A method and device for predicting multi-region short-time distribution positions of intercity network car-hailing

Abstract

This invention provides a method and apparatus for predicting the short-term, multi-regional distribution location of intercity ride-hailing vehicles, relating to the field of vehicle dispatching technology. The method includes: dividing the intercity ride-hailing route into several sub-regions based on acquired road network data of the two cities, and constructing intercity entrances and exits for the two cities; constructing initial vehicle reporting sets for each sub-region of the two cities based on acquired passenger order data; establishing a separate multi-task adaptive Kalman filter model for each vehicle; constructing a set of actual vehicle distribution, a set of predicted available vehicle distribution, and a set of vehicles scheduled to leave the city for the two cities; updating the set of vehicles scheduled to leave the city after constructing the set of vehicles en route; and re-reporting vehicles after they leave the city and resetting the corresponding Kalman filter model. This invention establishes a separate Kalman filter model for each vehicle, which can better capture the characteristics of each vehicle and improve the accuracy of the short-term, multi-regional distribution location of intercity ride-hailing vehicles.

Description

Technical Field

[0001] This invention relates to the field of vehicle dispatching technology, and more specifically, to a method and apparatus for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions. Background Technology

[0002] With the widespread adoption of ride-hailing services, intercity transportation demand has become highly dynamic and complex, especially during peak hours and holidays, when travel demand fluctuates significantly across different regions. Traditional static scheduling methods cannot meet this ever-changing demand, leading to increased passenger waiting times and higher driver empty-run rates, impacting service quality and operational efficiency. Intercity transportation demand is influenced by various factors such as weekday and weekend travel habits, large-scale events, and unforeseen incidents, exhibiting high spatiotemporal variability. Traditional scheduling methods, based on fixed empirical rules, lack the ability to dynamically respond to real-time data and struggle to cope with complex demand changes. Therefore, intercity ride-hailing services urgently need a location prediction method capable of real-time prediction and dynamic adjustment of vehicle distribution to adapt to the rapidly changing demand environment, thereby improving passenger service satisfaction, optimizing driver efficiency, and enhancing the overall operational efficiency of ride-hailing platforms.

[0003] Capacity changes are influenced by scheduling strategies. Predictions based on data-driven deep learning algorithms lack interpretability, cannot predict the location and status of specific vehicles, and cannot update capacity distribution in real time according to changes in scheduling strategies, resulting in low prediction accuracy. Existing patent document CN117593043B, "Method, Apparatus and Equipment for Estimating the Short-Term Future Time Distribution of Intercity Ride-Hailing Vehicles," is not applicable when vehicle speed data is unavailable, and does not address the situation of each vehicle in multiple areas within a city.

[0004] In view of this, the applicant hereby submits this application after studying the existing technology. Summary of the Invention

[0005] The present invention aims to provide a method and device for predicting the short-term distribution of intercity ride-hailing vehicles in multiple regions, in order to solve the problems that existing models cannot explain the prediction of transportation capacity, cannot predict the specific location and situation of each vehicle, and have low accuracy.

[0006] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution:

[0007] A method for predicting the short-term, multi-regional distribution of intercity ride-hailing vehicles includes:

[0008] S1, based on the real-time road network data of intercity ride-hailing services in cities A and B, divide the area into several sub-regions and construct the designated intercity entrance / exit e for city A. A and the intercity entrance / exit e of city B B ;

[0009] S2, based on the passenger order data of each sub-region of city A and city B, construct the initial vehicle registration set K for each sub-region of city A and city B respectively. A and K B ;

[0010] S3, based on the initial vehicle reporting set K A and K B A multi-task adaptive Kalman filter model is established for each vehicle. This model takes the vehicle's highest detour coefficient at the current moment, the number of passengers, and the previous state data as input, and outputs the vehicle's next predicted state data and predicted departure time.

[0011] S4, at each scheduling decision time t within the established platform scheduling decision time, based on the vehicle reporting set K A Construct the set of real vehicle distributions for all sub-regions of city A at time t. And the predicted distribution set of available vehicles in all sub-regions of city A from time t to t+n. n is the estimated step size; similarly, based on the vehicle reporting set K B Construct a set of real vehicle distribution data for all sub-regions of city B. and the set of predicted available vehicle distributions

[0012] S5, using each scheduling decision time t as a trigger event, iterates through the real vehicle distribution set of city A. The detour coefficient of the vehicle at time t is calculated, and the detour coefficients and passenger numbers of non-empty and not fully loaded vehicles within the city at time t are input into the vehicle's adaptive Kalman filter model, outputting the vehicle's predicted departure time from the city. Based on the predicted departure time from the city Construct a set of pre-departure vehicles for multiple areas in city A; similarly, construct a set of pre-departure vehicles for multiple areas in city B.

[0013] S6, where vehicle k from city A passes through intercity entrance / exit e at time t. A To trigger the event, iterate through the set of real vehicle distributions in city A. Calculate the set of arrival times for all passengers in vehicles that have already left the city. Construct the set of vehicles en route to city A According to the vehicles gathered along the way The arrival time and drop-off location of the last passenger are used to update the predicted available vehicle distribution set for city B. Similarly, consider vehicle k from city B passing through intercity entrance / exit e at time t. BTo trigger the event, construct a set of vehicles en route to city B. And update the predicted available vehicle distribution set for city A

[0014] S7, based on the real-time data of the actual departure times of vehicles that have already left the city. Update the data of the corresponding vehicle's adaptive Kalman filter model until the vehicle reports for duty again, then reset the vehicle's adaptive Kalman filter model.

