Destination prediction method, device, electronic device and storage medium

Abstract

The present invention provides a destination prediction method, device, electronic device and storage medium, wherein the method comprises: obtaining driving information of a vehicle to be predicted; estimating the driving intention of the vehicle to be predicted based on the driving information, and performing trajectory prediction based on the driving information and the driving intention to obtain an estimated destination of the vehicle to be predicted. The method, device, electronic device and storage medium provided by the present invention estimate the driving intention through the driving information of the vehicle, thereby combining the driving information and the driving intention to predict the destination; the combination of driving intention can guide the direction of travel of the estimated trajectory, thereby ensuring that the final estimated trajectory is consistent with the user's driving habits while also being able to fit the user's driving intention, thereby greatly improving the reliability and accuracy of destination prediction in non-navigation scenarios.

Description

Technical Field

[0001] The present invention relates to the field of intelligent driving technology, and in particular to a destination prediction method, device, electronic device and storage medium. Background Art

[0002] In the navigation system of the related art, the user is required to set the destination manually or by voice input, and then the navigation system automatically searches to obtain the required navigation path.

[0003] In actual applications, users usually do not turn on navigation when traveling to places they often go via familiar routes, such as daily commuting, picking up children from school, etc. Therefore, the in-vehicle intelligent central control cannot obtain the destination of this trip through the navigation system for business recommendations. How to predict the destination of the trip and promote business without using the destination set by the user in the navigation system is still a problem that needs to be studied and solved urgently. Summary of the invention

[0004] The present invention provides a destination prediction method, device, electronic device and storage medium, which are used to solve the defect that the destination prediction cannot be realized in a non-navigation scenario in the prior art.

[0005] The present invention provides a destination prediction method, comprising:

[0006] Obtain driving information of the vehicle to be predicted;

[0007] The driving intention of the vehicle to be predicted is estimated based on the driving information, and the trajectory is estimated based on the driving information and the driving intention to obtain an estimated destination of the vehicle to be predicted.

[0008] According to a destination prediction method provided by the present invention, the estimating the driving intention of the vehicle to be predicted based on the driving information includes:

[0009] Determining the intention regularity information at the current moment based on the driving information at the current moment and the intention regularity information at the previous moment before the current moment;

[0010] Based on the intention regularity information at the current moment, the driving intention at the current moment is predicted.

[0011] According to a destination prediction method provided by the present invention, the method of predicting the driving intention at the current moment based on the intention regularity information at the current moment includes:

[0012] Determining the driving intention corresponding to the intention law information at the current moment based on the relationship between the historical intention law information and the driving intention;

[0013] The relationship between the historical intention regularity information and the driving intention is determined based on the historical driving information of the user who is currently driving the vehicle to be predicted.

[0014] According to a destination prediction method provided by the present invention, the trajectory prediction is performed based on the driving information and the driving intention to obtain the estimated destination of the vehicle to be predicted, including:

[0015] Performing trajectory matching on the driving information to obtain a matching probability that each historical driving trajectory of the vehicle to be predicted is a corresponding estimated trajectory of the driving information;

[0016] Based on the matching probability of each historical driving trajectory and the driving intention, the estimated trajectory is determined, and the destination of the estimated trajectory is determined as the estimated destination.

[0017] According to a destination prediction method provided by the present invention, performing trajectory matching on the driving information to obtain the matching probability of each historical driving trajectory of the vehicle to be predicted being the estimated trajectory corresponding to the driving information includes:

[0018] Determine the trajectory characteristics of the current moment based on the driving information and driving intention at the current moment, and the driving intention of the driving information at each moment before the current moment;

[0019] Based on the trajectory features of each historical driving trajectory, a matching probability between the trajectory features at the current moment and each historical driving trajectory is determined.

[0020] According to a destination prediction method provided by the present invention, the trajectory characteristics at the current moment are determined based on the driving information and driving intention at the current moment, as well as the driving intention of the driving information at each moment before the current moment, including:

[0021] Based on the driving information at the current moment and the driving information at each moment before the current moment, determining the trajectory regularity information at the current moment;

[0022] Determining the trajectory code at the current moment based on the trajectory regularity information at the current moment and the driving intention;

[0023] Based on the trajectory code at the current moment and the trajectory feature at a moment before the current moment, the trajectory feature at the current moment is determined.

