Trajectory prediction method and device, electronic equipment and storage medium

Abstract

The application discloses a trajectory prediction method and device, electronic equipment and a storage medium. The method comprises the following steps: determining an initial trajectory point at a starting time; inputting weighted trajectory information from the initial trajectory point to a trajectory point at a current time of a current trajectory into a prediction model within a predetermined time window to obtain prediction information of a trajectory point at a next time of the current time of the current trajectory; determining at least one predicted trajectory based on the weighted trajectory information of the trajectory point at the current time and the prediction information of the predicted trajectory point at the next time; and determining a predicted trajectory satisfying a first predetermined condition as a target predicted trajectory based on the at least one predicted trajectory. Thus, compared with the related art which only predicts a next position, the target predicted trajectory in the embodiment of the application has integrity and continuity; the influence of historical trajectory points on the predicted position of the predicted trajectory point is fully considered, and the accuracy and effectiveness of the target predicted trajectory are improved.

Description

Trajectory prediction methods, devices, electronic equipment and storage media Technical Field

[0001] This application relates to the field of information technology, and in particular to a trajectory prediction method, apparatus, electronic device, and storage medium. Background Technology

[0002] Predicting the long-term location coordinates of mobile users is fundamental to many emerging applications, such as anomaly detection, location-based service recommendation systems, tracking and positioning, and smart homes. However, mobile users' movement behavior is influenced by both their internal drives and environmental factors, making it complex, diverse, and uncertain. Accurately obtaining predicted user trajectories remains a challenging problem.

[0003] Existing technologies for predicting user movement trajectories mostly employ multi-layer dynamic Bayesian network models or neural network models, which only predict the next location of the mobile user and cannot predict the complete and continuous trajectory of the mobile user. Summary of the Invention

[0004] In view of this, embodiments of this application provide a trajectory prediction method, apparatus, electronic device, and storage medium.

[0005] The technical solution of this application embodiment is implemented as follows:

[0006] In a first aspect, embodiments of this application provide a trajectory prediction method, the method comprising:

[0007] Determine the initial trajectory point at the start time;

[0008] Within a predetermined time window, the weighted trajectory information from the initial trajectory point to the current trajectory at the current moment is input into the prediction model to obtain the prediction information of the trajectory point at the next moment of the current trajectory; wherein, the weighted trajectory information is determined based on the trajectory information of the trajectory point and the weight corresponding to the trajectory point; the trajectory point at the next moment includes the predicted position of at least one predicted trajectory point; the prediction information of the trajectory point at the next moment includes at least the trajectory information;

[0009] Based on the weighted trajectory information of the trajectory point at the current time and the prediction information of the predicted trajectory point at the next time, at least one predicted trajectory is determined;

[0010] Based on at least one of the predicted trajectories, the predicted trajectory that satisfies a first predetermined condition is determined as the target predicted trajectory.

[0011] In some embodiments, determining the predicted trajectory that satisfies a first predetermined condition as the target predicted trajectory based on at least one of the predicted trajectories includes:

[0012] The total similarity of the predicted trajectory is determined based on the similarity between at least one of the predicted trajectories and at least one historical trajectory.

[0013] The predicted trajectory corresponding to the total similarity that satisfies the first predetermined condition is determined as the target predicted trajectory.

[0014] In some embodiments, the prediction information of the trajectory point at the next moment further includes the probability value;

[0015] The step of determining the predicted trajectory that satisfies a first predetermined condition as the target predicted trajectory based on at least one of the predicted trajectories includes:

[0016] Based on the probability value corresponding to at least one of the predicted trajectories, the predicted trajectory corresponding to the probability value that satisfies the first predetermined condition is determined as the target predicted trajectory.

[0017] In some embodiments, the predetermined time window includes time windows 1 to n; the starting time of the first time window is the mi-th time, and the current time of the first time window is the m-th time; n, m, and i are all integers greater than 0; m is greater than i.

[0018] The method further includes: determining the prediction information of the trajectory point at time (m+1) based on the trajectory information from the trajectory point at time (m-1) to the trajectory point at time (m).

[0019] In some embodiments, the starting time of the kth time window among the 2nd to nth time windows is the (m-i+k-1)th time window; the current time of the kth time window is the (m+k-1)th time window; where k is an integer;

[0020] The method further includes: determining the prediction information of the trajectory point at the (m-i+k-1)th time based on the trajectory information from the trajectory point at the (m-i+k-1)th time to the trajectory point at the (m+k-1)th time.

[0021] In some embodiments, the method includes:

[0022] Obtain trajectory information of at least one trajectory point from at least one historical trajectory and trajectory information of at least one trajectory point from the current trajectory;

[0023] Based on the trajectory information of the trajectory points of the historical trajectory and the trajectory information of the trajectory points of the current trajectory, at least one metric value between the historical trajectory and the current trajectory is determined;

[0024] The determination of the initial trajectory point at the start time includes:

[0025] The initial trajectory point of the historical trajectory whose metric value satisfies the second predetermined condition is determined as the initial trajectory point of the starting time of the first time window.

[0026] In some embodiments, determining at least one metric value between the historical trajectory and the current trajectory based on the trajectory information of trajectory points in the historical trajectory and the trajectory information of trajectory points in the current trajectory includes:

[0027] The first parameter is determined based on the time interval between the current time of the trajectory point of the current trajectory and the historical time corresponding to the trajectory point of at least one historical trajectory;

[0028] Using dynamic normalization, based on the product of the trajectory information of the trajectory points of the historical trajectory and the trajectory information of the trajectory points of the current trajectory and the first parameter, at least one first metric value of the historical trajectory and the current trajectory is determined; and / or, using distance editing, based on the sum of the trajectory information of the trajectory points of the historical trajectory and the trajectory information of the trajectory points of the current trajectory and the first parameter, at least one second metric value of the historical trajectory and the current trajectory is determined.

[0029] In some embodiments, the historical trajectory includes: a first historical trajectory and a second historical trajectory;

[0030] The method further includes:

[0031] Obtain the metrics of a first historical trajectory and at least one second historical trajectory;

[0032] If the metric value is less than or equal to the first threshold, the second historical trajectory corresponding to the metric value is determined to be a valid trajectory.

[0033] The measurement value is determined based on the number of times the trajectory information of at least one trajectory point of the first historical trajectory overlaps with the trajectory information of the trajectory points of the effective trajectory;

[0034] If the measured value is greater than or equal to the second threshold, the trajectory point of the first historical trajectory corresponding to the measured value is determined as the trajectory point for determining the predicted trajectory.