[0015] The present invention also provides a device for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions, comprising:

[0016] The multi-region dispatch environment construction unit is used to divide the intercity ride-hailing routes in cities A and B into several sub-regions based on real-time acquired road network data, and to construct the intercity entrance / exit points e for designated city A. A and the intercity entrance / exit e of city B B ;

[0017] The multi-regional vehicle registration set construction unit is used to construct the initial vehicle registration set K for each sub-region of city A and city B based on the passenger order data obtained from each sub-region of city A and city B. A and K B ;

[0018] The multi-task model building unit is used to build upon the initial vehicle reporting set K. A and K B A multi-task adaptive Kalman filter model is established for each vehicle. This model takes the vehicle's highest detour coefficient at the current moment, the number of passengers, and the previous state data as input, and outputs the vehicle's next predicted state data and predicted departure time.

[0019] A multi-regional vehicle distribution set construction unit is used to obtain the actual vehicle distribution set and the predicted available vehicle distribution set. At each scheduling decision time t within the established platform scheduling decision time, based on the vehicle reporting set K... A Construct the set of real vehicle distributions for all sub-regions of city A at time t. And the predicted distribution set of available vehicles in all sub-regions of city A from time t to t+n. n is the estimated step size; similarly, based on the vehicle reporting set K B Construct a set of real vehicle distribution data for all sub-regions of city B. and the set of predicted available vehicle distributions

[0020] The multi-regional pre-departure vehicle set construction unit is used to obtain the multi-regional pre-departure vehicle set. It iterates through the real vehicle distribution set of city A, with each scheduling decision time t as the trigger event. The detour coefficient of the vehicle at time t is calculated, and the detour coefficients and passenger numbers of non-empty and not fully loaded vehicles within the city at time t are input into the vehicle's adaptive Kalman filter model, outputting the vehicle's predicted departure time from the city. Based on the predicted departure time from the city Construct a set of pre-departure vehicles for multiple areas in city A; similarly, construct a set of pre-departure vehicles for multiple areas in city B.

[0021] The vehicle set construction unit is used to obtain the vehicle set in transit, such as vehicle k in city A passing through intercity entrance / exit e at time t. A To trigger the event, iterate through the set of real vehicle distributions in city A. Calculate the set of arrival times for all passengers in vehicles that have already left the city. Construct the set of vehicles en route to city A According to the vehicles gathered along the way The arrival time and drop-off location of the last passenger are used to update the predicted available vehicle distribution set for city B. Similarly, consider vehicle k from city B passing through intercity entrance / exit e at time t. B To trigger the event, construct a set of vehicles en route to city B. And update the predicted available vehicle distribution set for city A

[0022] The model update unit is used to update the model based on the actual departure times of vehicles that have already left the city, which are obtained in real time. Update the data of the corresponding vehicle's adaptive Kalman filter model until the vehicle reports for duty again, then reset the vehicle's adaptive Kalman filter model.

[0023] The present invention also provides a device for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions, including a processor and a memory. The memory stores a computer program that can be executed by the processor to realize the method for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions as described above.

[0024] The present invention also provides a computer-readable storage medium storing computer-readable instructions, which, when executed by a processor of the device on which the computer-readable storage medium is located, implement the method described above for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions.

[0025] In summary, compared with the prior art, the present invention has the following beneficial effects:

[0026] (1) The present invention establishes a Kalman filter model for each vehicle separately, and the data between vehicles do not interfere with each other and are updated independently, which can better capture the characteristics of each vehicle.

[0027] (2) The multi-task adaptive Kalman filter model constructed in this invention, in addition to using the previous state data as model input, also inputs the current number of passengers and the highest detour coefficient of the vehicle into the Kalman filter model to obtain the predicted departure time. The closer the highest detour coefficient and the number of passengers are to the constraint value, the more likely the vehicle is to leave the city directly, making the vehicle distribution prediction interpretable and making the vehicle "traceable".

[0028] (3) This invention can predict the location and status of specific vehicles in multiple areas without obtaining the vehicle's speed, and update vehicle information in real time. It can effectively improve prediction accuracy, adapt to rapidly changing demand environment and scheduling strategies, optimize driver work efficiency, and improve the overall operational efficiency of ride-hailing platforms.

[0029] (4) The present invention divides the city into several sub-regions, and can obtain the vehicle situation of a specific area of ​​a city in real time, reducing data transmission and improving operating efficiency. Attached Figure Description

[0030] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0031] Figure 1 This is a schematic diagram of a method for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions, provided in Embodiment 1 of the present invention.

[0032] Figure 2 This is a schematic diagram of the multi-task adaptive Kalman filter model provided in Embodiment 1 of the present invention.

[0033] Figure 3 This is a schematic diagram of vehicle trajectory prediction provided in Embodiment 1 of the present invention.

[0034] Figure 4 This is a real-time prediction process for the short-term distribution of intercity ride-hailing vehicles in multiple regions, provided in Embodiment 1 of the present invention.

[0035] Figure 5 This is a graph showing the predicted number of available vehicles in a certain area of ​​city B, as provided in Embodiment 1 of the present invention.

[0036] Figure 6This is a graph showing the predicted number of available seats in a certain area of ​​city B, as provided in Embodiment 1 of the present invention.

[0037] Figure 7 This is a schematic diagram of a device for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions, provided in Embodiment 2 of the present invention.

[0038] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0040] Example 1

[0041] Embodiment 1 of the present invention provides a method for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions. This method can be implemented by a prediction device for the short-term distribution location of intercity ride-hailing vehicles in multiple regions (hereinafter referred to as the prediction device), and in particular, it can be executed by one or more processors within the prediction device.

[0042] In this embodiment, the prediction device may be an electronic device equipped with a processor, which carries a computer program for the prediction method of the short-term distribution location of intercity ride-hailing vehicles in multiple regions and the computer program can be executed, such as a computer, workstation, server, etc. The present invention does not make any specific limitations.