[0024] According to a destination prediction method provided by the present invention, the method of estimating the driving intention of the vehicle to be predicted based on the driving information, and performing trajectory estimation based on the driving information and the driving intention to obtain the estimated destination of the vehicle to be predicted includes:

[0025] The driving information is input into a dedicated destination prediction model of a user who is currently driving the vehicle to be predicted, and the dedicated destination prediction model estimates the driving intention based on the driving information, and performs trajectory prediction based on the driving information and the driving intention, to obtain an estimated destination output by the dedicated destination prediction model;

[0026] The exclusive destination prediction model is obtained by performing transfer learning on the general trajectory prediction model based on the historical driving information and historical destinations of the user.

[0027] According to a destination prediction method provided by the present invention, the driving information includes driving trajectory information, or includes the driving trajectory information and driving environment information.

[0028] According to a destination prediction method provided by the present invention, the estimated destination of the vehicle to be predicted is obtained, and then the method further includes:

[0029] Acquiring historical business data of the estimated destination;

[0030] Based on the driving information, selecting a service to be recommended from historical service data;

[0031] Generate and send a recommendation reminder for the service to be recommended.

[0032] The present invention also provides a destination prediction device, comprising:

[0033] An information acquisition unit, used to acquire driving information of a vehicle to be predicted;

[0034] The destination estimation unit is used to estimate the driving intention of the vehicle to be predicted based on the driving information, and to perform trajectory estimation based on the driving information and the driving intention to obtain the estimated destination of the vehicle to be predicted.

[0035] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any of the above-mentioned destination prediction methods when executing the computer program.

[0036] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any of the above-mentioned destination prediction methods are implemented.

[0037] The destination prediction method, device, electronic device and storage medium provided by the present invention estimate the driving intention through the driving information of the vehicle, and thus predict the destination in combination with the driving information and the driving intention; the combination of the driving intention can guide the direction of the estimated trajectory, thereby ensuring that the final estimated trajectory conforms to the user's driving habits while also being able to fit the user's driving intention, thereby greatly improving the reliability and accuracy of destination prediction in non-navigation scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly described below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 It is a flowchart of the destination prediction method provided by the present invention;

[0040] Figure 2 is a flowchart of step 120 in the destination prediction method provided by the present invention;

[0041] Figure 3 is a flowchart of step 121 in the destination prediction method provided by the present invention;

[0042] Figure 4 is a flowchart of step 122 in the destination prediction method provided by the present invention;

[0043] Figure 5 is a flowchart of step 1221 in the destination prediction method provided by the present invention;

[0044] Figure 6 It is a flowchart of step 1221-1 in the destination prediction method provided by the present invention;

[0045] Figure 7 It is a structural schematic diagram of the destination estimation model provided by the present invention;

[0046] Figure 8 It is a structural schematic diagram of the closed-loop recommendation system provided by the present invention;

[0047] Fig. 9 It is a structural schematic diagram of the destination prediction device provided by the present invention;

[0048] Fig.10 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] With the rapid development of the smart car industry, the in-car smart central control has become the link between human and machine interaction. The closed loop of services and accurate business recommendations beyond navigation have also become the direction and research hotspot of user expectations and industry efforts.

[0051] Currently, the service loop and business recommendations are mostly of the following two types: one is to recommend routes for various situations based on the start and end points set by the user, such as avoiding congestion, the least traffic lights, the least time, the best parking route, etc.; the other is to match the route representation of historical routes based on the start and end points preset by the user to infer the recommendation probability result of the user's next request, so as to recommend routes that are more in line with the user's habits. In either case, it is achieved on the premise that the user turns on the navigation and sets the start and end points.

[0052] However, in daily operations, users usually do not turn on navigation when traveling to places they often go via familiar routes, such as daily commuting, picking up children from school, etc. Therefore, the in-vehicle intelligent central control cannot obtain the destination of this trip through the navigation system to close the service loop and make business recommendations. To address this problem, the current solution is to combine the user's current vehicle location with the user's previous navigation information to estimate the destination, but this method only involves matching historical destinations, and its reliability and accuracy cannot be guaranteed.