[0035] Secondly, embodiments of this application provide a trajectory prediction device, the device comprising:

[0036] The first determining module is used to determine the initial trajectory point at the start time;

[0037] The calculation module is used to input the weighted trajectory information of the trajectory points from the initial trajectory point to the current time point of the current trajectory into the prediction model within a predetermined time window, so as to obtain the prediction information of the trajectory points at the next time point of the current trajectory; wherein, the weighted trajectory information is determined based on the trajectory information of the trajectory points and the weights corresponding to the trajectory points; the trajectory points at the next time point include the predicted positions of at least one predicted trajectory point; the prediction information of the trajectory points at the next time point includes at least the trajectory information;

[0038] The second determining module is used to determine at least one of the predicted trajectories based on the weighted trajectory information of the trajectory points at the current time and the prediction information of the predicted trajectory points at the next time.

[0039] The third determining module is used to determine, based on at least one of the predicted trajectories, the predicted trajectory that satisfies a first predetermined condition as the target predicted trajectory.

[0040] Thirdly, embodiments of this application provide an electronic device, the electronic device comprising:

[0041] Memory, which stores computer-readable instructions;

[0042] The processor, connected to the memory, is configured to implement the trajectory prediction method provided in the first aspect by executing the computer-readable instructions.

[0043] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions thereon, which, when executed by a processor, implement the trajectory prediction method provided in the first aspect above.

[0044] This application provides a trajectory prediction method, apparatus, server, and storage medium. The method includes: determining an initial trajectory point at a starting time; within a predetermined time window, inputting weighted trajectory information from the initial trajectory point to the current trajectory point at the current time into a prediction model to obtain prediction information for the trajectory point at the next time step of the current trajectory; determining at least one predicted trajectory based on the weighted trajectory information of the current trajectory point and the prediction information of the predicted trajectory point at the next time step; and determining a target predicted trajectory based on the at least one predicted trajectory that satisfies a first predetermined condition.

[0045] In the above trajectory prediction process, compared with related technologies that only predict the next position, the target predicted trajectory in this application embodiment has completeness and continuity; it fully considers the influence of historical trajectory points on the predicted position of the predicted trajectory points, thereby improving the accuracy and effectiveness of the target predicted trajectory.

[0046] Furthermore, if the obtained target prediction trajectory is applied to a tracking and positioning scenario, the target prediction trajectory can provide accurate and reliable location information of the target object; if the obtained target prediction trajectory is applied to a smart home scenario, the prediction information of the target prediction trajectory can also be used to control the opening and / or closing of smart home devices, thereby improving the user experience. Attached Figure Description

[0047] Figure 1 is a flowchart illustrating a trajectory prediction method provided in an embodiment of this application;

[0048] Figure 2 is a flowchart illustrating a time window provided in an embodiment of this application;

[0049] Figure 3 is a flowchart illustrating a trajectory prediction method provided in an embodiment of this application;

[0050] Figure 4 is a flowchart illustrating a trajectory prediction method provided in an embodiment of this application;

[0051] Figure 5 is a schematic diagram of the result of determining a measurement value according to an embodiment of this application;

[0052] Figure 6 is a flowchart illustrating a trajectory prediction method provided in an embodiment of this application;

[0053] Figure 7 is a schematic diagram of the structure of a trajectory prediction device provided in an embodiment of this application;

[0054] Figure 8 is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0055] The present application will now be described in further detail with reference to the accompanying drawings and embodiments.

[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application.

[0057] This application provides a trajectory prediction method. Figure 1 is a schematic diagram of the implementation process of the trajectory prediction method provided in this application. As shown in Figure 1, the method mainly includes the following steps:

[0058] Step S110: Determine the initial trajectory point at the start time;

[0059] Step S120: Within a predetermined time window, input the weighted trajectory information from the initial trajectory point to the trajectory point at the current moment of the current trajectory into a prediction model to obtain prediction information on the trajectory point at the next moment of the current moment of the current trajectory; wherein, the weighted trajectory information is determined based on the trajectory information of the trajectory point and the weight corresponding to the trajectory point; the trajectory point at the next moment includes the predicted positions of at least one predicted trajectory point; the prediction information on the trajectory point at the next moment includes at least the trajectory information;

[0060] Step S130: Based on the weighted trajectory information of the trajectory point at the current moment and the prediction information of the predicted trajectory point at the next moment, determine at least one of the predicted trajectories;

[0061] Step S140: Based on at least one of the predicted trajectories, determine the predicted trajectory that satisfies the first predetermined condition as the target predicted trajectory.

[0062] In the embodiments of the present application, the trajectory prediction method involved can be executed by a terminal. The terminal can be any mobile terminal or fixed terminal, such as, but not limited to, a mobile communication device, a computer, a robot, a vehicle-mounted device, or a wearable device, etc.

[0063] Here, the predetermined time window includes at least one time window.

[0064] Exemplarily, the predetermined time window includes the 1st to the nth time windows. Among them, the 1st time window can be a time window with an arbitrary historical moment as the start moment and the next moment of the current moment as the end moment; the start moment of the ith time window is later than the start moment of the (i - 1)th time window; the difference between the start moment of the ith time window and the start moment of the (i - 1)th time window, and the difference between the start moment of the (i - 1)th time window and the start moment of the (i - 2)th time window are within a predetermined time range; 2 < i ≤ n, and n is a positive integer.

[0065] Exemplarily, as shown in Figure 2, the start moment of the 1st time window is t i , and the end moment is t1; the start moment of the 2nd time window is t i+1 , and the end moment is t2.

[0066] Here, the start moment is the first moment of the predetermined time window.

[0067] Here, the weighted trajectory information is determined based on the trajectory information of the trajectory point and the weight corresponding to the trajectory point.

[0068] In some embodiments, the trajectory information includes at least: target object identifier, location information, and time information; the trajectory information may also include semantic information. The time information includes: arrival time and / or departure time.

[0069] Here, location information includes at least: geographic location information, GPS location information, and latitude and longitude information. For example, location information could be D Street, District C, City B, Province A, Country X; or location information could be Hospital X, Province A; or location information could be 13°N, 25°E.

[0070] In some embodiments, the weights corresponding to historical trajectory points are determined based on the similarity between the current trajectory point of the target object and the trajectory points of historical trajectories; the weights corresponding to historical trajectory points are positively correlated with the similarity.