[0043] In this embodiment, the Kalman Filter Model is an efficient recursive filter capable of estimating the state of a dynamic system from a series of noisy observations. Based on the state-space representation of linear dynamic systems, the Kalman Filter Model assumes that the system state is linear and that both process and observation noise are Gaussian distributed. The model iteratively estimates the system state through two main steps: prediction and update.

[0044] Prediction phase: Predict the current state based on the system's previous state and control input; simultaneously predict the current state covariance.

[0045] Update phase: Update the prediction using new measurement data; calculate a gain (Kalman gain) that determines which part of the prediction and measurement is more reliable; update the estimated state and covariance.

[0046] like Figure 1 As shown, a method for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions includes steps S1 to S7.

[0047] S1, based on the real-time road network data of intercity ride-hailing services in cities A and B, divide the area into several sub-regions and construct the designated intercity entrance / exit e for city A. A and the intercity entrance / exit e of city B B .

[0048] The road network data includes the internal road network structure of the two cities and intercity entrances / exits. A and e B The expressway connects the two cities; the two ends of the expressway connect the only intercity entrances and exits of the two cities respectively. Intercity ride-hailing vehicles leave the city through the intercity entrances and exits, pass through the expressway, and then arrive in the other city through the intercity entrance and exit of the other city.

[0049] Divide city A into |U| regions A U U = {1,…,u,…,|U|}, where u is the area code of city A; divide city B into |W| areas B. W W = {1,…,w,…,|W|}, where w is the area code of city B; |U| and |W| are the number of areas divided between city A and city B.

[0050] S2, based on the passenger order data of each sub-region of city A and city B, construct the initial vehicle registration set K for each sub-region of city A and city B respectively. A and K B .

[0051] Based on the number of drivers operating on routes in each sub-region u of city A, construct an initial vehicle registration set K consisting of the same number of vehicles as the number of drivers. A , The vehicle registration set for sub-region u of city A;

[0052] The fact that vehicle k reports for duty means that vehicle k can be used for subsequent dispatching. After delivering the last passenger, vehicle k will join the vehicle reporting set of its city.

[0053] The order data for each passenger p can be represented by tuples. Composition, in which o p It is the boarding location for passenger p, d p It's the drop-off point, u p It is the area code of the boarding location, w p It is the area code of the drop-off location, u p and w p They do not belong to the same city. That's the earliest departure time. That's the latest train departure time;

[0054] Let the set of time numbers for platform scheduling decisions be T, T = {0, 1, ..., t, ..., |T|}, where any time interval [t-1, t] is equal.

[0055] The vehicle k at time t has: rated capacity The position at time t is l t Vehicle status s t The vehicle's area code is u t The assigned passenger set P t ={p1,…,p i ,,…p q}, that is, possessing tuples Furthermore, vehicle k must be within passenger p's travel time window. Passengers were picked up inside; q represents the number of passengers.

[0056] Similarly, construct the initial vehicle registration set K for city B. B .

[0057] S3, based on the initial vehicle reporting set K A and K N A multi-task adaptive Kalman filter model is established for each vehicle. This model takes the vehicle's highest detour coefficient at the current moment, the number of passengers, and the previous state data as input, and outputs the vehicle's next predicted state data and predicted departure time.

[0058] When calculating the detour coefficient of vehicle k at the current time, based on the initial vehicle reporting set, the detour coefficient of vehicle k after each passenger insertion is calculated, and the highest detour coefficient r at the current time is updated. t ;

[0059] The detour coefficient = the total travel distance of vehicle k after inserting a new passenger divided by the total travel distance of the vehicle before inserting a new passenger;

[0060] If the calculated detour coefficient is greater than the highest detour coefficient r at the current timet When that happens, update the highest detour coefficient r. t Otherwise, remain unchanged.

[0061] For example, when vehicle k receives its first passenger order, the total distance is 10 kilometers, and the maximum detour coefficient is initialized to 0. At time t, after receiving the second passenger order, it needs to detour to pick up passenger 2. At this time, the total distance is set to 15 kilometers, and the detour coefficient = 15 / 10 = 1.5, which is greater than the current maximum detour coefficient of 0. Therefore, the maximum detour coefficient is updated to 1.5. When picking up the third passenger, the total distance becomes 21 kilometers, and the detour coefficient at this time = 21 / 15 = 1.4. Since 1.4 < 1.5, the maximum detour coefficient remains unchanged.

[0062] In this embodiment, as Figure 2 As shown, since the scheduling of each vehicle is independent, an adaptive Kalman filter model is built for each vehicle. This model can produce different outputs based on different model inputs (such as the vehicle's current maximum detour coefficient and the number of passengers). First, the Kalman filter model is initialized. Then, at each scheduling decision time t, the Kalman filter model is called to predict the next state of the vehicle, reflecting the situation of each vehicle. This is then updated independently to avoid mutual interference between models. The specific steps are as follows:

[0063] Step 1: Initialize the state vector and error covariance matrix of the Kalman filter model:

[0064] In this embodiment, let the state vector be... Error covariance matrix State transition matrix Observation matrix Observation noise covariance matrix Process noise covariance matrix

[0065] Where, d s It is the initial distance of vehicle k from the city exit, t o It is the time offset, d v It is the distance variance, t v It is the time variance;

[0066] Step 2: Based on the number of passengers q at time t of vehicle k and the maximum detour coefficient r at the current time. t The process noise covariance matrix Q and the state transition matrix F are adjusted using the following formulas:

[0067]

[0068] in, and These are the weighting coefficients for the number of passengers in the process noise covariance matrix and the state transition matrix. and These are the weighting coefficients of the detour coefficients in the process noise covariance matrix and the state transition matrix; This is the detour coefficient threshold;

[0069] Step 3: Based on the previous state vector, process noise covariance matrix, and estimated error covariance matrix, predict the next state vector and error covariance, as shown in the following formula:

[0070] x t|t-1 =Fx t-1|t-1 ;

[0071] P ′ t|t-1 =FP ′ t-1|t-1 F * +Q;

[0072] Where, x t-1|t-1 It is the estimated state vector at time t-1, x t|t-1 P is the predicted state vector at time t. ′ t|t-1 P is the prediction error covariance matrix at time t. ′ t-1|t-1 is the estimation error covariance matrix at time t-1, and * is the transpose matrix;

[0073] Then, predict the departure time from the city. The calculation method is: the time offset of the predicted state vector at the current time + time t.