[0053] In view of the above problems, the present invention provides a destination prediction method for realizing destination prediction in non-navigation scenarios. Figure 1 is a flow chart of the destination prediction method provided by the present invention, such as Figure 1 As shown, the execution subject of the method can be an in-vehicle intelligent central control, or a smart device connected to the in-vehicle intelligent network, such as a smart phone, a tablet computer, etc.

[0054] The method includes:

[0055] Step 110, obtaining driving information of the vehicle to be predicted.

[0056] Here, the vehicle to be predicted is the vehicle whose destination needs to be predicted. The driving information of the vehicle to be predicted refers to the information after the vehicle is started. It can be the information from the time the vehicle is started to the current time, or it can be the information calculated from the current time to the preset time.

[0057] Driving information may specifically include the vehicle's driving trajectory information, such as the location, time, vehicle head direction, and road of each trajectory point during the vehicle's driving, as well as the mileage and time taken to reach each trajectory point from the starting point. In addition, driving information may also include environmental information during the vehicle's driving process, which may be the vehicle's internal environment or the external environment, such as the number of people in the vehicle, their identities, etc., as well as the degree of congestion on the road where the vehicle is located, and surrounding buildings and landmarks on the road where the vehicle is located.

[0058] Step 120 , estimating the driving intention of the vehicle to be predicted based on the driving information, and performing trajectory estimation based on the driving information and the driving intention to obtain an estimated destination of the vehicle to be predicted.

[0059] Specifically, after obtaining driving information, a commonly used method is to directly estimate the destination based on the driving information, such as matching the driving information with historical driving information, and directly determining the destination corresponding to the matched historical driving information as the estimated destination of the current trip, or learning the mapping relationship between driving information and destination from historical driving information and its corresponding destination, so as to directly obtain the estimated destination corresponding to the driving information of the current trip. Although the above method can achieve simple prediction of the destination, in real life, there is not a simple mapping relationship between driving information and destination, where the driving information contains the complex intentions of the user who is currently driving the vehicle to be predicted. Predicting the destination based on relationship mapping alone does not conform to the actual situation, and the accuracy and reliability are very poor.

[0060] Taking the actual situation into consideration, when an embodiment of the present invention predicts the destination based on driving information, it first estimates the driving intention of the vehicle to be predicted. The driving intention here reflects the current driving direction of the vehicle, such as turning left, turning right, or going straight. Specifically, it can be the probability of the vehicle executing a left turn, a right turn, and execution.

[0061] Furthermore, when estimating driving intention based on driving information, the law of driving intention reflected by the driving information in the historical driving process of the vehicle can be combined to estimate the driving intention of the current vehicle to be predicted in accordance with the law. For example, the driving information in the historical driving process can be used as a training sample, and the trained neural network model can be used to train the trained neural network model so that the law of driving intention reflected in the driving information can be learned. When the driving intention is subsequently estimated, the driving information can be directly input into the neural network model for prediction. For example, the driving intention probability distribution of the driving information in the historical driving process can be calculated by mathematical statistics, and the probability distribution law of the driving intention can be applied to the subsequent driving intention estimation. The embodiment of the present invention does not make specific limitations on this.

[0062] After obtaining the driving intention of the vehicle to be predicted, the trajectory of the vehicle to be predicted can be estimated based on its driving information and driving intention, and the estimated destination of the trip can be determined based on the estimated trajectory. Specifically, in the process of trajectory estimation, the driving intention can guide the direction of the estimated trajectory based on the driving information, so that the trajectory estimation can refer to the complex intentions of the user who is currently driving the vehicle to be predicted, and further fit the actual application situation. The final estimated trajectory can not only meet the user's driving habits and historical behaviors, but also fit the user's current driving intention, thereby improving the reliability and accuracy of trajectory prediction.

[0063] Furthermore, when performing trajectory estimation based on driving information and driving intention, the driving information, driving intention and historical trajectory including the destination in the historical driving process are used as training samples. By training the neural network model, the trained neural network model can learn the rules of the driving trajectory reflected by the driving information and driving intention. When performing trajectory estimation subsequently, the driving information and driving intention can be directly input into the neural network model for prediction. For example, the probability distribution of the frequently visited destinations of the driving information in the historical driving process can be calculated by mathematical statistics in combination with the driving intention, and the probability distribution rules of the frequently visited destinations can be applied to the subsequent trajectory estimation. The embodiments of the present invention do not make specific limitations on this.

