's operating conditions, which effectively avoids the unrealistic speed or acceleration mutations common in traditional deterministic car-following models.

[0019] Accurate Analysis of Complex Behaviors: The fusion module (S5) employs a multi-objective optimization strategy to identify and optimize lane-changing behavior while maintaining safe distance constraints. This addresses the issue of insufficient accuracy in complex interaction scenarios using traditional models and eliminates the bias caused by inaccurate lane-change identification, which leads to an overestimation of emissions from adjacent lanes.

[0020] Efficient utilization of multi-source data: This method is based on two types of real-world road detection data: fixed sensor data and floating car data. Through the DTW-traffic wave equivalent bidirectional car-following mechanism, it makes full use of the complementarity of multi-source information and effectively solves the trajectory reconstruction problem in sparse detection data environment.

[0021] Spatiotemporal continuity optimization: For vehicles that do not change lanes, time-varying weighted dynamic programming (S4) is used for trajectory fusion, which ensures smooth transition and high continuity of trajectory in time and space dimensions, and improves the reliability of trajectory in emission factor calculation. Attached Figure Description

[0022] Figure 1 This is the main flow of a macro-micro fusion trajectory reconstruction method for emission estimation according to the present invention;

[0023] Figure 2 This is a schematic diagram of the dual-filter principle in step 2 of this embodiment of the invention;

[0024] Figure 3 This is a schematic diagram of the regional macroscopic velocity field reference in an embodiment of the present invention;

[0025] Figure 4 This is a schematic diagram of the probability distribution matrix of random acceleration at the current moment in an embodiment of the present invention;

[0026] Figure 5 This is a schematic diagram of upstream and downstream floating car trajectory matching in an embodiment of the present invention;

[0027] Figure 6 This is a schematic diagram of the time delay / space delay ratio of the vehicle to be reconstructed in an embodiment of the present invention;

[0028] Figure 7 This is a schematic diagram of the candidate trajectories of the vehicle to be reconstructed according to an embodiment of the present invention;

[0029] Figure 8 This is a schematic diagram of the vehicle fusion trajectory to be reconstructed according to an embodiment of the present invention;

[0030] Figure 9 This is a schematic diagram of the decision area for the lane change point of the vehicle to be reconstructed according to an embodiment of the present invention;

[0031] Figure 10 This is a schematic diagram of the location of the lane-changing point of the vehicle to be reconstructed in an embodiment of the present invention;

[0032] Figure 11 This is a schematic diagram of the vehicle fusion trajectory to be reconstructed after a lane-changing behavior occurs, according to an embodiment of the present invention.

[0033] Figure 12 This is a schematic diagram of the complete reconstructed trajectory according to an embodiment of the present invention. Detailed Implementation

[0034] To provide a clearer understanding of the technical features, objectives, and effects of the present invention, specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0035] A macro-micro fusion trajectory reconstruction method applied to emissions estimation, such as Figure 1 As shown, it includes the following steps:

[0036] S1. Generate a regional macroscopic velocity field reference;

[0037] S2. Correct abnormal operating conditions of floating cars;

[0038] S3. Determine the candidate trajectory set for non-floating vehicles;

[0039] S4. Determine the fused trajectory of vehicles that have not changed lanes among non-floating vehicles based on candidate trajectories;

[0040] S5. Determine the fused trajectory of lane-changing vehicles among non-floating vehicles based on candidate trajectories;

[0041] S6. Integrate the two types of trajectories, verify the operating conditions, and generate a complete vehicle trajectory.

[0042] This invention relies on detection data from two types of sensors:

[0043] Mobile sensor floating car data includes vehicle identification (ID), spatial coordinates of the road at each time stamp (Time), instantaneous velocity, and instantaneous acceleration.

[0044] Fixed sensor data includes vehicle identification (ID), arrival time of the vehicle at a fixed road section (Distance), instantaneous velocity, and instantaneous acceleration.

[0045] Furthermore, in step S1, the regional macroscopic velocity field reference is generated by sparse trajectory data collected by road sensors, and the specific steps are as follows:

[0046] (1.1) Sparse trajectory data is collected using road sensors. Then, the interpolated velocity is obtained by weighted interpolation of the trajectory data using an adaptive dual filter based on free flow and congested flow. .

[0047] Specifically, the adaptive dual filter employs anisotropic kernels that skew the principal axis, and by analyzing the congestion level at spatiotemporal points, fuses the interpolation speeds obtained from the dual kernels. The calculation formula is as follows:

[0048]

[0049] in, The velocity at that spatiotemporal point is obtained by filter interpolation based on congested flow; The velocity at that spatiotemporal point is obtained based on free-flow filter interpolation; This is the weight adjustment function. The calculation formula is as follows:

[0050]

[0051] in, For the spatiotemporal point where the actual velocity value is obtained from the sensor; The speed threshold between free flow and congested flow is specified as 50 km / h in this embodiment; N and M are the maximum numbers obtained by dividing the total spatial distance and total time in the dataset by the smallest unit, respectively. In this embodiment, the smallest spatial unit is 0.1m and the smallest time unit is 0.04s. Distance coordinates for a specific number; A time coordinate with a specific number; Indicates in The transition width obtained near the value is specified as 10 km / h in this embodiment;

[0052] It is a free-flow based filter; It is a filter based on congested flow, and the calculation formula is as follows:

[0053]

[0054] Among them, spatial interpolation width Typically, it is half the height of the anisotropic kernel in the filter, and is set to 50 in this embodiment; the time interpolation width Typically, it is half the length of the anisotropic kernel in the filter; in this embodiment, it is set to 36. The specific structure of the dual filter is as follows: Figure 2 As shown.