[0074] S4, at each scheduling decision time t within the established platform scheduling decision time, based on the vehicle reporting set K A Construct the set of real vehicle distributions for all sub-regions of city A at time t. And the predicted distribution set of available vehicles in all sub-regions of city A from time t to t+n. n is the estimated step size; similarly, based on the vehicle reporting set K B Construct a set of real vehicle distribution data for all sub-regions of city B. and the set of predicted available vehicle distributions

[0075] The real vehicle distribution set This refers to all sub-regions of city A that have registered for the train at time t but have not passed through the intercity entrance / exit e. A And the remaining capacity on the vehicle The set of vehicles is expressed as:

[0076]

[0077] in, Let represent the set of actual vehicle distributions in subregion u of city A at time t;

[0078] The predicted set of available vehicle distributions This refers to all sub-regions of city A that register for a flight within the time period [t, t+n] without passing through the intercity entrance / exit e. A And the remaining capacity on the vehicle The set of vehicles is expressed as:

[0079]

[0080] At the beginning of each scheduling decision time t, let Right now To ensure the real-time nature and accuracy of the prediction process;

[0081] in, Let represent the set of predicted available vehicle distributions for sub-region u of city A during the time interval [t, t+n].

[0082] Similarly, construct the real vehicle distribution set for city B. and the set of predicted available vehicle distributions

[0083] S5, using each scheduling decision time t as a trigger event, iterates through the real vehicle distribution set of city A. The detour coefficient of the vehicle at time t is calculated, and the detour coefficients and passenger numbers of non-empty and not fully loaded vehicles within the city at time t are input into the vehicle's adaptive Kalman filter model, outputting the vehicle's predicted departure time from the city. Based on the predicted departure time from the city Construct a set of pre-departure vehicles from multiple areas of city A; similarly, consider vehicle k from city B passing through intercity entrance / exit e at time t. B To trigger the event, construct a set of pre-departure vehicles from multiple areas of city B.

[0084] Using each scheduling decision time t as the trigger event, iterate through the real vehicle distribution sets of all sub-regions of city A.

[0085] If vehicles in sub-region u of city A If vehicle k is neither empty nor fully loaded, then the highest detour coefficient r of vehicle k at time t will be determined. t The number of passengers q and the previous state data are input into the adaptive Kalman filter model of vehicle k to obtain the predicted departure time of vehicle k.

[0086] If the departure time is predicted Then vehicle k is selected from the set of real vehicle distributions in region u of city A. and the set of predicted available vehicle distributions Remove vehicle k from the list and add it to the pre-departure vehicle set. In the middle, construct a pre-departure vehicle set:

[0087]

[0088] in, This represents the set of vehicles scheduled to leave city A in subregion u during the time interval [t, t+n]. Vehicle k is used to predict the departure time. Add to middle;

[0089] Similarly, construct the set of vehicles expected to leave city B.

[0090] S6, where vehicle k from city A passes through intercity entrance / exit e at time t. A To trigger the event, iterate through the set of real vehicle distributions in city A. Calculate the set of arrival times for all passengers in vehicles that have already left the city. Construct the set of vehicles en route to city A According to the vehicles gathered along the way The arrival time and drop-off location of the last passenger are used to update the predicted available vehicle distribution set for city B. Similarly, consider vehicle k from city B passing through intercity entrance / exit e at time t. B To trigger the event, construct a set of vehicles en route to city B. And update the predicted available vehicle distribution set for city A

[0091] Iterate through the set of real vehicle distributions in all sub-regions of city A. If vehicles in sub-region u of city A Vehicle k in the middle has passed the intercity entrance / exit e. A Then vehicle k will be from and Remove from;

[0092] like Figure 3 As shown, calculate the set of expected arrival times for all passengers on vehicle k.

[0093]

[0094] in, p represents the i-th passenger i The arrival time, The arrival time of the last passenger. The calculation method is as follows:

[0095] via intercity entrance / exit e A With passenger p i drop-off location d p The distance between them is divided by the default road speed. For example, highways are typically set at 120 km / h, and city roads at 50 km / h.

[0096] Construct the set of vehicles en route to city A like Then the vehicle k will be determined by its arrival time. Add to the en route vehicle collection to city A like Then vehicle k will be based on the arrival time of the last passenger. and the area code of the drop-off location w p After updating the information of vehicle k, add it to the real vehicle distribution set of city B.

[0097] Similarly, construct the set of vehicles en route to city B. And update the predicted available vehicle distribution set for city A

[0098] S7, based on the real-time data of the actual departure times of vehicles that have already left the city. Update the data of the corresponding vehicle's adaptive Kalman filter model until the vehicle reports for duty again, then reset the vehicle's adaptive Kalman filter model.