[0071] Here, the prediction information includes at least the predicted trajectory information, and may also include probability values.

[0072] Here, the predicted trajectory information includes at least: target object identifier, predicted location information, and predicted time information; the predicted trajectory information may also include predicted semantic information. The predicted time information includes: predicted arrival time and / or predicted departure time.

[0073] Here, the probability value indicates the likelihood that the target object will reach the predicted location of the trajectory point and / or the likelihood that the predicted trajectory is the target's predicted trajectory.

[0074] Here, the predicted trajectory is a complete continuous trajectory, which includes at least two trajectory points. For example, the predicted trajectory is from Cinema A to Metro Station B, and then from Metro Station B to Company C; or, the predicted trajectory is from City B in Province A to City D in Province C.

[0075] In some embodiments, at least one predicted trajectory is determined based on the current trajectory point and the predicted position of the predicted trajectory point at the next time step. For example, the current trajectory point is the target object's home, the predicted position of the first predicted trajectory point at the next time step is shopping mall A near the home, and the first predicted trajectory is determined based on the target object's home and shopping mall A; the predicted position of the second predicted trajectory point at the next time step is bus stop C, and the predicted position of the predicted trajectory point at the time after that is the target object's company, and the second predicted trajectory is determined based on the target object's home, bus stop C, and company; the predicted position of the third predicted trajectory point at the next time step is bank B, and the third predicted trajectory is determined based on the target object's home and bank B.

[0076] Here, the preset prediction model can be a neural network model. Exemplarily, the neural network model can be a long short-term memory network model, a recurrent neural network model, a convolutional neural network model, etc.

[0077] In some embodiments, the first predetermined condition can be that the similarity between the predicted trajectory and the historical trajectory is the largest or the second largest. Exemplarily, the similarity of the first predicted trajectory is S1, the similarity of the second predicted trajectory is S2, and the similarity of the third predicted trajectory is S3; if S1 ≤ S3 < S2, then the second predicted trajectory corresponding to S2 is determined as the target predicted trajectory.

[0078] In some embodiments, the target predicted trajectory can be applied to tracking and positioning. For example, during the tracking process of a target object, the possible positions and times of the target object can be judged according to the target predicted trajectory, and positioning can be carried out in a timely manner.

[0079] In some embodiments, the target predicted trajectory can be applied to smart life. For example, the terminal determines that the target predicted trajectory of the target object is to leave home at the first moment, arrive at the subway station to take the subway at the second moment, and arrive at the company to go to work at the third moment. The terminal can output recommendation information to the target object before the first moment, and the recommendation information includes reminder information, control information, and subway schedule; among them, the reminder information is used to remind the target object to go out in time to avoid being late; the control information is used to recommend that the target object turn on or off smart home devices at a certain moment. Exemplarily, the target object can set to turn off the air conditioner and stereo and turn on the sweeping robot according to the control information; the subway schedule is used to remind the user of the real-time arrival time of the subway, and the target object can adjust the time to go out according to the real-time arrival time of the subway, reduce the waiting time for the subway, and improve the user experience.

[0080] Thus, compared with only predicting the next position in the related art, the target predicted trajectory in the embodiments of the present application has integrity and continuity; fully considers the influence of historical trajectory points on the predicted position of the predicted trajectory points, improves the accuracy and effectiveness of the target predicted trajectory; and if the obtained target predicted trajectory is applied to the tracking and positioning scenario, the target predicted trajectory can provide accurate and reliable position information of the target object; if the obtained target predicted trajectory is applied to the smart home scenario, the prediction information of the target predicted trajectory can also be used to control the turning on and / or off of smart home devices, improving the user experience.

[0081] In some embodiments, step S140 includes:

[0082] Based on the similarities between at least one of the predicted trajectories and at least one historical trajectory, determine the total similarity of the predicted trajectories;

[0083] The predicted trajectory corresponding to the total similarity that satisfies the first predetermined condition is determined as the target predicted trajectory.

[0084] In some embodiments, any one of the K predicted trajectories can be compared with N historical trajectories to determine N similarities. The largest similarity among the N similarities is determined as the total similarity of the predicted trajectory. The K predicted trajectories can be compared to determine K total similarities. The predicted trajectory with the largest total similarity among the K total similarities is determined as the target predicted trajectory. Here, K and N are both positive integers.

[0085] Thus, by selecting the most similar predicted trajectory from at least one predicted trajectory as the target predicted trajectory, errors can be reduced and the accuracy of the target predicted trajectory can be improved.

[0086] In some embodiments, the prediction information of the trajectory point at the next moment further includes the probability value;

[0087] Step S140 includes:

[0088] Based on the probability value corresponding to at least one of the predicted trajectories, the predicted trajectory corresponding to the probability value that satisfies the first predetermined condition is determined as the target predicted trajectory.

[0089] In some embodiments, the first predetermined condition may be that the probability value of the predicted trajectory being the target predicted trajectory is the largest or the second largest.

[0090] Here, the probability value indicates the likelihood that the predicted trajectory is the target predicted trajectory.

[0091] In one embodiment, the prediction trajectory with the largest probability value among the K predicted trajectories is selected as the target predicted trajectory.

[0092] Therefore, determining the predicted trajectory corresponding to the highest probability value as the target predicted trajectory can further reduce errors and improve the accuracy and consistency of the target predicted trajectory.

[0093] In some embodiments, the predetermined time window includes time windows 1 to n; the starting time of the first time window is the mi-th time, and the current time of the first time window is the m-th time; n, m, and i are all integers greater than 0; m is greater than i.

[0094] The method further includes: determining the prediction information of the trajectory point at time (m+1) based on the trajectory information from the trajectory point at time (m-1) to the trajectory point at time (m).

[0095] In some embodiments, determining the predicted information of the trajectory point at time (m+1) based on the trajectory information from the trajectory point at time (m-1) to the trajectory point at time (m) includes:

[0096] Using a prediction model, based on the trajectory information from the trajectory point at time m-1 to the trajectory point at time m, the prediction information for the trajectory point at time m+1 is determined.

[0097] In one embodiment, the prediction model is a gated recurrent unit (GRU). Traditional GRUs typically input sequence information, which consists of trajectory information from historical time points. However, not every historical time point has trajectory information; the target object may not generate trajectory information at a particular historical time point or within a specific historical time period, for example, if the target object remains in one place for an extended period. To improve the sparsity of the GRU input information, temporal information is introduced. Weights are assigned based on the importance of the predicted position of the predicted trajectory point at the next time point, according to both sequence and temporal information. The weights indicate the proportion of information contained in the sequence and temporal information during the trajectory prediction process, with 0 ≤ weight ≤ 1.