[0055] (1.2) Apply moving average smoothing to the interpolated velocity to obtain the final regional macroscopic velocity field reference. The calculation formula is as follows:

[0056]

[0057] in, The unit spatial interval is set to 1m in this embodiment; The unit time interval is set to 0.4s in this embodiment; The window size is set to 70 in this embodiment; and It is a two-dimensional spatiotemporal index for roads.

[0058] Furthermore, in step S2, the correction of abnormal operating conditions of the floating car is determined by the collected floating car driving data, and the specific steps are as follows:

[0059] (2.1) First, the vehicle's jerk data is obtained by differential processing of the floating car's instantaneous acceleration data; then, at a speed of 0.2 m / s², the jerk data is... 2 The acceleration is divided into discrete intervals for quantization, and the corresponding jerk distribution in each interval is statistically analyzed. Then, the 97.5% quantile and 1% quantile of the jerk distribution in each acceleration interval are extracted and used as the upper and lower bounds. Finally, two continuous acceleration-jerk dynamic constraint functions are established by fitting functions to the upper and lower bounds respectively.

[0060] It should be noted that "probe vehicle" refers to a vehicle that collects road information, typically equipped with positioning devices to record its spatiotemporal coordinates and operating conditions while driving on the road; "non-probe vehicle" refers to a vehicle that normally drives on the road, whose spatiotemporal coordinate information cannot be directly obtained. Because floating vehicle data often exhibits abnormal operating conditions due to errors generated during the collection and stitching process, this embodiment classifies instantaneous velocities and accelerations exceeding the 97.5th percentile or falling below the 1st percentile of the floating vehicle sample dataset as abnormal operating conditions.

[0061] (2.2) Divide the velocity discrete intervals with a quantization interval of 0.5 m / s, and statistically analyze the corresponding acceleration distribution in each interval. Then, extract the 97.5% quantile and 1% quantile of the acceleration distribution in each velocity interval and use them as the upper and lower bounds. Finally, establish two continuous velocity-acceleration dynamic constraint functions by fitting functions to the upper and lower bounds respectively.

[0062] (2.3) Based on the vehicle driving state (uniform speed, acceleration and deceleration), the floating car speed and acceleration within the constraint boundary are divided. By statistically analyzing the hidden state transition matrix from the previous time to the current time and the joint probability density function of the acceleration condition from the previous time to the current time, the basic parameters of the HMM (Hidden Markov Chain) model are obtained.

[0063] The formula for calculating the hidden state transition matrix from the previous time step to the current time step is as follows:

[0064]

[0065] in, It represents the current driving state, including: acceleration (acceleration greater than 1), deceleration (acceleration less than -1), and constant speed (acceleration greater than -1 and less than 1). It represents the driving state at the previous moment; through the hidden state transition matrix Utilizing the previous moment Determine the current driving status. .

[0066] The formula for calculating the conditional joint probability density function of acceleration from the previous time step to the current time step is as follows:

[0067]

[0068] In this formula, all parameters are limited to the hidden state at the current time. and the interval to which the velocity belonged at the previous moment. Within the scope of statistics, The variance of the acceleration distribution at the previous moment; The variance of the acceleration distribution at the current moment; It is the acceleration of the previous moment. and current acceleration The correlation coefficient; It is the current moment. The distribution mean; It was the previous moment The mean of the distribution.

[0069] (2.4) If the operating data at a given moment is determined to be abnormal, the velocity and acceleration from the previous moment are input into the HMM model. Combining the velocity and acceleration from the previous moment, the hidden state of the previous moment is inferred. Subsequently, based on the hidden state transition matrix, the state estimate for the current moment (e.g., acceleration / uniform speed / deceleration) is predicted. Finally, under the current predicted state, based on the current velocity range and the conditional constraints of the acceleration from the previous moment, random sampling is performed using the acceleration conditional probability density function to generate an estimated acceleration value for the current moment that conforms to traffic dynamics constraints. .

[0070] Furthermore, in step S3, the candidate trajectory set is determined by the collected upstream and downstream fixed sensor data and the dynamic bidirectional car-following equation based on traffic wave-DTW equivalence. The specific steps are as follows:

[0071] (3.1) Using the upstream floating car speed sequence after working condition correction and downstream floating car speed sequence Substitute the Dynamic Time Warping (DTW) algorithm into the algorithm to generate matching pairs with the goal of making the velocities between two points as close as possible.

[0072]

[0073] Among them, the optimal path It is a set consisting of a series of matching points; W It is the set of all feasible paths in the spatiotemporal search space that satisfy the monotonicity and continuity constraints; Indicates the first feasible path k There are 1 matching pairs, and their values ​​range from 1 to 2. ; Indicates that the upstream floating car is in the first k Sampling position during the second match; Indicates that the downstream floating car is in the first k Sampling position during the second match; The constraints mean that the matching path can only move in one direction, either diagonally, horizontally, or vertically, within the spatiotemporal search space.

[0074] It should be noted that fixed sensors (such as coils) are usually deployed in the road cross-section to record the number of vehicles passing through the sensor and instantaneous operating conditions (speed and acceleration). In this embodiment, the upstream and downstream sensor interval is set to 500m to collect the instantaneous traffic status of the upstream and downstream of the road.

[0075] Dynamic Time Warping (DTW) is an algorithm that performs nonlinear morphological calibration on two sequences by identifying an optimal matching path that minimizes the cumulative distance between the aligned sequences within a spatiotemporal search space. Let the upstream and downstream sequences be X = (x1, ..., x...). n ) and Y=(y1,…,y m The algorithm determines d(x) that minimizes the distance metric. i ,y j Alignment is achieved by using an alignment path. In the physical context of this embodiment, the distance metric is strictly defined as the difference between instantaneous velocity vectors, aiming to drive the matching of spatiotemporal coordinates through the consistency of the velocity field.