[0099] In this embodiment, as Figure 4 As shown, initialize t=0 and obtain the real vehicle distribution at time t in real time; set the time number i=0 for the platform scheduling decision, obtain the detour coefficient and passenger number of vehicle k at the current time, perform real-time multi-task Kalman filter prediction, and output the real-time prediction of the departure time of vehicle k, until the real vehicle distribution set of all sub-regions of city A is traversed. All vehicles; then t = t + 1, continue iteratively calculating the real-time prediction of vehicle distribution in all areas of city A during the time period [t, t + n].

[0100] If the vehicle set of subregion u of city A Vehicle k in the middle has passed the intercity entrance / exit e. A Based on the actual departure time of vehicle k Location and intercity entrance / exit when the prediction is triggered A distance The adaptive Kalman filter model for vehicle k is updated as follows:

[0101] Step 1: Calculate the observation residual y t The formula is:

[0102]

[0103] Step 2: Calculate the covariance S of the observed residuals t The formula is:

[0104] S t =HP ′ t|t-1 H * +R;

[0105] Step 3: Calculate the Kalman gain Z t The formula is:

[0106]

[0107] Step 4: Update the state vector x t|t The formula is:

[0108] x t|t =x t|t-1 +Z t y t ;

[0109] Step 5: Update the error covariance matrix P ′ t|t The formula is:

[0110] P ′ t|t =P ′ t|t-1 -K t HK t|t-1 ;

[0111] Similarly, update the set of actual vehicle distributions for all sub-regions of city B. The adaptive Kalman filter model corresponding to vehicles that have already left the city;

[0112] After vehicle k re-enrolls, when resetting the adaptive Kalman filter model for vehicle k, a learning rate lr is introduced to allow the model to have a memory of the historical training process. The reset process is as follows:

[0113] 1) Calculate the average outbound travel time t of vehicle k on the current day. avg and distance d avg ;

[0114] 2) Adjust the state vector x of vehicle k;

[0115]

[0116] 3) Reset the error covariance matrix P of vehicle k ′ That is, initializing P ′ .

[0117] In another specific embodiment, to verify the effectiveness of the model and model solution proposed in this invention, an actual example will be used below to illustrate the application of this invention.

[0118] This study selects order data from a specific intercity ride-hailing route operated by a certain company on June 2, 2023, as the research object. After one day of scheduling, the evaluation metrics for the prediction results at each departure time are the root mean square error (RMSE), mean absolute error (MAE), and symmetric mean absolute percentage error (SMAPE). Lower values ​​for RMSE, MAE, and SMAPE indicate better prediction accuracy of the model. The formulas are shown below:

[0119]

[0120] Among them, y i This refers to the actual number of available vehicles (or seats) in the transportation capacity data. The number of available vehicles (or seats) is predicted, where ti is the number of time slices. The vehicle distribution at adjacent times is considered constant, and the number of available seats in a region can be calculated from the number of available vehicles in that region. The experimental parameters are set as shown in Table 1.

[0121] Table 1. Experimental parameter settings

[0122]

[0123] The comparison algorithms used are HA (Historical Average), XGBOOST (eXtremeGradient Boosting, an ensemble learning algorithm based on gradient boosting framework), LSTM (Long Short-Term Memory, a neural network model), and T-GCN (Temporal Graph Convolutional Network, a temporal graph convolutional network model for traffic flow prediction).

[0124] The dataset used for the comparison algorithm consists of intercity round-trip ride-hailing data for this route from April 1, 2023 to June 4, 2023. The ratio of the training set, validation set, and test set is 7:2:1. The prediction results are shown in Tables 2 and 3.

[0125] Table 2. Prediction Results of Available Vehicle Count for City A <—> City B

[0126]

[0127] Table 3. Predicted Number of Available Seats for City A <—> City B

[0128]

[0129] As shown in Tables 2 and 3, the prediction results of this invention are superior to other comparative algorithms in terms of the prediction results of the number of available vehicles and the number of available seats, and have higher prediction accuracy.

[0130] from Figure 5 and Figure 6 It can be seen that the predicted number of available vehicles and the actual number of vehicles, as well as the predicted number of available seats and the actual number of seats, are highly similar in the prediction results for the sub-regions. This confirms the feasibility of this embodiment in predicting the short-term distribution of intercity ride-hailing vehicles in multiple regions, and verifies that the solution quality of this embodiment is high and can meet the engineering needs of operators.

[0131] In summary, compared with the prior art, the present invention has the following beneficial effects:

[0132] This invention addresses the low accuracy of existing data-driven models in predicting transportation capacity. It establishes a separate model for each vehicle, better capturing their unique characteristics. Based on the current number of passengers and detour coefficients, it adaptively inputs the data into a Kalman filter model, making vehicle distribution predictions interpretable and allowing for traceable vehicle movement. This invention is applicable to situations where vehicle speeds are unavailable and the problem does not involve multiple urban areas. It can predict the location and status of specific vehicles within multiple areas, updating their status in real time. This effectively improves prediction accuracy, adapts to rapidly changing demand environments and scheduling strategies, optimizes driver efficiency, enhances the overall operational efficiency of ride-hailing platforms, and provides high-value scheduling references for ride-hailing platforms. Furthermore, it can be used in subsequent vehicle relocation and scheduling strategies for supply and demand balancing.

[0133] Example 2

[0134] like Figure 7 As shown, the second embodiment of the present invention also provides a device for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple areas, including:

[0135] The multi-region dispatch environment construction unit is used to divide the intercity ride-hailing routes in cities A and B into several sub-regions based on real-time acquired road network data, and to construct the intercity entrance / exit points e for designated city A. A and the intercity entrance / exit e of city B B ;

[0136] The multi-regional vehicle registration set construction unit is used to construct the initial vehicle registration set K for each sub-region of city A and city B based on the passenger order data obtained from each sub-region of city A and city B. A and K B ;

[0137] The multi-task model building unit is used to build upon the initial vehicle reporting set K. A and K B A multi-task adaptive Kalman filter model is established for each vehicle. This model takes the vehicle's highest detour coefficient at the current moment, the number of passengers, and the previous state data as input, and outputs the vehicle's next predicted state data and predicted departure time.