[0098] As shown in Figure 3, the gated loop unit includes: a forget gate f n Input gate i n Output gate Old cell state h n-1 Temporary unit status and the new unit state h n .

[0099] Forgotten Gate f n The calculation method is as follows:

[0100] Input gate i n The calculation method is as follows:

[0101] Output gate The calculation method is as follows:

[0102] Temporary unit state The calculation method is as follows:

[0103] New unit state h n The calculation method is as follows:

[0104] Among them, A f A i A h A T , b is a linear transformation matrix; fb i b h b T , For linear transformation offset; This is sequence information, indicating the trajectory information of the trajectory point at time m. That is, the trajectory information of the trajectory point at the current moment of the first time window; old state unit h n-1 The trajectory information of the trajectory point at time m-1, h n-1 That is, the trajectory information of the trajectory points of the previous time point of the current time window in the first time window; This is time information, indicating the time interval from time (m-1) to time (m+1); temporary cell state. Used to update old state unit h n-1 To the new unit state h n New unit state h n It indicates the trajectory information of the trajectory point at time m+1; σ represents the logical sigmoid function, which is used as an activation function, and the nonlinear function tanh is also used as an activation function.

[0105] in, Let p be a linear transformation matrix, u be the encoding of the target object identifier, and p be a linear transformation matrix. n-1 Let s be the position of the trajectory point at time m-1. n-1 This represents the semantic information of the trajectory point at time m-1. Let t be a linear transformation matrix, u be the encoding of the target object identifier, and t be... n For the (m+1)th time, t d It represents the time interval from the (m-1)th time to the (m+1)th time.

[0106] Thus, based on the introduced time information The gated loop unit determines the predicted trajectory information, improves the sparsity of the input information of the gated loop unit, and improves the accuracy of the trajectory point prediction information at the (m+1)th time.

[0107] In some embodiments, the starting time of the kth time window among the 2nd to nth time windows is the (m-i+k-1)th time window; the current time of the kth time window is the (m+k-1)th time window; where k is an integer;

[0108] The method further includes: determining the prediction information of the trajectory point at the (m-i+k-1)th time based on the trajectory information from the trajectory point at the (m-i+k-1)th time to the trajectory point at the (m+k-1)th time.

[0109] In some embodiments, determining the prediction information of the trajectory point at the (m + k)-th moment based on the trajectory information from the trajectory point at the (m - i + k - 1)-th moment to the trajectory point at the (m + k - 1)-th moment includes:

[0110] Using a prediction model, based on the trajectory information from the trajectory point at the (m - i + k - 1)-th moment to the trajectory point at the (m + k - 1)-th moment, determine the prediction information of the trajectory point at the (m + k)-th moment.

[0111] In some embodiments, the prediction information includes predicted trajectory information and a probability value. The probability value indicates the likelihood that the target object reaches this trajectory point and / or the likelihood that the predicted trajectory is the target predicted trajectory.

[0112] In one embodiment, as shown in Figure 4, a flowchart of a trajectory prediction method is provided. The method includes the following steps:

[0113] Step S400: Based on the trajectory information from the trajectory point at the (m - 1)-th moment to the trajectory point at the m-th moment in the first time window, determine the prediction information of the trajectory point at the (m + 1)-th moment.

[0114] Step S410: In the k-th time window among the first to the n-th time windows, based on the trajectory information from the trajectory point at the (m - i + k - 1)-th moment to the trajectory point at the (m + k - 1)-th moment, determine the prediction information of the trajectory point at the (m + k)-th moment, where 1 < k ≤ n and k is an integer;

[0115] The trajectory point at the (m + k - 1)-th moment has j pieces of prediction information. Any one of the j pieces of prediction information includes a predicted position and a probability value corresponding to the predicted position.

[0116] The positions of the trajectory points from the (m - i + k - 1)-th moment to the (m + k - 2)-th moment and the j predicted positions of the trajectory point at the (m + k - 1)-th moment determine j predicted trajectories and probability values; based on any one of the j predicted trajectories, the j predicted positions and probability values of the trajectory point at the (m + k)-th moment can be determined. Among them, the predicted position corresponding to the largest probability value is selected from the j predicted positions of the trajectory point at the (m + k)-th moment as the preferred predicted position of this trajectory point at the (m + k)-th moment.

[0117] Step S420: In the n-th time window, based on the trajectory information from the trajectory point at the (m - i + n - 2)-th moment to the trajectory point at the m-th moment and the prediction information of the trajectory points from the (m + 1)-th moment to the (m + n - 2)-th moment, determine the prediction information of the trajectory point at the (m + n - 1)-th moment.

[0118] Step S430: Determine a target predicted trajectory based on the predicted positions of the trajectory points from the (m + 1)-th moment to the (m + n - 1)-th moment.

[0119] Determine j predicted trajectories and probability values based on the j predicted positions of the trajectory points from the (m + 1)-th moment to the (m + n - 1)-th moment; determine the similarity between any one of the j predicted trajectories and each historical trajectory, and determine the predicted trajectory corresponding to the maximum similarity among the j predicted trajectories as the target predicted trajectory.

[0120] In this way, the trajectory information of the trajectory points at each historical moment and the predicted information of the trajectory points at the predicted moment are fully considered, and the predicted trajectory with the maximum similarity is selected as the target predicted trajectory, ensuring the accuracy of the target predicted trajectory and facilitating the timely tracking and positioning of the target object.

[0121] In some embodiments, obtain the trajectory information of at least one trajectory point of at least one historical trajectory and the trajectory information of at least one trajectory point of the current trajectory;

[0122] Based on the trajectory information of the trajectory points of the historical trajectory and the trajectory information of the trajectory points of the current trajectory, determine the similarity metric value between at least one of the historical trajectories and the current trajectory;

[0123] The determination of the initial trajectory point at the starting moment includes:

[0124] Determine the initial trajectory point of the historical trajectory whose metric value satisfies the second predetermined condition as the initial trajectory point at the starting moment of the first time window.

[0125] Here, the metric value indicates the similarity between the historical trajectory and the current trajectory. Among them, the metric value is negatively correlated with the similarity.