[0064] In addition, the above-mentioned steps of estimating driving intention based on driving information, and estimating driving trajectory based on driving information and driving intention to finally determine the destination can also be implemented through an end-to-end neural network model. The driving information, driving intention and historical trajectory including the destination in the historical driving process can be used as training samples. By training the neural network model, the trained end-to-end neural network model can learn the rules of driving intention reflected in the driving information, as well as the rules of driving trajectory reflected by the driving information and driving intention. Therefore, when predicting the candidate destination, it is only necessary to input the driving information into the end-to-end neural network model, so that the end-to-end neural network model can realize the estimation of driving intention and the estimation of driving trajectory based on driving intention.

[0065] The neural network model used here to implement trajectory prediction or intent estimation can be implemented based on a neural network architecture with time series data processing capabilities, such as RNN (Recurrent Neural Network) and LSTM (Long Short-Term Memory). In addition, the neural network model used to implement trajectory prediction or intent estimation can also be an n-gram (N-gram model), which can be combined with the viterbi algorithm to find the top N results with the highest matching model scores as the pre-judgment and recommendation results.

[0066] The method provided by the embodiment of the present invention estimates the driving intention through the driving information of the vehicle, and thus predicts the destination by combining the driving information and the driving intention; the combination of driving intention can guide the direction of travel of the estimated trajectory, thereby ensuring that the final estimated trajectory is consistent with the user's driving habits while also being able to fit the user's driving intention, thereby greatly improving the reliability and accuracy of destination prediction in non-navigation scenarios.

[0067] Based on the above embodiments, Figure 2 is a flow chart of step 120 in the destination prediction method provided by the present invention, such as Figure 2 As shown, step 120 includes:

[0068] Step 121, estimating the driving intention of the vehicle to be predicted based on the driving information;

[0069] Step 122 , performing trajectory estimation based on driving information and driving intention to obtain an estimated destination of the vehicle to be estimated.

[0070] On this basis, Figure 3 is a flow chart of step 121 in the destination prediction method provided by the present invention, such as Figure 3 As shown, step 121 includes:

[0071] Step 1211, based on the driving information at the current moment and the intention regularity information at the previous moment before the current moment, determine the intention regularity information at the current moment.

[0072] Step 1212, predicting the driving intention at the current moment based on the intention rule information at the current moment.

[0073] Specifically, driving information can be divided based on the time of collection. The driving information at each time can be used to reflect the information of the vehicle to be predicted at the corresponding time, such as the vehicle position, the direction of the front of the vehicle, the road where the vehicle is located, the speed at the corresponding time, the mileage and driving time at the corresponding time, the number of people in the car at the corresponding time, the identity of the people, the congestion level of the road where the vehicle is located, surrounding buildings, landmarks, etc.

[0074] In order to realize intention recognition based on driving information, the intention pattern of the vehicle to be predicted can be statistically analyzed based on the driving information at each moment. Here, the intention pattern information can reflect the driving pattern during driving, such as the pattern of turning, changing lanes, and going straight. Specifically in step 1211, the driving information is sorted according to the corresponding moment, and the intention pattern information of the current moment is obtained by integrating the driving information at each moment. For the first moment, the intention pattern information of the first moment can be determined based on the driving information at the first moment; for each moment after the first moment, the intention pattern information of the corresponding moment can be determined based on the driving information at the corresponding moment and the intention pattern information of the previous moment of the corresponding moment, so that the intention pattern information of each moment can not only reflect the driving information at the corresponding moment, but also cover the intention pattern information of each moment before the corresponding moment.

[0075] Here, the determination of the intention regularity information can be achieved by a pre-set acquisition rule or a pre-trained neural network model. The first moment can be the moment of the preset time period forward when the intention recognition is performed at the current moment, or it can be the starting moment of the trip, which is not specifically limited in the embodiment of the present invention. For example, the first moment can be set to 1 minute after the start of the trip.