[0138] A multi-regional vehicle distribution set construction unit is used to obtain the actual vehicle distribution set and the predicted available vehicle distribution set. At each scheduling decision time t within the established platform scheduling decision time, based on the vehicle reporting set K... A Construct the set of real vehicle distributions for all sub-regions of city A at time t. And the predicted distribution set of available vehicles in all sub-regions of city A from time t to t+n. n is the estimated step size; similarly, based on the vehicle reporting set K B Construct a set of real vehicle distribution data for all sub-regions of city B. and the set of predicted available vehicle distributions

[0139] The multi-regional pre-departure vehicle set construction unit is used to obtain the multi-regional pre-departure vehicle set. It iterates through the real vehicle distribution set of city A, with each scheduling decision time t as the trigger event. The detour coefficient of the vehicle at time t is calculated, and the detour coefficients and passenger numbers of non-empty and not fully loaded vehicles within the city at time t are input into the vehicle's adaptive Kalman filter model, outputting the vehicle's predicted departure time from the city. Based on the predicted departure time from the city Construct a set of pre-departure vehicles for multiple areas in city A; similarly, construct a set of pre-departure vehicles for multiple areas in city B.

[0140] The vehicle set construction unit is used to obtain the vehicle set in transit, such as vehicle k in city A passing through intercity entrance / exit e at time t. A To trigger the event, iterate through the set of real vehicle distributions in city A. Calculate the set of arrival times for all passengers in vehicles that have already left the city. Construct the set of vehicles en route to city A According to the vehicles gathered along the way The arrival time and drop-off location of the last passenger are used to update the predicted available vehicle distribution set for city B. Similarly, consider vehicle k from city B passing through intercity entrance / exit e at time t. B To trigger the event, construct a set of vehicles en route to city B. And update the predicted available vehicle distribution set for city A

[0141] The model update unit is used to update the model based on the actual departure times of vehicles that have already left the city, obtained in real time. Update the data of the corresponding vehicle's adaptive Kalman filter model until the vehicle reports for duty again, then reset the vehicle's adaptive Kalman filter model.

[0142] Example 3

[0143] The third embodiment of the present invention also provides a device for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions, which includes a memory and a processor. The memory stores a computer program, which can be executed by the processor to realize the method for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions as described above.

[0144] Example 4

[0145] The fourth embodiment of the present invention also provides a computer-readable storage medium storing computer-readable instructions. When the computer-readable instructions are executed by the processor of the device where the computer-readable storage medium is located, they implement the above-described method for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions.

[0146] In the several embodiments provided in this invention, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus and method embodiments described above are merely illustrative. For example, the flowcharts in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of the invention. In this regard, each block in the flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0147] In addition, the functional modules in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0148] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, electronic device, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks. It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. In the absence of further restrictions, an element defined by the phrase "comprising a..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0149] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0150] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0151] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0152] The use of "first" and "second" in the embodiments is merely to distinguish similar objects and does not represent a specific ordering of objects. It is understood that "first" and "second" can be interchanged in a specific order or sequence where permitted. It should be understood that the objects distinguished by "first" and "second" can be interchanged where appropriate so that the embodiments described herein can be implemented in an order other than those illustrated or described herein.

[0153] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting the short-term distribution location of intercity ride-hailing vehicles across multiple regions, characterized in that, include: S1, based on the real-time road network data of intercity ride-hailing services in cities A and B, divide the area into several sub-regions and construct the designated intercity entrance / exit e for city A. A and the intercity entrance / exit e of city B B ; S2, based on the passenger order data of each sub-region of city A and city B, construct the initial vehicle registration set K for each sub-region of city A and city B respectively. A and K B ; S3, based on the initial vehicle reporting set K A and K B A multi-task adaptive Kalman filter model is established for each vehicle. This model takes the vehicle's highest detour coefficient at the current moment, the number of passengers, and the previous state data as input, and outputs the vehicle's next predicted state data and predicted departure time. S4, at each scheduling decision time t within the established platform scheduling decision time, based on the initial vehicle reporting set K A Construct the set of real vehicle distributions for all sub-regions of city A at time t. And the predicted distribution set of available vehicles in all sub-regions of city A from time t to t+n. n is the estimated step size; similarly, based on the initial vehicle reporting set K... B Construct a set of real vehicle distribution data for all sub-regions of city B. and the set of predicted available vehicle distributions S5, using each scheduling decision time t as a trigger event, iterates through the real vehicle distribution set of city A. The detour coefficient of the vehicle at time t is calculated, and the detour coefficients and passenger numbers of non-empty and not fully loaded vehicles within the city at time t are input into the vehicle's adaptive Kalman filter model, outputting the vehicle's predicted departure time from the city. Based on the predicted departure time from the city Construct a set of pre-departure vehicles for multiple areas in city A; similarly, construct a set of pre-departure vehicles for multiple areas in city B. S6, where vehicle k from city A passes through intercity entrance / exit e at time t. A To trigger the event, iterate through the set of real vehicle distributions in city A. Calculate the set of arrival times for all passengers in vehicles that have already left the city. Construct the set of vehicles en route to city A According to the vehicles gathered along the way The arrival time and drop-off location of the last passenger are used to update the predicted available vehicle distribution set for city B. Similarly, consider vehicle k from city B passing through intercity entrance / exit e at time t. B To trigger the event, construct a set of vehicles en route to city B. And update the predicted available vehicle distribution set for city A S7, based on the real-time data of the actual departure times of vehicles that have already left the city. Update the data of the corresponding vehicle's adaptive Kalman filter model until the vehicle reports for duty again, then reset the vehicle's adaptive Kalman filter model.