[0076] (3.2) Move the vehicle n The spatiotemporal Euclidean distance between the upstream and downstream fixed sensor observation points and the matching pairs generated by the upstream and downstream floating car trajectories is calculated sequentially, and the matching pair with the smallest spatiotemporal Euclidean distance is selected as the best matching pair for the fixed sensor observation points of the vehicle.

[0077] (3.3) Based on the spatiotemporal information of each vehicle arrival section recorded by upstream and downstream fixed sensors and the spatiotemporal information of the two trajectory points of their corresponding best matching pair, the vehicle is obtained. n The ratio of time delay to idle delay is calculated using the following formula:

[0078]

[0079] in, and These are based on upstream fixed sensors and vehicles. n The time delay ratio and the idle delay ratio; and Vehicles recorded by upstream fixed sensors n Arrival time and space information; and It is a vehicle n The best matching pair of upstream fixed points corresponds to the spatiotemporal information of the upstream floating car trajectory points; and It is the spatiotemporal information of the downstream floating car trajectory points corresponding to the best matching pair.

[0080]

[0081] in, and These are based on downstream fixed sensors and vehicles. n The time delay ratio and the idle delay ratio; and These are vehicles recorded by downstream fixed sensors. n Arrival time and space information; and They are vehicles n The best matching pair of the downstream fixed point corresponds to the spatiotemporal information of the upstream floating car trajectory point; and These are the spatiotemporal information of the downstream floating car trajectory points corresponding to the best matching pair.

[0082] (3.4) Combining the vehicle arrival data recorded by upstream and downstream fixed sensors with the floating car trajectory data, the candidate trajectories are calculated by substituting them into the dynamic bidirectional car-following equation based on traffic wave-DTW equivalence. This generates a candidate trajectory set for each vehicle based on upstream and downstream fixed sensors. The calculation formula is as follows:

[0083]

[0084] in, and It is the first k The spatiotemporal information of the upstream floating car for each matching pair; and It is the first k The spatiotemporal information of downstream floating cars for each matching pair; and It is based on data from upstream fixed sensors, vehicle n The corresponding number kThe time and space delays required for the upstream floating car of each matching pair to translate, and the current vehicle n In time The position is equivalent to the previous car n -1 is subtracted at time t. .

[0085]

[0086] in, and It is the first k The spatiotemporal information of the upstream floating car for each matching pair; and It is the first k The spatiotemporal information of downstream floating cars for each matching pair; and These are based on data from downstream fixed sensors, and the vehicle... n The corresponding number k The time and space delays required for the upstream floating car of each matching pair to translate, and the current vehicle n In time The position is equivalent to the previous car n -1 is subtracted at time t. .

[0087] Furthermore, in step S4, the fused trajectory of vehicles that have not changed lanes is determined by a dynamic programming algorithm based on a set of candidate vehicle trajectories generated from upstream and downstream fixed sensors and nonlinear time-varying weights. The specific steps are as follows:

[0088] (4.1) The fused trajectory of vehicles that have not changed lanes is derived from the candidate trajectory set of vehicles from upstream and downstream fixed sensors. The fused trajectory and the calculation formula for the operating conditions are as follows:

[0089]

[0090] in, and These are weighted functions for vehicle candidate trajectories generated from upstream and downstream fixed sensors, and their values ​​change over time. and These are the spatial positions of the two candidate vehicle trajectories generated based on upstream and downstream fixed sensors at the corresponding times. The spatial location of the fused trajectory at the corresponding moment; and The instantaneous velocity and acceleration are given in the spatiotemporal coordinates corresponding to the fused trajectory; T represents the smallest time unit, which is 0.04s.

[0091] (4.2) The weighting function of the nonlinear time-varying weights is determined by the dynamic programming algorithm, and the calculation formula is as follows:

[0092]

[0093] in, This represents the reference velocity obtained from the corresponding coordinate point of the regional macroscopic velocity field reference. Indicates the steepness of the transition, with a value between 0 and 1; Representing the transition midpoint, this is achieved by traversing the vehicles in a dynamic programming algorithm. n The common time of the two candidate trajectories is determined; The instantaneous velocity of the spatiotemporal coordinate point corresponding to the fused trajectory can be calculated by the weighting function.

[0094] Furthermore, in step S5, the fused trajectory of the lane-changing vehicle is determined by the vehicle candidate set generated based on the upstream and downstream fixed sensors and the vehicle candidate set generated based on the optimal lane-changing position. The specific steps are as follows:

[0095] (5.1) For vehicles that engage in lane changing behavior, a multi-objective optimization algorithm is used to calculate the optimal lane change position. The calculation formula is as follows:

[0096]

[0097] in, and These represent the estimated speed reference surfaces for lane a and lane b (the two lanes where the switching action occurred); and These are candidate trajectories collected by a single sensor in different lanes; This represents the adjustment range of the two candidate trajectories, which is obtained from the absolute half difference of the two trajectories; and These represent the lead distances of the vehicle in front and the vehicle behind, respectively. This indicates the minimum safe distance, which is specified as 5 m in this embodiment; The speed difference threshold is specified as 1.8 km / h in this embodiment; The distance difference threshold is specified as 0.5m in this embodiment to prevent zero from appearing in the numerator or denominator.

[0098] (5.2) Based on the optimal lane change point determined by the optimization algorithm, a new candidate trajectory is generated on the original lane and the target lane respectively. The method for generating the candidate trajectory is the same as in step 3.

[0099] (5.3) The new candidate trajectory generated based on the lane change point in the original lane is fused with the single candidate trajectory inferred by the fixed sensor in the same lane in a time-varying manner (the method is the same as step S4) to determine the optimal weight, thereby determining the fused trajectory of the lane-changing vehicle in the original lane before the lane change; the new candidate trajectory generated based on the lane change point in the lane change lane is fused with the single candidate trajectory inferred by the fixed sensor in the same lane in a time-varying manner (the method is the same as step S4) to determine the optimal weight, thereby determining the fused trajectory of the lane-changing vehicle in the original lane after the lane change.