[0126] Here, the second predetermined condition may be that the metric value of the predicted trajectory and the historical trajectory is the smallest or the second smallest. Exemplarily, the metric value of the first predicted trajectory is S4, the metric value of the second predicted trajectory is S5, and the metric value of the third predicted trajectory is S6; if S6 ≤ S4 < S5, then it is determined that the third predicted trajectory corresponding to S6 has the maximum similarity with the historical trajectory, and the third predicted trajectory corresponding to S6 is determined as the target predicted trajectory.

[0127] In some embodiments, the determination of the similarity metric value between at least one of the historical trajectories and the current trajectory based on the trajectory information of the trajectory points of the historical trajectory and the trajectory information of the trajectory points of the current trajectory includes:

[0128] Determine a first parameter based on the time interval between the current moment of the trajectory points of the current trajectory and the historical moments corresponding to the trajectory points of at least one historical trajectory;

[0129] Using dynamic normalization, based on the product of the trajectory information of the trajectory points of the historical trajectory and the trajectory information of the trajectory points of the current trajectory and the first parameter, at least one first metric value of the historical trajectory and the current trajectory is determined; and / or, using distance editing, based on the sum of the trajectory information of the trajectory points of the historical trajectory and the trajectory information of the trajectory points of the current trajectory and the first parameter, at least one second metric value of the historical trajectory and the current trajectory is determined.

[0130] Here, Dynamic Time Warping (DTW) is used to measure the similarity between two time series of different lengths. The unknown quantities are stretched or shortened (compressed) until they match the length of the reference template. During this process, the unknown sequences will be distorted or bent so that their features correspond to the standard pattern.

[0131] Here, the Edit Distance (ED) method is used to transform one string into another string with the minimum number of editing operations required, including adding, deleting, or replacing a character.

[0132] In some embodiments, a dynamic regularization method is used to determine a first metric value based on the product of the trajectory information of the trajectory points of the historical trajectory and the trajectory information of the trajectory points of the current trajectory with the first parameter, including:

[0133] Wherein, Δ indicates the time interval between the current moment corresponding to the trajectory point of the current trajectory and the historical moment corresponding to the trajectory point of the historical trajectory; The first parameter indicates the correlation between different historical trajectories within the Δ time interval; The trajectory information indicating the trajectory points of the current trajectory; The trajectory information of the trajectory points indicating the historical trajectory, where i indicates the historical time of the trajectory point; T pre Indicates the trajectory point of the current trajectory; T i pre Track points indicating historical trajectories; W DTW (D Ti This indicates the first metric comparing the historical trajectory with the current trajectory.

[0134] Exemplarily, an experiment is provided to determine a first metric value of at least one historical trajectory and the current trajectory based on dynamic warping, the experiment including the following steps:

[0135] 1) Preprocess 853 user datasets into user ID, semantic information, latitude and longitude, arrival time and departure time; among them, the user dataset is the weighted trajectory information of trajectory points of historical trajectories.

[0136] 2) Based on the different weighted trajectory information of trajectory points in each user's historical trajectory, the weighted trajectory information of trajectory points with time sequence of 85% of each user's historical trajectory is randomly selected as the dataset, and the weighted trajectory information of trajectory points of the remaining historical trajectory is used as the test set.

[0137] 3) Initialize the weights of the entire network, use stochastic gradient descent with learning rate decay to reduce the loss function, and select 212 hidden layers to minimize the cross-entropy loss.

[0138] 4) Obtaining Experimental Results: The root mean square error (RMSE) was used to evaluate the validity of the experiment. As shown in Figure 5, the following is the average RMSE variation for 853 users. Figure 5 shows that the RMSE gradually decreases as the amount of historical trajectory data increases; the RMSE is the first measure of the difference between the historical trajectory and the current trajectory.

[0139] In some embodiments, a distance editing method is used to determine at least one second metric value between the historical trajectory and the current trajectory, based on the trajectory information of the trajectory points of the historical trajectory and the sum of the trajectory information of the trajectory points of the current trajectory and the first parameter, including:

[0140]

[0141] in, The first parameter indicates the correlation between different historical trajectories within the Δ time interval; The trajectory information indicating the trajectory points of the current trajectory; The trajectory information of the trajectory points indicating the historical trajectory, where i indicates the historical time of the trajectory point; T pre Indicates the trajectory point of the current trajectory; T i pre Track points indicating historical trajectories; W ED (D Ti This indicates the second metric comparing the historical trajectory with the current trajectory.

[0142] Here, the smaller the first and second metrics, the higher the similarity between the trajectory points representing the historical trajectory and the trajectory points of the current trajectory. Weights are assigned to each trajectory point of the historical trajectory based on the similarity. The weights are positively correlated with the similarity.

[0143] In some embodiments, the method further includes:

[0144] The first metric and the second metric corresponding to at least one trajectory point of the historical trajectory are normalized to determine the total metric of at least one trajectory point of the historical trajectory.

[0145] Based on the total metric value, the weights of the trajectory points of the historical trajectory are determined.

[0146] In some embodiments, normalizing the first and second metric values ​​corresponding to at least one trajectory point of the historical trajectory to determine the total metric value of the trajectory points of at least one historical trajectory includes:

[0147] Determine the first mean and first variance of the first metric of the historical trajectory;

[0148] Determine the second mean and second variance of the second metric of the historical trajectory;

[0149] A first difference is determined based on the difference between a first metric value of at least one trajectory point of the historical trajectory and the first mean value;

[0150] The first ratio is determined based on the ratio of the first coefficient to the first variance;

[0151] A second difference is determined based on the difference between the second metric value and the second mean value of the trajectory points of the historical trajectory corresponding to the first metric value.

[0152] The second ratio is determined based on the ratio of the third coefficient to the second variance;

[0153] Based on the sum of the first ratio and the second ratio, the total metric value of the trajectory points of the historical trajectory corresponding to the first metric value is determined.

[0154] For example, the normalization process for the first metric and the second metric corresponding to at least one trajectory point of the historical trajectory includes:

[0155] Among them, W DTW (D Ti This indicates the first metric comparing the historical trajectory with the current trajectory. A second metric indicating the historical trajectory and the current trajectory; μ DWT The first mean of the first metric, μ, indicating the relationship between all historical trajectories and the current trajectory. ED σ represents the second mean of the second metric indicating the relationship between all historical trajectories and the current trajectory. DWT Indicates the first variance, σ, of the first metric between all historical trajectories and the current trajectory. ED The second variance of a second metric indicating the difference between all historical trajectories and the current trajectory; Indicates the total metric value of the historical trajectory and the current trajectory.