2. The method for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions according to claim 1, characterized in that... In step S1: The road network data includes the internal road network structure of the two cities and intercity entrances / exits. A and e B The expressway connects the two cities; the two ends of the expressway connect the only intercity entrances and exits of the two cities respectively. Intercity ride-hailing vehicles leave the city through the intercity entrances and exits, pass through the expressway, and then arrive in the other city through the intercity entrance and exit of the other city. Divide city A into |U| regions A U U = {1,…,u,…,|U|}, where u is the area code of city A; Divide city B into |W| regions B W W = {1,…,w,…,|W|}, where w is the area code of city B.

3. The method for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions according to claim 2, characterized in that... In step S2: Based on the number of drivers operating on routes in each sub-region u of city A, construct an initial vehicle registration set K consisting of the same number of vehicles as the number of drivers. A , The vehicle registration set for sub-region u of city A; The vehicle k reporting means that vehicle k can be used for subsequent dispatching. After the last passenger is delivered, vehicle k will join the vehicle reporting set of the city. The order data for each passenger p consists of tuples. Composition, in which o p It is the boarding location for passenger p, d p It's the drop-off point, u p It is the area code of the boarding location, w p It is the area code of the drop-off location, u p and w p They do not belong to the same city. That's the earliest departure time. That's the latest train departure time; The vehicle k at time t has: rated capacity The position at time t is l t Vehicle status s t The vehicle's area code is u t The assigned passenger set P t ={p1,…,p i ,,…p q }, that is, possessing tuples Furthermore, vehicle k must be within passenger p's travel time window. Passengers were picked up inside; q represents the number of passengers; Similarly, construct the initial vehicle registration set K for city B. B .

4. The method for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions according to claim 3, characterized in that... When calculating the detour coefficient of vehicle k at the current moment, based on the initial vehicle reporting set, the detour coefficient of vehicle k after each passenger insertion is calculated, and the highest detour coefficient r at the current moment is updated. t ; The detour coefficient = the total travel distance of vehicle k after inserting a new passenger divided by the total travel distance of the vehicle before inserting a new passenger; If the calculated detour coefficient is greater than the highest detour coefficient r at the current moment t When that happens, update the highest detour coefficient r. t Otherwise, remain unchanged.

5. The method for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions according to claim 4, characterized in that... In step S3, when establishing a multi-task adaptive Kalman filter model for each vehicle, the Kalman filter model is first initialized. Then, at each scheduling decision time t, the Kalman filter model is called to predict the next state of the vehicle to reflect the situation of each vehicle and is updated independently to avoid mutual interference between models. The specific steps are as follows: Step 1: Initialize the state vector and error covariance matrix of the Kalman filter model: Define state vector The error covariance matrix is ​​P ′ The state transition matrix is ​​F, the observation matrix is ​​H, and the observation noise covariance matrix is... The process noise covariance matrix is ​​Q; Where, d s It is the initial distance of vehicle k from the city exit, t o It is the time offset, d v It is the distance variance, t v It is the time variance; Step 2: Based on the number of passengers q at time t of vehicle k and the maximum detour coefficient r at the current time. t The process noise covariance matrix Q and the state transition matrix F are adjusted using the following formulas: in, and These are the weighting coefficients for the number of passengers in the process noise covariance matrix and the state transition matrix. and These are the weighting coefficients of the detour coefficients in the process noise covariance matrix and the state transition matrix; This is the detour coefficient threshold; Step 3: Based on the previous state vector, process noise covariance matrix, and estimated error covariance matrix, predict the next state vector and error covariance, as shown in the following formula: x t|t-1 =Fx t-1|t-1 ; P ′ t|t-1 =FP ′ t-1|t-1 F * +Q; Where, x t-1|t-1 It is the estimated state vector at time t-1, x t|t-1 P is the predicted state vector at time t. ′ t|t-1 P is the prediction error covariance matrix at time t. ′ t-1|t-1 is the estimation error covariance matrix at time t-1, and * is the transpose matrix; Then, predict the departure time from the city. The time offset of the predicted state vector at the current time plus time t.

6. The method for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions according to claim 5, characterized in that... In step S4, the real vehicle distribution set This refers to all sub-regions of city A that have registered for the train at time t but have not passed through the intercity entrance / exit e. A And the remaining capacity on the vehicle The set of vehicles is expressed as: in, Let represent the set of actual vehicle distributions in subregion u of city A at time t; The predicted set of available vehicle distributions This refers to all sub-regions of city A that register for a flight within the time period [t, t+n] without passing through the intercity entrance / exit e. A And the remaining capacity on the vehicle The set of vehicles is expressed as: At the beginning of each scheduling decision time t, let Right now To ensure the real-time nature and accuracy of the prediction process; in, Let represent the set of predicted available vehicle distributions for sub-region u of city A during the time interval [t, t+n]. Similarly, construct the real vehicle distribution set for city B. and the set of predicted available vehicle distributions 7. The method for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions according to claim 6, characterized in that... Step S5 specifically includes: Using each scheduling decision time t as the trigger event, iterate through the real vehicle distribution sets of all sub-regions of city A. If vehicles in sub-region u of city A If vehicle k is neither empty nor fully loaded, then the highest detour coefficient r of vehicle k at time t will be determined. t The number of passengers q and the previous state data are input into the adaptive Kalman filter model of vehicle k to obtain the predicted departure time of vehicle k. If the departure time is predicted Then vehicle k is selected from the set of real vehicle distributions in region u of city A. and the set of predicted available vehicle distributions Remove vehicle k from the list and add it to the pre-departure vehicle set. In the middle, construct a pre-departure vehicle set: in, This represents the set of vehicles scheduled to leave city A in subregion u during the time interval [t, t+n]. Vehicle k is used to predict the departure time. Add to middle; Similarly, construct the set of vehicles expected to leave city B.