[0100] Furthermore, in step S6, by merging two types of non-floating vehicles with different behaviors, and after correcting the abnormal trajectory conditions using the method in step S2, the complete reconstructed trajectory is output.

[0101] Example

[0102] Taking data collected by two fixed sensors in a two-lane area of ​​a one-way road and a mobile sensor with a penetration rate of 5% traveling on a designated road area as an example, based on the macro-micro fusion trajectory reconstruction method for emission estimation proposed in this invention, a reconstructed trajectory suitable for high-precision spatiotemporal emission estimation is generated. The specific implementation steps are as follows:

[0103] S1. Using the velocity and acceleration in time and space corresponding to the sparse trajectory data collected by two types of road sensors as input, a regional macroscopic velocity field reference is generated. The data collection information for fixed sensors and floating car data is shown in Tables 1 and 2:

[0104] Table 1. Data collected by fixed sensors (partial list)

[0105]

[0106] Table 2. Data collected by floating car sensors (partial list)

[0107]

[0108] (1) The data is weighted and interpolated using an adaptive dual filter based on free flow and congested flow. The filter uses an anisotropic kernel with the principal axis skewed. The interpolation speed obtained by fusing the dual kernels is obtained by analyzing the congestion level at spatiotemporal points. The calculation formula is as follows:

[0109]

[0110] in, The velocity at that spatiotemporal point is obtained by filter interpolation based on congested flow; The velocity at that spatiotemporal point is obtained based on free-flow filter interpolation; This is the weight adjustment function, and its calculation formula is as follows:

[0111]

[0112] in, For the spatiotemporal point where the actual velocity value is obtained from the sensor; This represents the speed threshold between free flow and congested flow, which is specified as 50 km / h in this embodiment; Indicates in The transition width obtained near the value is specified as 10 km / h in this embodiment; It is a free-flow based filter; It is a filter based on congested flow, and the calculation formula is as follows:

[0113]

[0114] in, In this embodiment, it is set to 50; In this embodiment, it is set to 36.

[0115] (2) Apply moving average smoothing to obtain the final regional macroscopic velocity field reference. The calculation formula is as follows:

[0116]

[0117] in, The unit spatial interval is set to 1m in this embodiment; The unit time interval is set to 0.4s in this embodiment; The window size is set to 70 in this embodiment; and It is a two-dimensional spatiotemporal index for roads.

[0118] The generated regional macroscopic velocity field reference is as follows Figure 3 As shown in the figure, this graph reflects the velocity interpolation smoothing result of the spatiotemporal coordinates output based on the velocity information input from two types of datasets. Blue represents a higher velocity and a more free and unobstructed spatiotemporal traffic state; red represents a lower velocity and a more congested spatiotemporal traffic state. The coordinates of the velocity field reference and their corresponding velocity information are shown in Table 3.

[0119] Table 3 Summary of Regional Macroscopic Velocity Field Reference Information (Partial List)

[0120]

[0121] S2. Correct abnormal operating conditions of floating cars, specifically including the following steps:

[0122] (1) First, the vehicle's acceleration data is obtained by differential processing of the instantaneous acceleration data of the floating car; then, at 0.2 m / s 2 The acceleration is divided into discrete intervals for quantization, and the corresponding jerk distribution in each interval is statistically analyzed. Then, the 97.5% quantile and 1% quantile of the jerk distribution in each acceleration interval are extracted and used as the upper and lower bounds. Finally, a continuous acceleration-jerk dynamic constraint function is established by fitting a function to these discrete boundary points.

[0123] (2) Divide the velocity discrete intervals with a quantization interval of 0.5 m / s, and statistically analyze the corresponding acceleration distribution in each interval; then, extract the 97.5% quantile and 1% quantile of the acceleration distribution in each velocity interval, and use them as the upper and lower bounds; finally, establish a continuous velocity-acceleration dynamic constraint function by fitting these discrete boundary points.

[0124] (3) Based on the vehicle's driving state (uniform speed, acceleration, and deceleration), the floating car's speed and acceleration within the constraint boundary are divided. By statistically analyzing the hidden state transition matrix from the previous time step to the current time step and the observation matrix of the acceleration from the previous time step to the current time step, i.e., the conditional joint probability density function of the acceleration at the current time step, the basic parameters of the HMM model are obtained. The calculation formula is as follows:

[0125]

[0126] in, It represents the current driving state, including: acceleration (acceleration greater than 1), deceleration (acceleration less than -1), and constant speed (acceleration greater than -1 and less than 1). It represents the driving state at the previous moment; through the hidden state transition matrix Utilizing the previous moment Determine the current driving status. .

[0127]

[0128] In this formula, all parameters are limited to the hidden state at the current time. and the interval to which the velocity belonged at the previous moment. Within the scope of statistics, The variance of the acceleration distribution at the previous moment; The variance of the acceleration distribution at the current moment; It is the acceleration of the previous moment. and current acceleration The correlation coefficient; It is the current moment. The distribution mean; It was the previous moment The mean of the distribution.

[0129] (4) Input the velocity and acceleration of the previous moment into the HMM model, and infer the hidden state of the previous moment by combining the velocity and acceleration of the previous moment. Then, based on the hidden state transition matrix, predict the state estimate of the current moment (e.g., acceleration / uniform speed / deceleration). Finally, under the current predicted state, based on the current velocity range and the conditional constraints of the acceleration of the previous moment, use the acceleration conditional probability density function to perform random sampling to generate the current moment acceleration estimate that meets the traffic dynamics constraints. .