[0076] Based on any of the above embodiments, step 1212 includes:

[0077] Based on the relationship between the historical intention regularity information and the driving intention, determining the driving intention corresponding to the intention regularity information at the current moment;

[0078] The relationship between the historical intention regularity information and the driving intention is determined based on the historical driving information of the user who is currently driving the vehicle to be predicted.

[0079] Specifically, after obtaining the intention rule information at the current moment, the intention rule information at the current moment can be combined with the relationship between the historical intention rule information of the vehicle to be predicted at each moment in the historical driving process and the driving intention to determine the probability distribution of the intention rule information at the current moment for each driving intention. For example, the probability of turning left is 72%, the probability of going straight is 13%, and the probability of turning right is 15%. Therefore, it can be determined that the driving intention at the current moment is to turn left.

[0080] In particular, considering that different people have different driving habits when driving a vehicle, and different people may be familiar with different routes and frequently visited destinations, the embodiment of the present invention specifically distinguishes users when training the relationship between historical intention regularity information and driving intention, that is, for the historical driving information of different users as drivers, the relationship between historical intention regularity information and driving intention corresponding to each user is trained respectively. For any user as a driver, the intention regularity information can be extracted from the historical driving information of the user's multiple trips, and the historical driving information at each moment in multiple time periods can be intercepted, and the driving intention can be marked, thereby obtaining the intention regularity information and driving intention at each moment as training samples and labels, and training is performed on this basis to obtain the relationship between the user's historical intention regularity information and driving intention.

[0081] Based on any of the above embodiments, Figure 4 is a flow chart of step 122 in the destination prediction method provided by the present invention, such as Figure 4 As shown, step 122 includes:

[0082] Step 1221 , performing trajectory matching on the driving information, and obtaining the matching probability of each historical driving trajectory of the vehicle to be predicted being the estimated trajectory corresponding to the driving information.

[0083] Specifically, the historical driving trajectory may be the driving trajectory of the vehicle previously driven by the user who is currently driving the vehicle to be predicted. The selection of the historical driving trajectory may be based on the number of times the user has driven the vehicle along the driving trajectory within a period of time. For example, if a user goes from home to work every day within a week, the driving trajectory from home to work may be used as the user's historical driving trajectory. For another example, if a user goes to a shopping mall after get off work two days a week, the driving trajectory from the company to the mall may be used as the user's historical driving trajectory.

[0084] After the historical driving trajectories of the user are collected, the travel information under each historical travel trajectory can be obtained for the subsequent matching of driving information and calculation of distribution probability.

[0085] When performing trajectory estimation in combination with driving information and driving intention, the driving information may be first matched with the historical driving information corresponding to each historical driving trajectory, thereby obtaining a matching probability that each historical driving trajectory is the estimated trajectory corresponding to the driving information.

[0086] Step 1222: Based on the matching probability of each historical driving trajectory and the driving intention, an estimated trajectory is determined, and the destination of the estimated trajectory is determined as the estimated destination.

[0087] Specifically, compared with simply applying driving information to match historical driving trajectories to estimate trajectories, the reliability of the estimated trajectory can be further improved by combining the driving intention to guide the direction of the estimated trajectory. For example, the direction of travel of each historical driving trajectory can be analyzed, and it can be determined whether the direction of travel of each historical driving trajectory is consistent with the direction of travel indicated by the driving intention. If they are consistent, the matching probability of the historical driving trajectory is increased. If they are inconsistent, the matching probability of the historical driving trajectory is weakened. On this basis, according to the matching probability of each historical driving trajectory, a trajectory is selected from each historical driving trajectory as the estimated trajectory, and the destination of the estimated trajectory is determined as the estimated destination.

[0088] Based on any of the above embodiments, Figure 5 is a flow chart of step 1221 in the destination prediction method provided by the present invention, such as Figure 5 As shown, step 1221 includes:

[0089] Step 1221-1, based on the driving information and driving intention at the current moment, and the driving intention of the driving information at each moment before the current moment, determine the trajectory characteristics at the current moment;

[0090] Step 1221 - 2 , based on the trajectory features of each historical driving trajectory, determine the matching probability between the trajectory features at the current moment and each historical driving trajectory.