[0156] In one embodiment, a weight is assigned to each historical trajectory based on the total metric value of the historical trajectory and the current trajectory. The smaller the total metric value, the greater the weight assigned to the historical trajectory corresponding to the total metric value.

[0157] In some embodiments, the historical trajectory includes: a first historical trajectory and a second historical trajectory;

[0158] The method further includes:

[0159] Obtain a metric value for each of the first historical trajectories and at least one of the second historical trajectories;

[0160] If the metric value is less than or equal to the first threshold, the second historical trajectory corresponding to the metric value is determined to be a valid trajectory.

[0161] The measurement value is determined based on the number of times the trajectory information of at least one trajectory point of the first historical trajectory overlaps with the trajectory information of the trajectory points of the effective trajectory;

[0162] If the measured value is greater than or equal to the second threshold, the trajectory point of the first historical trajectory corresponding to the measured value is determined as the trajectory point for determining the predicted trajectory.

[0163] Here, the first historical trajectory and the second historical trajectory can be any historical trajectory.

[0164] In some embodiments, the first threshold may be determined based on the user's historical experience.

[0165] For example, if the range of the first threshold is 0 to 1, the first threshold can be 0.2; if the range of the first threshold is 0 to 10, the first threshold can be 15.

[0166] In some embodiments, the second threshold may be determined based on the user's historical experience, or it may be determined based on the average number of times the trajectory information of at least one trajectory point of the first historical trajectory overlaps with the trajectory information of the trajectory points of the effective trajectory.

[0167] For example, the average number of times the trajectory information of at least one trajectory point of the first historical trajectory overlaps with the trajectory information of the trajectory point of the effective trajectory is 10, and the second threshold can be 8, 9, 10, etc.

[0168] In some embodiments, the first historical trajectory is an arbitrarily selected historical trajectory, and the second historical trajectory is each historical trajectory within half a month before and after the corresponding historical time of the first historical trajectory. A metric value is obtained between each first historical trajectory and at least one second historical trajectory; wherein, the second historical trajectory corresponding to a metric value less than or equal to a first threshold is a valid trajectory. The measurement value is determined based on the number of times the position information of each trajectory point in the first historical trajectory overlaps with the position information of each trajectory point in the valid trajectory; wherein, the trajectory points of the first historical trajectory corresponding to measurement values ​​greater than or equal to the second threshold are valid trajectory points and can be retained; the trajectory points of the first historical trajectory corresponding to measurement values ​​less than the second threshold are error trajectory points and need to be discarded. The valid trajectory points are input into the prediction model as trajectory points to determine the predicted trajectory, thereby obtaining the prediction information of the trajectory points of the predicted trajectory; wherein, the trajectory points to determine the predicted trajectory include all trajectory points from the initial trajectory point to the current time of the current trajectory.

[0169] In this way, valid trajectory points in historical trajectories can be retained, while error trajectory points with fewer overlaps can be eliminated, thereby improving the accuracy of predicted trajectories.

[0170] The following provides a specific example in conjunction with any of the above embodiments:

[0171] In one application scenario, as shown in Figure 6, a trajectory prediction method is provided for use on a terminal. The method includes:

[0172] Step S610: Obtain the trajectory information of the trajectory points of the historical trajectory and the trajectory information of the trajectory points of the current trajectory;

[0173] Here, the trajectory information of the trajectory points of the historical trajectory includes: the trajectory information from the initial trajectory point at the starting time to the trajectory point at the previous time of the current trajectory.

[0174] Step S620: Based on the trajectory information of trajectory points of historical trajectories and the trajectory information of trajectory points of the current trajectory, determine the measurement value of at least one trajectory point of the historical trajectory and the trajectory point of the current trajectory.

[0175] Based on the dynamic regularization method and / or edit distance method, determine the metric value between the trajectory point of at least one historical trajectory and the trajectory point of the current trajectory; assign weights to the trajectory points of the historical trajectory based on the metric value to determine the weighted trajectory information of the trajectory points of the historical trajectory; wherein, the larger the metric value, the smaller the assigned weight.

[0176] Step S630: Determine the valid trajectory points among the trajectory points of the historical trajectory based on the metric values;

[0177] Here, the historical trajectory includes: a first historical trajectory and a second historical trajectory; obtaining the measurement value of a first historical trajectory and at least one second historical trajectory respectively; if the measurement value is less than or equal to 0.3, the second historical trajectory corresponding to the measurement value is determined as a valid trajectory; based on the number of times the position information of at least one trajectory point of the first historical trajectory coincides with the position information of the trajectory point of the valid trajectory, the measurement value is determined; if the measurement value is greater than or equal to 50, the trajectory point of the first historical trajectory corresponding to the measurement value is determined as a valid trajectory point.

[0178] Step S640: Determine the prediction information of the trajectory points at the next time step in the first time window;

[0179] Using a recurrent neural network model, the starting time within the first time window is the mi-th time, and the current time within the first time window is the m-th time; where m and i are both integers greater than 0; where m is greater than i; based on the trajectory information from the trajectory point at time m-1 to the trajectory point at time m, the predicted information for the trajectory point at time m+1 is determined.

[0180] Step S650: Determine the prediction information of the trajectory points for the next time step in the second to nth time windows;

[0181] Using a recurrent neural network model, the starting time of the kth time window from the 2nd to the nth time window is the (m-i+k-1)th time window; the current time of the kth time window is the (m+k-1)th time window; where k is an integer; based on the trajectory information from the trajectory point at the (m-i+k-1)th time window to the trajectory point at the (m+k-1)th time window, the prediction information of the trajectory point at the (m+k)th time window is determined.

[0182] Step S660: Determine the target predicted trajectory based on the predicted information of the trajectory points of the next time step of the current time step in the first to nth time windows;

[0183] Based on the j predicted positions of the trajectory points from time m+1 to time m+n-1, j predicted trajectories and probability values ​​are determined; based on the similarity between any one of the j predicted trajectories and each historical trajectory, the predicted trajectory with the highest similarity among the j predicted trajectories is determined as the target predicted trajectory.