[0130] The observation matrix of the trained HMM model is as follows Figure 4 As shown in Table 4, the original operating data of the floating car and the corresponding corrected operating data are presented.

[0131] Table 4 Original operating data and corrected data of the floating car (partial list)

[0132]

[0133] S3. Using the collected dual-sensor data and the dynamic bidirectional car-following equation based on traffic wave-DTW simulcasting, a candidate trajectory set is generated. The specific steps are as follows:

[0134] (1) Using the upstream floating car speed sequence after working condition correction and downstream floating car speed sequence Substitute the DTW (Data Transfer Method) into the target, and generate matching pairs with the closest possible speed between the two points.

[0135]

[0136] Among them, the optimal path It is a set consisting of a series of matching points; W It is the set of all feasible paths in the spatiotemporal search space that satisfy the monotonicity and continuity constraints; Indicates the first feasible path k There are 1 matching pairs, and their values ​​range from 1 to 2. ; Indicates that the upstream floating car is in the first k Sampling position during the second match; Indicates that the downstream floating car is in the first k Sampling position during the second match; The constraints mean that the matching path can only move in one direction, either diagonally, horizontally, or vertically, within the spatiotemporal search space.

[0137] The generated trajectory point matching pairs are as follows Figure 5 As shown, PV1 represents the upstream floating car, PV2 represents the downstream floating car, the red line represents the matching pair generated by the upstream and downstream floating cars according to the DTW algorithm based on the approximation of speed, the green trajectory represents the floating car trajectory, the blue dots represent the vehicle trajectory points detected by the upstream and downstream fixed detectors, and the asterisks represent the vehicle trajectory points recorded by the downstream sensor.

[0138] (2) Move the vehicle k The spatiotemporal Euclidean distance between the fixed sensor observation point and the matching pairs generated by the upstream and downstream floating car trajectories is calculated sequentially, and the matching pair with the smallest spatiotemporal Euclidean distance is selected as the best matching pair for the fixed sensor observation point of the vehicle.

[0139] (3) Based on the spatiotemporal information of each vehicle arrival section recorded by upstream and downstream fixed sensors and the spatiotemporal information of the two trajectory points of their corresponding best matching pair, the vehicle is obtained. k The ratio of time delay to idle delay is calculated using the following formula:

[0140]

[0141] in, and Based on upstream fixed sensors, vehicles n The time delay ratio and the idle delay ratio; and Vehicles recorded by upstream fixed sensors n Arrival time and space information; and It is a vehicle n The best matching pair of upstream fixed points corresponds to the spatiotemporal information of the upstream floating car trajectory points; and It is the spatiotemporal information of the downstream floating car trajectory points corresponding to the best matching pair.

[0142]

[0143] in, and Based on downstream fixed sensors, vehicles n The time delay ratio and the idle delay ratio; and Vehicles recorded by downstream fixed sensors n Arrival time and space information; and It is a vehicle n The best matching pair of the downstream fixed point corresponds to the spatiotemporal information of the upstream floating car trajectory point; and It is the spatiotemporal information of the downstream floating car trajectory points corresponding to the best matching pair.

[0144] The generated vehicle k is based on the best matching pair obtained from the sensors and the ratio of time delay to idle delay, as follows: Figure 6 As shown.

[0145] (4) Combining the vehicle arrival data recorded by upstream and downstream fixed sensors with the floating car trajectory data, the candidate trajectories are calculated by substituting them into the dynamic bidirectional car-following equation based on traffic wave-DTW equivalence. This generates a candidate trajectory set for each vehicle based on upstream and downstream fixed sensors. The calculation formula is as follows:

[0146]

[0147] in, and It is the first k The spatiotemporal information of the upstream floating car for each matching pair; and It is the first k The spatiotemporal information of downstream floating cars for each matching pair; and It is based on data from upstream fixed sensors, vehicle n The corresponding number k The time and space delays required for the upstream floating car of each matching pair to translate, and the current vehicle n In time The position is equivalent to the previous car n -1 is subtracted at time t. .

[0148]

[0149] in, and It is the first k The spatiotemporal information of the upstream floating car for each matching pair; and It is the first k The spatiotemporal information of downstream floating cars for each matching pair; and It is based on data from downstream fixed sensors, vehicle n The corresponding number k The time and space delays required for the upstream floating car of each matching pair to translate, and the current vehicle n In time The position is equivalent to the previous car n -1 is subtracted at time t. .

[0150] The generated candidate trajectory data is shown in Table 5. The corresponding spatiotemporal two-dimensional paths of the candidate trajectories are shown in Table 5. Figure 7 As shown.

[0151] Table 5. Candidate vehicle trajectories generated based on upstream and downstream fixed sensors (partial list)

[0152]

[0153] S4. The fused trajectory of vehicles that have not changed lanes is determined by a dynamic programming algorithm based on a set of candidate vehicle trajectories generated from upstream and downstream fixed sensors and nonlinear time-varying weights. The specific steps are as follows:

[0154] (1) The fused trajectory of vehicles that have not changed lanes is derived from the candidate trajectory set of vehicles from upstream and downstream fixed sensors. The fused trajectory and its calculation formula are as follows:

[0155]

[0156] in, and These are weighted functions for vehicle candidate trajectories generated from upstream and downstream fixed sensors, and their values ​​change over time. and These are the spatial positions of the two candidate vehicle trajectories generated based on upstream and downstream fixed sensors at the corresponding times. The spatial location of the fused trajectory at the corresponding moment; and The instantaneous velocity and acceleration are given in the spatiotemporal coordinates corresponding to the fused trajectory; T represents the smallest time unit, which is 0.04s.