[0091] Specifically, when encoding the trajectory features at the current moment, in the time dimension, it is necessary to consider not only the driving information at the current moment, but also the driving information at each moment before the current moment; in the feature dimension, it is necessary to consider not only the driving information at each moment, but also the driving intention at each moment, thereby ensuring that the trajectory features at the current moment obtained by encoding can fully reflect the driving trajectory of the vehicle to be predicted in this trip.

[0092] After obtaining the trajectory features at the current moment, the trajectory features of each historical driving trajectory can be matched with it, so as to obtain the matching probability of the trajectory features at the current moment and each historical driving trajectory as the matching probability of each historical driving trajectory being the estimated trajectory corresponding to the driving information.

[0093] The method provided by the embodiment of the present invention combines driving intention in the trajectory matching process, thereby improving the reliability and accuracy of trajectory feature representation and helping to improve the accuracy of matching probability.

[0094] Based on any of the above embodiments, Figure 6 is a flow chart of step 1221-1 in the destination prediction method provided by the present invention, such as Figure 6 As shown, step 1221-1 includes:

[0095] Step 1221-1a, based on the driving information at the current moment and the driving information at each moment before the current moment, determine the trajectory regularity information at the current moment.

[0096] Specifically, in order to achieve trajectory matching prediction based on driving information, the driving trajectory pattern of the vehicle to be predicted can be statistically analyzed based on the driving information at each moment, and the trajectory pattern information here can reflect the driving path pattern during driving. Specifically in step 1221-1a, the driving information is sorted according to the corresponding time, and the trajectory pattern information at the current moment is obtained by integrating the driving information at each moment.

[0097] For example, for the first moment, the trajectory regularity information of the first moment can be determined based on the driving information of the first moment; for each moment after the first moment, the trajectory regularity information of the corresponding moment can be determined based on the driving information of the corresponding moment and the trajectory regularity information of the previous moment. In this way, the trajectory regularity information of each moment can not only reflect the driving information of the corresponding moment, but also cover the trajectory regularity information of each moment before the corresponding moment.

[0098] Here, the determination of the trajectory regularity information can be achieved by a pre-set acquisition rule or a pre-trained neural network model. The first moment can be the moment of a preset time period forward when performing trajectory matching prediction at the current moment, or it can be the starting moment of the trip, which is not specifically limited in the embodiment of the present invention.

[0099] In addition, the intention regularity information and trajectory regularity information obtained based on driving information may be the same or different. When they are the same, regularity extraction may be performed based on a neural network model with the same parameters. When they are different, regularity extraction may be performed based on a neural network model with different parameters or different structures. The embodiment of the present invention does not make any specific limitation on this.

[0100] Step 1221-1b, based on the trajectory regularity information and driving intention at the current moment, determine the trajectory code at the current moment.

[0101] Specifically, after obtaining the trajectory regularity information at the current moment, it can be encoded in combination with the driving intention at the current moment, so as to obtain the trajectory code at the current moment. For example, the two can be spliced ​​to obtain the trajectory code.

[0102] Step 1221 - 1c, based on the trajectory code at the current moment and the trajectory features at the previous moment before the current moment, determine the trajectory features at the current moment.

[0103] Specifically, when extracting trajectory features for the current moment, it can also be implemented in combination with the trajectory coding of each moment. For example, for each moment after the first moment, the trajectory features of the corresponding moment can be determined based on the trajectory coding of the corresponding moment and the trajectory features of the previous moment of the corresponding moment, so that the trajectory features of each moment can not only reflect the information of the driving code of the corresponding moment, but also cover the trajectory features of each moment before the corresponding moment.

[0104] Based on any of the above embodiments, step 120 includes:

[0105] The driving information is input into the exclusive destination prediction model of the user who is currently driving the vehicle to be predicted, and the exclusive destination prediction model estimates the driving intention based on the driving information, and performs trajectory prediction based on the driving information and the driving intention to obtain the estimated destination output by the exclusive destination prediction model;

[0106] The exclusive destination prediction model is obtained by transfer learning of the general trajectory prediction model based on the user's historical driving information and historical destinations.

[0107] Specifically, the estimation of driving intention based on driving information and the estimation of driving trajectory based on driving information and driving intention can be achieved through an end-to-end exclusive destination prediction model. Taking into account the differences in driving habits and familiar routes of different users, the exclusive destination prediction model corresponds one-to-one to the user.