8. The method for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions according to claim 7, characterized in that... Step S6 specifically includes: Iterate through the set of real vehicle distributions in all sub-regions of city A. If vehicles in sub-region u of city A Vehicle k in the middle has passed the intercity entrance / exit e. A Then vehicle k will be moved from and Remove from; Calculate the set of arrival times for all passengers on vehicle k. in, p represents the i-th passenger i The arrival time, The arrival time of the last passenger. The calculation method is as follows: via intercity entrance / exit e A With passenger p i drop-off location d p The distance between them, divided by the default road speed; Construct the set of vehicles en route to city A like Then the vehicle k will be determined by its arrival time. Add to the en route vehicle collection to city A like Then vehicle k will be based on the arrival time of the last passenger. and the area code of the drop-off location w p After updating the information of vehicle k, add it to the real vehicle distribution set of city B. Similarly, construct the set of vehicles en route to city B. And update the predicted available vehicle distribution set for city A 9. The method for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple regions according to claim 8, characterized in that... Step S7 specifically includes: Iterate through the set of real vehicle distributions in all sub-regions of city A. If the vehicle set of subregion u of city A Vehicle k in the middle has passed the intercity entrance / exit e. A Based on the actual departure time of vehicle k Location and intercity entrance / exit when the prediction is triggered A distance The adaptive Kalman filter model for vehicle k is updated as follows: Step 1: Calculate the observation residual y t The formula is: Step 2: Calculate the covariance S of the observed residuals t The formula is: S t =HP ′ t|t-1 H * +R; Step 3: Calculate the Kalman gain Z t The formula is: Step 4: Update the state vector x t|t The formula is: x t|t =x t|t-1 +Z t y t ; Step 5: Update the error covariance matrix P ′ t|t The formula is: P ′ t|t =P ′ t|t-1 -K t HK t|t-1 ; Similarly, update the set of actual vehicle distributions for all sub-regions of city B. The adaptive Kalman filter model corresponding to vehicles that have already left the city; After vehicle k re-enrolls, when resetting the adaptive Kalman filter model for vehicle k, a learning rate lr is introduced to allow the model to have a memory of the historical training process. The reset process is as follows: 1) Calculate the average outbound travel time t of vehicle k on the current day. avg and distance d avg ; 2) Adjust the state vector x of vehicle k; Where lr is the learning rate; 3) Reset the error covariance matrix P of vehicle k ′ That is, initializing P ′ .

10. A device for predicting the short-term distribution location of intercity ride-hailing vehicles in multiple areas, characterized in that, include: The multi-region dispatch environment construction unit is used to divide the intercity ride-hailing routes in cities A and B into several sub-regions based on real-time acquired road network data, and to construct the intercity entrance / exit points e for designated city A. A and the intercity entrance / exit e of city B B ; The multi-regional vehicle registration set construction unit is used to construct the initial vehicle registration set K for each sub-region of city A and city B based on the passenger order data obtained from each sub-region of city A and city B. A and K B ; The multi-task model building unit is used to build upon the initial vehicle reporting set K. A and K B A multi-task adaptive Kalman filter model is established for each vehicle. This model takes the vehicle's highest detour coefficient at the current moment, the number of passengers, and the previous state data as input, and outputs the vehicle's next predicted state data and predicted departure time. A multi-regional vehicle distribution set construction unit is used to obtain the actual vehicle distribution set and the predicted available vehicle distribution set. At each scheduling decision time t within the established platform scheduling decision time, based on the initial vehicle reporting set K... A Construct the set of real vehicle distributions for all sub-regions of city A at time t. And the predicted distribution set of available vehicles in all sub-regions of city A from time t to t+n. n is the estimated step size; similarly, based on the initial vehicle reporting set K... B Construct a set of real vehicle distribution data for all sub-regions of city B. and the set of predicted available vehicle distributions The multi-regional pre-departure vehicle set construction unit is used to obtain the multi-regional pre-departure vehicle set. It iterates through the real vehicle distribution set of city A, with each scheduling decision time t as the trigger event. The detour coefficient of the vehicle at time t is calculated, and the detour coefficients and passenger numbers of non-empty and not fully loaded vehicles within the city at time t are input into the vehicle's adaptive Kalman filter model, outputting the vehicle's predicted departure time from the city. Based on the predicted departure time from the city Construct a set of pre-departure vehicles for multiple areas in city A; similarly, construct a set of pre-departure vehicles for multiple areas in city B. The vehicle set construction unit is used to obtain the vehicle set in transit, such as vehicle k in city A passing through intercity entrance / exit e at time t. A To trigger the event, iterate through the set of real vehicle distributions in city A. Calculate the set of arrival times for all passengers in vehicles that have already left the city. Construct the set of vehicles en route to city A According to the vehicles gathered along the way The arrival time and drop-off location of the last passenger are used to update the predicted available vehicle distribution set for city B. Similarly, consider vehicle k from city B passing through intercity entrance / exit e at time t. B To trigger the event, construct a set of vehicles en route to city B. And update the predicted available vehicle distribution set for city A The model update unit is used to update the model based on the actual departure times of vehicles that have already left the city, which are obtained in real time. Update the data of the corresponding vehicle's adaptive Kalman filter model until the vehicle reports for duty again, then reset the vehicle's adaptive Kalman filter model.

Images