[0157] (2) The weighting function of the nonlinear time-varying weights is determined by the dynamic programming algorithm, and the calculation formula is as follows:

[0158]

[0159] in, This represents the reference velocity obtained from the corresponding coordinate point of the regional macroscopic velocity field reference. Indicates the steepness of the transition, with a value between 0 and 1; Representing the transition midpoint, this is achieved by traversing the vehicles in a dynamic programming algorithm. n The common time of the two candidate trajectories is determined; The instantaneous velocity of the spatiotemporal coordinate point corresponding to the fused trajectory can be calculated by the weighting function.

[0160] The fused trajectory data generated by candidate trajectory fusion is shown in Table 6. The corresponding candidate trajectory fusion process is as follows: Figure 8 As shown.

[0161] Table 6. Information on fused trajectory data (partial list)

[0162]

[0163] S5. The fused trajectory of lane-changing vehicles is determined by a vehicle candidate set generated based on upstream and downstream fixed sensors and a vehicle candidate set generated based on the optimal lane-changing position. The specific steps are as follows:

[0164] (1) For vehicles that engage in lane changing behavior, a multi-objective optimization algorithm is used to calculate the optimal lane change position. The calculation formula is as follows:

[0165]

[0166] in, and These represent the estimated speed reference surfaces for lane a and lane b, respectively. and These are candidate trajectories collected by a single sensor in different lanes; This represents the adjustment range that two candidate trajectories can be varied. This range is obtained by the absolute half difference between the two trajectories, and all spatiotemporal points within this range are considered as the candidate set of change points. and These represent the lead distances of the vehicle in front and the vehicle behind, respectively. This indicates the minimum safe distance, which is specified as 5 m in this invention; The speed difference threshold is specified as 1.8 km / h in this embodiment; The distance difference threshold is specified as 0.5m in this embodiment to prevent zero from appearing in the numerator or denominator.

[0167] The generated lane change point candidate region is as follows Figure 9 As shown, PV1 represents the upstream floating car of lane 1, PV2 represents the downstream floating car of lane 1, PV3 represents the upstream floating car of lane 2, and PV4 represents the downstream floating car of lane 2. The red trajectory represents the upstream and downstream floating car trajectories of lane 1 and lane 2, the blue dashed trajectory represents the fused trajectory of lane 1 and lane 2 generated according to the present invention, the green strip area represents the candidate area for lane change point decision, the blue dots represent the vehicle trajectory points passed by both upstream and downstream fixed sensors, and the red dots represent the vehicle trajectory points passed by only a single fixed sensor.

[0168] (2) Based on the optimal lane change point determined by the optimization algorithm, a new candidate trajectory is generated on the original lane and the target lane respectively using the method in step 3.

[0169] (3) The new candidate trajectory generated based on the lane change point in the original lane is fused with the single candidate trajectory inferred by the fixed sensor in the same lane in step S4 to determine the optimal weight, thereby determining the fused trajectory of the lane-changing vehicle in the original lane before the lane change; the new candidate trajectory generated based on the lane change point in the lane change lane is fused with the single candidate trajectory inferred by the fixed sensor in the same lane in step S4 to determine the optimal weight, thereby determining the fused trajectory of the lane-changing vehicle in the original lane after the lane change.

[0170] The merged trajectory data of lane-changing vehicles is shown in Table 7. The determination of the corresponding lane-changing points and the fusion process are as follows: Figure 10 and Figure 11 As shown.

[0171] Table 7. Merged trajectory data of lane-changing vehicles (partial list)

[0172]

[0173] S6. By merging two types of non-floating vehicles with different behaviors, the trajectory working conditions are corrected using the method in S2, and the complete reconstructed trajectory is output.

[0174] The complete trajectory data is shown in Table 8, and the corresponding complete trajectory diagram is shown below. Figure 12 As shown, the trajectory reconstructed in this model is mapped by speed. The yellower the trajectory segment, the greater the speed, and the bluer the trajectory segment, the smaller the speed. The red trajectory represents the floating car trajectory, and the purple-blue dots represent the vehicle trajectory points that pass through the upstream and downstream fixed sensors (without distinguishing lane changes).

[0175] Table 8. Complete trajectory data information (partial list only)

[0176]

[0177] In this embodiment, the trajectory used to estimate the crossover ratio (CBR) of emissions and actual road emissions is 85.65%, the F1-score reaches 92.27%, and the recall rate is 96.35%. Compared with traditional methods, it improves the CBR by 26.85%, the F1-score by 18.06%, and the recall rate by 21.47%. It breaks through the technical barrier of scale coupling in traditional models. Through a single and robust integration mechanism, it deeply couples vehicle dynamics, spatiotemporal dynamic wave propagation characteristics, and multi-lane heterogeneous behavior, achieving a comprehensive leap in the overall performance of emission hotspot identification.

[0178] The above embodiments are intended to illustrate the present invention in detail; however, the present invention is not limited to this single example. Any modifications, alterations, substitutions, combinations, simplifications, etc., made to the embodiments without departing from the spirit and principle of the present invention should be considered as equivalent substitutions and included within the protection scope of the present invention.