[0184] As shown in Figure 7, based on the same inventive concept as the trajectory prediction method provided in the foregoing embodiments, this application also provides a trajectory prediction device 700, the device 700 comprising:

[0185] The first determining module 701 is used to determine the initial trajectory point at the start time;

[0186] The calculation module 702 is used to input the weighted trajectory information of the trajectory points from the initial trajectory point to the current moment of the current trajectory into the prediction model within a predetermined time window, so as to obtain the prediction information of the trajectory points at the next moment of the current moment of the current trajectory; wherein, the weighted trajectory information is determined based on the trajectory information of the trajectory points and the weights corresponding to the trajectory points; the trajectory points at the next moment include the predicted positions of at least one predicted trajectory point; the prediction information of the trajectory points at the next moment includes at least the trajectory information;

[0187] The second determining module 703 is used to determine at least one predicted trajectory based on the weighted trajectory information of the trajectory point at the current time and the prediction information of the predicted trajectory point at the next time.

[0188] The third determining module 704 is used to determine, based on at least one of the predicted trajectories, the predicted trajectory that satisfies a first predetermined condition as the target predicted trajectory.

[0189] In some embodiments, the third determining module 704 is configured to determine the total similarity of the predicted trajectory based on the similarity between at least one predicted trajectory and at least one historical trajectory.

[0190] The predicted trajectory corresponding to the total similarity that satisfies the first predetermined condition is determined as the target predicted trajectory.

[0191] In some embodiments, the prediction information of the trajectory point at the next moment further includes the probability value;

[0192] The third determining module 704 is used to determine, based on at least one probability value corresponding to the predicted trajectory, the predicted trajectory corresponding to the probability value that satisfies the first predetermined condition as the target predicted trajectory.

[0193] In some embodiments, the predetermined time window includes time windows 1 to n; the starting time of the first time window is the mi-th time, and the current time of the first time window is the m-th time; n, m, and i are all integers greater than 0; m is greater than i.

[0194] The calculation module 702 is used to determine the prediction information of the trajectory point at time m+1 based on the trajectory information from the trajectory point at time m-1 to the trajectory point at time m.

[0195] In some embodiments, the starting time of the kth time window among the 2nd to nth time windows is the (m-i+k-1)th time window; the current time of the kth time window is the (m+k-1)th time window; where k is an integer;

[0196] The calculation module 702 is used to determine the prediction information of the trajectory point at the (m+k)th time based on the trajectory information from the trajectory point at the (m-i+k-1)th time to the trajectory point at the (m+k-1)th time.

[0197] In some embodiments, the device 700 includes:

[0198] The acquisition module is used to acquire trajectory information of at least one trajectory point of at least one historical trajectory and trajectory information of at least one trajectory point of the current trajectory.

[0199] The fourth determining module is used to determine at least one similarity metric between the historical trajectory and the current trajectory based on the trajectory information of the trajectory points of the historical trajectory and the trajectory information of the trajectory points of the current trajectory.

[0200] The first determining module 701 is used to determine the initial trajectory point of the historical trajectory whose metric value satisfies the second predetermined condition, as the initial trajectory point of the starting time of the first time window.

[0201] In some embodiments, the fourth determining module is used to determine a first parameter based on the time interval between the current time of the trajectory point of the current trajectory and the historical time corresponding to the trajectory point of at least one historical trajectory.

[0202] Using dynamic normalization, based on the product of the trajectory information of the trajectory points of the historical trajectory and the trajectory information of the trajectory points of the current trajectory with the first parameter, at least one first metric value of the historical trajectory and the current trajectory is determined; and / or, using distance editing, based on the sum of the trajectory information of the trajectory points of the historical trajectory and the trajectory information of the trajectory points of the current trajectory with the first parameter, at least one second metric value of the historical trajectory and the current trajectory is determined. In some embodiments, the historical trajectory includes: a first historical trajectory and a second historical trajectory;

[0203] The device 700 further includes:

[0204] The fourth determining module is used to obtain the measurement values ​​of a first historical trajectory and at least one second historical trajectory respectively;

[0205] The judgment module is used to determine that the second historical trajectory corresponding to the metric value is a valid trajectory if the metric value is less than or equal to the first threshold.

[0206] The fourth determining module is used to determine the measurement value based on the number of times the trajectory information of at least one trajectory point of the first historical trajectory overlaps with the trajectory information of the trajectory points of the effective trajectory.

[0207] The judgment module is further configured to determine the trajectory point of the first historical trajectory corresponding to the measurement value as the trajectory point for determining the predicted trajectory if the measured value is greater than or equal to the second threshold.

[0208] In practical applications, the calculation module, first determination module, second determination module, and third determination module of the trajectory prediction device can be implemented by the processor in the trajectory prediction device. Of course, the processor needs to run the computer program in the memory to implement its functions.

[0209] As shown in Figure 8, this application embodiment provides an electronic device, the electronic device comprising:

[0210] Memory 801 is used to store computer-readable instructions;

[0211] The processor 802, connected to the memory, is configured to implement the methods provided in any of the foregoing embodiments by executing computer-readable instructions.

[0212] The memory 801 can be of various types, such as random access memory, read-only memory, flash memory, etc. The memory can be used for information storage, for example, storing computer-executable instructions. The computer-executable instructions can be various program instructions, such as object program instructions and / or source program instructions.

[0213] The processor 802 can be various types of processors, such as a central processing unit, microprocessor, digital signal processor, programmable array, application-specific integrated circuit, or image processor. The processor can be connected to the memory via a bus, which can be an integrated circuit bus, etc.

[0214] As shown in Figure 8, the electronic device may also include a network interface 803, which can be used to interact with peer devices via a network.

[0215] This application also provides a computer storage medium storing computer-executable instructions, which, when executed, can implement the methods provided in any of the foregoing embodiments.

[0216] The computer-readable storage medium provided in the embodiments of this application can be any storage medium capable of storing program code, such as ROM, PROM, EPROM, EEPROM, Flash Memory, magnetic surface memory, optical disc, or CD-ROM.

[0217] In the embodiments provided in this application, it should be understood that the disclosed methods and apparatus can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple modules or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be indirect coupling or communication connection through some interfaces, devices, or units, and can be electrical, mechanical, or other forms.

[0218] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0219] In addition, each functional unit in the various embodiments of this application can be integrated into one processing module, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0220] It should be noted that terms such as "first" and "second" are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0221] Furthermore, the technical solutions described in the embodiments of this application can be combined arbitrarily without conflict.