[0108] Accordingly, before executing step 120, a destination prediction model exclusive to the user currently driving the vehicle to be predicted can be pre-trained, and the specific training method includes: first, collecting the user's historical driving information and its corresponding historical destinations. Then, based on the pre-trained general trajectory prediction model, the user's historical driving information and its corresponding historical destinations are applied for transfer learning, thereby obtaining the user's exclusive destination prediction model.

[0109] The pre-trained general trajectory prediction model is trained based on samples of historical driving information and historical destinations that do not distinguish between users, for example, it can be trained based on the MGSIM dataset. Migration training based on the general trajectory prediction model can accelerate the convergence speed of the exclusive destination prediction model and reduce the computing resources required to provide exclusive destination prediction models for each user.

[0110] On this basis, the specific execution steps covered in step 120 provided in the above-mentioned embodiments can all be converted into execution steps within the user's exclusive destination prediction model. Within the user's exclusive destination prediction model, each functional module can be divided to execute the corresponding steps, thereby realizing user-exclusive destination prediction.

[0111] Based on any of the above embodiments, the driving information includes driving trajectory information, or includes driving trajectory information and driving environment information.

[0112] Here, the driving trajectory information refers to information related to the driving trajectory of the vehicle to be predicted, such as the position, time, front direction, road of each trajectory point during the vehicle's driving, and the mileage and time from the starting point to each trajectory point.

[0113] Driving environment information refers to information related to the driving environment of the vehicle to be predicted, but the in-vehicle environment can also be the external environment, such as the number of people in the car, their identities, etc., and the congestion level of the road where the vehicle is located, the surrounding buildings and landmarks of the road where the vehicle is located, etc.

[0114] When the driving information includes both driving trajectory information and driving environment information, the driving intention is estimated based on the driving information, and the driving trajectory is estimated based on the driving information and the driving intention, so as to obtain a more accurate and reliable estimated destination.

[0115] Furthermore, with regard to the acquisition of driving trajectory information, it can be obtained through the built-in positioning and identification module of the vehicle to be predicted, for example, the vehicle position and the road the vehicle is on can be determined through the GPS positioning module, and the direction of the vehicle head can be determined through the direction sensor; with regard to the acquisition of driving environment information, the images of people in the car can be collected through the camera installed in the car, and the number of people can be obtained in conjunction with target detection technology, and the identity of the people can be obtained in conjunction with face recognition, or the voice of the people in the car can be collected through the audio collection equipment installed in the car, and the identity of the people can be obtained in conjunction with voiceprint recognition, and the images of the outside environment can also be collected through the camera installed outside the car, and the surrounding buildings and landmarks can be obtained in conjunction with target detection technology, or the vehicle position can be determined through the GPS positioning module, and then the degree of road congestion and surrounding buildings and landmarks can be obtained.

[0116] Based on any of the above embodiments, the driving information can be expressed as the following formula:

[0117]

[0118] Among them, I (t) is the driving information at time t, and E (t) are the driving trajectory information and driving environment information at time t, T p is the time span. The driving environment information can be expressed as follows:

[0119] E (t) =f(t)+g(p (t) )

[0120] Where f(t) reflects the time information at time t, p (t) is the occupant information at time t, g(p (t) ) is the encoding vector of the occupant information at time t.

[0121] Based on any of the above embodiments, Figure 7 is a schematic diagram of the structure of the destination estimation model provided by the present invention, such as Figure 7 As shown, the destination prediction model can be used to implement the operation of step 120, where the destination prediction model can be a dedicated model for the user who is currently driving the vehicle to be predicted.

[0122] The destination estimation model can be divided into four parts: input module, intention recognition module, trajectory output module and output module.

[0123] The input module is used to realize the driving information I at the current time T. (T) , and the driving information I at each time before the current time (t) ,t=(TT p ,…,T-1). Since the input module is connected to the intention recognition module and the trajectory output module respectively, the driving information at each moment needs to be input to the subsequent modules through the corresponding fully connected layer FC. Among them, the fully connected layer FC connected to the intention recognition module is set with a vector space label.