Claims

1. A macro-micro fusion trajectory reconstruction method applied to emissions estimation, characterized in that, Includes the following steps: S1. Generate a regional macroscopic velocity field reference; S2. Correct abnormal operating conditions of floating cars; S3. Determine the candidate trajectory set for non-floating vehicles; S4. Determine the fused trajectory of vehicles that have not changed lanes among non-floating vehicles based on candidate trajectories; S5. Determine the fused trajectory of lane-changing vehicles among non-floating vehicles based on candidate trajectories; S6. Integrate the two types of trajectories, verify the operating conditions, and generate a complete vehicle trajectory; The aforementioned regional macroscopic velocity field reference is generated by sparse trajectory data collected from road sensors, and the specific steps are as follows: (1.1) Collect sparse trajectory data using road sensors; then use an adaptive dual filter based on free flow and congested flow to perform weighted interpolation on the trajectory data to obtain the interpolated velocity. ; Specifically, the adaptive dual filter employs anisotropic kernels that skew the principal axis, and by analyzing the congestion level at spatiotemporal points, fuses the interpolation speeds obtained from the dual kernels. The calculation formula is as follows: in, The velocity at that spatiotemporal point is obtained by filter interpolation based on congested flow; The velocity at that spatiotemporal point is obtained based on free-flow filter interpolation; It is a weight adjustment function; the calculation formula is as follows: in, For the spatiotemporal point where the actual velocity value is obtained from the sensor; This represents the velocity threshold between free flow and congested flow; N and M are the maximum numbers obtained by dividing the total spatial distance and total time in the dataset by the smallest unit, respectively; Distance coordinates for a specific number; A time coordinate with a specific number; Indicates in The transition width obtained near the value; It is a free-flow based filter; It is a filter based on congested flow, and the calculation formula is as follows: Among them, spatial interpolation width Half the height of the anisotropic kernel in the filter; time interpolation width It is half the length of the anisotropic kernel in the filter; (1.2) Apply moving average smoothing to the interpolated velocity to obtain the final regional macroscopic velocity field reference. The calculation formula is as follows: in, For unit spatial interval, For unit time intervals, For window size, and A two-dimensional spatiotemporal index for roads; The specific steps of step S2 are as follows: (2.1) First, the vehicle's jerk data is obtained by differential processing of the floating car's instantaneous acceleration data; then, at a speed of 0.2 m / s², the jerk data is... 2 The acceleration discrete interval is divided into intervals for quantization, and the corresponding jerk distribution in each interval is statistically analyzed. Then, the 97.5% quantile and 1% quantile of the jerk distribution in each acceleration interval are extracted and used as the upper and lower bounds. Finally, two continuous acceleration-jerk dynamic constraint functions are established by fitting functions to the upper and lower bounds respectively. (2.2) Divide the velocity discrete interval with 0.5 m / s as the quantization interval, and statistically analyze the corresponding acceleration distribution in each interval; then, extract the 97.5% quantile and 1% quantile of the acceleration distribution in each velocity interval, and use them as the upper and lower bounds; finally, establish two continuous velocity-acceleration dynamic constraint functions by fitting functions to the upper and lower bounds respectively. (2.3) Based on the vehicle driving state, the floating car speed and acceleration within the constraint boundary are divided. By statistically analyzing the hidden state transition matrix from the previous time to the current time and the acceleration conditional joint probability density function from the previous time to the current time, the basic parameters of the hidden Markov chain HMM model are obtained. The formula for calculating the hidden state transition matrix from the previous time step to the current time step is as follows: in, It represents the current driving state, including: acceleration, deceleration, and constant speed. It represents the driving state at the previous moment; through the hidden state transition matrix Utilizing the previous moment Determine the current driving status. ; The formula for calculating the conditional joint probability density function of acceleration from the previous time step to the current time step is as follows: In this formula, all parameters are limited to the hidden state at the current time. and the interval to which the velocity belonged at the previous moment. Within the scope of statistics, The variance of the acceleration distribution at the previous moment; The variance of the acceleration distribution at the current moment; It is the acceleration of the previous moment. and current acceleration The correlation coefficient; It is the current moment. The distribution mean; It was the previous moment The distribution mean; (2.4) If the operating condition data at this moment is determined to be abnormal, the velocity and acceleration of the previous moment are input into the HMM model. Combined with the velocity and acceleration of the previous moment, the hidden state of the previous moment is inferred. Subsequently, based on the hidden state transition matrix, the state estimate of the current moment is predicted. Finally, under the current predicted state, based on the current velocity range and the conditional constraints of the acceleration of the previous moment, random sampling is performed using the acceleration conditional probability density function to generate the current moment acceleration estimate that conforms to traffic dynamics constraints. .

2. The macro-micro fusion trajectory reconstruction method for emission estimation as described in claim 1, characterized in that, Data input includes the following two categories: Mobile sensor floating car data: includes vehicle identification, on-road spatial coordinates at each unit timestamp, instantaneous velocity, and instantaneous acceleration; Fixed sensor data includes vehicle identification, arrival time of the vehicle across a fixed road section, instantaneous speed, and instantaneous acceleration.