[0222] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A trajectory prediction method, characterized in that, The method includes: determining an initial trajectory point at a starting time; within a predetermined time window, inputting weighted trajectory information from the initial trajectory point to the current trajectory point at the current time into a prediction model to obtain prediction information for the trajectory point at the next time step of the current trajectory; wherein, the weighted trajectory information is determined based on the trajectory information of the trajectory point and the weight corresponding to the trajectory point; the weight corresponding to the trajectory point is the weight corresponding to a historical trajectory point determined based on the similarity between the current trajectory point of the target object and the trajectory points of historical trajectories; the trajectory point at the next time step includes the predicted position of at least one predicted trajectory point; the prediction information for the trajectory point at the next time step includes at least the trajectory information; and the weighted trajectory information based on the trajectory point at the current time step... The information is compared with the prediction information of the predicted trajectory points at the next time moment to determine at least one predicted trajectory; based on at least one predicted trajectory, the predicted trajectory that satisfies a first predetermined condition is determined as the target predicted trajectory; the method further includes: acquiring trajectory information of at least one trajectory point of at least one historical trajectory and trajectory information of at least one trajectory point of the current trajectory; based on the trajectory information of the trajectory points of the historical trajectory and the trajectory information of the trajectory points of the current trajectory, determining a metric value of at least one historical trajectory and the current trajectory; determining the initial trajectory point at the start time includes: determining the initial trajectory point of the historical trajectory whose metric value satisfies a second predetermined condition, as the initial trajectory point at the start time of the first time window.

2. The method according to claim 1, characterized in that, The step of determining the predicted trajectory that satisfies a first predetermined condition as the target predicted trajectory based on at least one of the predicted trajectories includes: determining the total similarity of the predicted trajectories based on the similarity between at least one of the predicted trajectories and at least one historical trajectory; and determining the predicted trajectory corresponding to the total similarity that satisfies the first predetermined condition as the target predicted trajectory.

3. The method according to claim 1, characterized in that, The prediction information of the trajectory point at the next moment also includes a probability value; the step of determining the predicted trajectory that satisfies the first predetermined condition as the target predicted trajectory based on at least one of the predicted trajectories includes: determining the predicted trajectory corresponding to the probability value that satisfies the first predetermined condition as the target predicted trajectory based on the probability value corresponding to at least one of the predicted trajectories.

4. The method according to claim 1, characterized in that, The predetermined time window includes time windows 1 to n; the starting time of the first time window is the mi-th time, and the current time of the first time window is the m-th time; n, m, and i are all integers greater than 0; m is greater than i; the method further includes: determining the prediction information of the trajectory point at the (m-1)-th time based on the trajectory information from the trajectory point at the (m-1)-th time to the trajectory point at the m-th time.

5. The method according to claim 1 or 4, characterized in that, The starting time of the kth time window from the 2nd to the nth time window is the (m-i+k-1)th time window; the current time of the kth time window is the (m+k-1)th time window; where k is an integer; the method further includes: determining the prediction information of the trajectory point at the (m+k)th time window based on the trajectory information from the trajectory point at the (m-i+k-1)th time window to the trajectory point at the (m+k-1)th time window.

6. The method according to claim 4, characterized in that, The step of determining at least one metric value between the historical trajectory and the current trajectory based on the trajectory information of the trajectory points of the historical trajectory and the trajectory information of the current trajectory includes: determining a first parameter based on the time interval between the current time of the trajectory point of the current trajectory and the historical time corresponding to the trajectory point of at least one historical trajectory; using dynamic normalization, determining a first metric value between at least one historical trajectory and the current trajectory based on the product of the trajectory information of the trajectory points of the historical trajectory and the trajectory information of the trajectory points of the current trajectory and the first parameter; and / or, using distance editing, determining a second metric value between at least one historical trajectory and the current trajectory based on the sum of the trajectory information of the trajectory points of the historical trajectory and the trajectory information of the trajectory points of the current trajectory and the first parameter.

7. The method according to claim 4 or 6, characterized in that, The historical trajectory includes a first historical trajectory and a second historical trajectory; the method further includes: obtaining a measurement value of a first historical trajectory and at least one second historical trajectory respectively; if the measurement value is less than or equal to a first threshold, determining the second historical trajectory corresponding to the measurement value as a valid trajectory; determining a measurement value based on the number of times the trajectory information of at least one trajectory point of the first historical trajectory coincides with the trajectory information of the trajectory point of the valid trajectory; if the measurement value is greater than or equal to a second threshold, determining the trajectory point of the first historical trajectory corresponding to the measurement value as the trajectory point for determining the predicted trajectory.

8. A trajectory prediction device, characterized in that, The device includes: a first determining module, configured to determine an initial trajectory point at a starting time; a calculation module, configured to input weighted trajectory information from the initial trajectory point to the current trajectory point at the current time into a prediction model within a predetermined time window, to obtain prediction information of the trajectory point at the next time step of the current trajectory; wherein the weighted trajectory information is determined based on the trajectory information of the trajectory point and the weight corresponding to the trajectory point; the weight corresponding to the trajectory point is the weight corresponding to a historical trajectory point determined based on the similarity between the current trajectory point of the target object and the trajectory points of historical trajectories; the trajectory point at the next time step includes the predicted position of at least one predicted trajectory point; the prediction information of the trajectory point at the next time step includes at least the trajectory information; a second determining module, configured to determine the trajectory point at the current time step... The weighted trajectory information is compared with the prediction information of the predicted trajectory point at the next time moment to determine at least one predicted trajectory; the third determining module is used to determine the predicted trajectory that satisfies the first predetermined condition as the target predicted trajectory based on at least one predicted trajectory; the acquisition module is used to acquire the trajectory information of at least one trajectory point of at least one historical trajectory and the trajectory information of at least one trajectory point of the current trajectory; the fourth determining module is used to determine the measurement value of at least one historical trajectory and the current trajectory based on the trajectory information of the trajectory points of the historical trajectory and the trajectory information of the trajectory points of the current trajectory; the first determining module is also used to determine the initial trajectory point of the historical trajectory whose measurement value satisfies the second predetermined condition, as the initial trajectory point of the starting time of the first time window.

9. An electronic device, characterized in that, include: A memory storing computer-readable instructions; a processor connected to the memory, configured to implement the trajectory prediction method provided in any one of claims 1 to 7 by executing the computer-readable instructions.

10. A computer storage medium, characterized in that, The computer storage medium stores computer-executable instructions; when the computer-executable instructions are executed by the processor, they can implement the trajectory prediction method according to any one of claims 1 to 7.

Images