3. The macro-micro fusion trajectory reconstruction method for emission estimation as described in claim 1, characterized in that, The specific steps of step S3 are as follows: (3.1) Using the upstream floating car speed sequence after working condition correction and downstream floating car speed sequence Substituting the dynamic time warping algorithm, with the goal of minimizing the velocities between two points, we generate matching pairs: Among them, the optimal path It is a set consisting of a series of matching points; W It is the set of all feasible paths in the spatiotemporal search space that satisfy the monotonicity and continuity constraints; Indicates the first feasible path k There are 1 matching pairs, and their values ​​range from 1 to 2. ; Indicates that the upstream floating car is in the first k Sampling position during the second match; Indicates that the downstream floating car is in the first k Sampling position during the second match; The constraints mean that the matching path can only move in one direction, either diagonally, horizontally, or vertically, within the spatiotemporal search space. The Dynamic Time Warping (DTW) algorithm achieves nonlinear morphological calibration of two sequences by identifying an optimal matching path that minimizes the cumulative distance between aligned sequences within a spatiotemporal search space. Let the upstream and downstream sequences be X = (x1, ..., x...). n ) and Y=(y1,…,y m The algorithm determines d(x) that minimizes the distance metric. i ,y j Alignment is achieved by using the alignment path; (3.2) Move the vehicle n The spatiotemporal Euclidean distance between the upstream and downstream fixed sensor observation points and the matching pairs generated by the upstream and downstream floating car trajectories is calculated sequentially, and the matching pair with the smallest spatiotemporal Euclidean distance is selected as the best matching pair for the fixed sensor observation points of the vehicle. (3.3) Based on the spatiotemporal information of each vehicle arrival section recorded by upstream and downstream fixed sensors and the spatiotemporal information of the two trajectory points of their corresponding best matching pair, the vehicle is obtained. n The ratio of time delay to idle delay is calculated using the following formula: in, and These are based on upstream fixed sensors and vehicles. n The time delay ratio and the idle delay ratio; and Vehicles recorded by upstream fixed sensors n Arrival time and space information; and It is a vehicle n The best matching pair of upstream fixed points corresponds to the spatiotemporal information of the upstream floating car trajectory points; and It is the spatiotemporal information of the downstream floating car trajectory points corresponding to the best matching pair; in, and These are based on downstream fixed sensors and vehicles. n The time delay ratio and the idle delay ratio; and These are vehicles recorded by downstream fixed sensors. n Arrival time and space information; and They are vehicles n The best matching pair of the downstream fixed point corresponds to the spatiotemporal information of the upstream floating car trajectory point; and These are the spatiotemporal information of the downstream floating car trajectory points corresponding to the best matching pair; (3.4) Combining the vehicle arrival data recorded by upstream and downstream fixed sensors with the floating car trajectory data, the candidate trajectories are calculated by substituting them into the dynamic bidirectional car-following equation based on traffic wave-DTW equivalence. This generates a candidate trajectory set for each vehicle based on upstream and downstream fixed sensors. The calculation formula is as follows: in, and It is the first k The spatiotemporal information of the upstream floating car for each matching pair; and It is the first k The spatiotemporal information of downstream floating cars for each matching pair; and It is based on data from upstream fixed sensors, vehicle n The corresponding number k The time and space delays required for the upstream floating car of each matching pair to translate, and the current vehicle n In time The position is equivalent to the previous car n -1 is subtracted at time t. ; in, and It is the first k The spatiotemporal information of the upstream floating car for each matching pair; and It is the first k The spatiotemporal information of downstream floating cars for each matching pair; and These are based on data from downstream fixed sensors, and the vehicle... n The corresponding number k The time and space delays required for the upstream floating car of each matching pair to translate, and the current vehicle n In time The position is equivalent to the previous car n -1 is subtracted at time t. .

4. The macro-micro fusion trajectory reconstruction method for emission estimation as described in claim 1, characterized in that, In step S4, the fused trajectory of vehicles that have not changed lanes is determined by a dynamic programming algorithm based on a set of candidate vehicle trajectories generated by upstream and downstream fixed sensors and nonlinear time-varying weights. The specific steps are as follows: (4.1) The fused trajectory of vehicles that have not changed lanes is derived from the candidate trajectory set of vehicles from upstream and downstream fixed sensors. The fused trajectory and the calculation formula for the operating conditions are as follows: in, and These are weighted functions for vehicle candidate trajectories generated from upstream and downstream fixed sensors, and their values ​​change over time. and These are the spatial positions of the two candidate vehicle trajectories generated based on upstream and downstream fixed sensors at the corresponding times. The spatial location of the fused trajectory at the corresponding moment; and The instantaneous velocity and acceleration in the spatiotemporal coordinates corresponding to the fused trajectory; T represents the smallest time unit; (4.2) The weighting function of the nonlinear time-varying weights is determined by the dynamic programming algorithm, and the calculation formula is as follows: in, This represents the reference velocity obtained from the corresponding coordinate point of the regional macroscopic velocity field reference. Indicates the steepness of the transition, with a value between 0 and 1; Representing the transition midpoint, this is achieved by traversing the vehicles in a dynamic programming algorithm. n The common time of the two candidate trajectories is determined; The instantaneous velocity of the spatiotemporal coordinate point corresponding to the fused trajectory can be calculated by the weighting function.

5. The macro-micro fusion trajectory reconstruction method for emission estimation as described in claim 1, characterized in that, In step S5, the fused trajectory of the lane-changing vehicle is determined by the vehicle candidate set generated based on the upstream and downstream fixed sensors and the vehicle candidate set generated based on the optimal lane-changing position. The specific steps are as follows: (5.1) For vehicles that engage in lane changing behavior, a multi-objective optimization algorithm is used to calculate the optimal lane change position. The calculation formula is as follows: in, and These represent the estimated speed reference surfaces for lane a and lane b, respectively. and These are candidate trajectories collected by a single sensor in different lanes; This represents the adjustment range of the two candidate trajectories, which is obtained from the absolute half difference of the two trajectories; and These represent the lead distances of the vehicle in front and the vehicle behind, respectively. Indicates the minimum safe distance; Indicates the speed difference threshold; This represents the distance difference threshold, designed to prevent zero from appearing in the numerator or denominator; (5.2) Based on the optimal lane change point determined by the optimization algorithm, a new candidate trajectory is generated on the original lane and the target lane respectively; (5.3) The new candidate trajectory generated based on the lane change point in the original lane is fused with the single candidate trajectory inferred by the fixed sensor in the same lane in a time-varying manner to determine the optimal weight, thereby determining the fused trajectory of the lane-changing vehicle in the original lane before the lane change; the new candidate trajectory generated based on the lane change point in the lane change lane is fused with the single candidate trajectory inferred by the fixed sensor in the same lane in a time-varying manner to determine the optimal weight, thereby determining the fused trajectory of the lane-changing vehicle in the original lane after the lane change.

6. The macro-micro fusion trajectory reconstruction method for emission estimation as described in claim 1, characterized in that, Step S6 merges the two types of non-floating vehicles with different behaviors, corrects the trajectory conditions using the method in S2, and outputs the complete reconstructed trajectory.

Images