[0124] The intention recognition module is used to execute step 121 in step 120, that is, to estimate the driving intention based on the driving information. Specifically, the intention regularity information at the current moment can be extracted through the LSTM network. The intention regularity information here can be understood as the hidden layer state output by the LSTM network in the intention recognition module, which can be identified as follows:

[0125] h (t) =f(h (t-1) ,I (t) )

[0126] In the formula, h (t) That is, the hidden state at time t, that is, the intention regularity information at time t. (t-1) ,I (t) ) is the function representation of the LSTM network.

[0127] Subsequently, the intention regularity information at the current moment is sorted out through the Softmax layer to obtain the probability matrix Ω=(w1,w2,w3) reflecting the driving intention, where w1,w2,w3 represent the probabilities of the vehicle changing lanes to the left, going straight, and changing lanes to the right, respectively. Furthermore, confidence thresholds can be set in advance for turning left, going straight, and turning right. For example, the confidence threshold for changing lanes to the left or right is 85%, while the confidence threshold for going straight is 75%. When the output probability of a certain category is greater than the corresponding confidence threshold, the category is determined to be the final predicted intention, and the probability of the category is adjusted to 1, and the probabilities of the other two categories are adjusted to 0.

[0128] The trajectory output module is used to execute step 122 in step 120, and can be further refined into an encoder, a decoder, and a mixed density network layer. The encoder is used to encode the input driving information into a vector of fixed length to characterize the regularity of the trajectory that has been driven, and the decoder is used to extract information from the vector to predict the trajectory of the vehicle. The mixed density layer is used to match the predicted trajectory with the historical trajectory to obtain the matching probability distribution of the predicted trajectory and each historical trajectory.

[0129] The encoder is used to perform step 1221-1a, that is, based on the driving information at the current moment and the driving information at each moment before the current moment, determine the trajectory regularity information at the current moment. The encoder can extract the trajectory regularity information at the current moment through the LSTM network. The trajectory regularity information here can be understood as the hidden layer state output by the LSTM network in the encoder.

[0130] After that, the trajectory output module can also perform step 1221-1b, that is, determine the trajectory code at the current moment based on the trajectory regularity information and driving intention at the current moment. Specifically, the probability matrix Ω reflecting the driving intention output by the intention recognition module can be fused with the trajectory regularity information at the current moment after passing through the fully connected layer, so as to obtain the trajectory code r at the current moment.

[0131] The decoder is used to execute step 1221-1c, that is, to determine the trajectory features of the current moment based on the trajectory code of the current moment and the trajectory features of the previous moment. In the decoder, the trajectory features can be extracted through the LSTM network, specifically, the trajectory features of the previous moment q (t-1) , the hidden state h at the previous moment (t-1) And the trajectory code r, get the hidden state h at the current moment (t) :

[0132] h (t) =f(h (t-1) ,q (t-1) ,r)

[0133] In the formula, f(h(t-1) ,q (t-1) , r) is the function representation of the LSTM network.

[0134] Then, the decoder calculates the hidden state h at the current moment. (t) , the trajectory characteristics q at the previous moment (t-1) And the trajectory code r, update the trajectory features at the current moment:

[0135] q (t) =g(h (t) ,q (t-1) , r)

[0136] In the formula, g(h (t) ,q (t-1) , r) is the function representation of trajectory feature extraction.

[0137] In the decoder, for the trajectory features at any moment, the dimension can be adapted through the fully connected layer FC, and the trajectory features after the dimension adaptation of the fully connected layer are recorded as z. The dimension adaptation here refers to the adaptation to the input dimension of the mixture-density-networks (MDN) connected to the decoder.

[0138] The mixed density network can execute step 1221-2, that is, based on the trajectory characteristics of each historical driving trajectory, determine the matching probability between the trajectory characteristics at the current moment and each historical driving trajectory. The mixed density network can give the probability distribution of different outputs and calculate the confidence of the output results. The overall idea is to use a set of kernel functions to linearly combine any input to approximate the output, and continuously iterate by representing the distribution parameters of the kernel function through the neural network.

[0139] In the embodiment of the present invention, the kernel function of the mixed density network is composed of K Gaussian functions, and the trajectory probability distribution is:

[0140]

[0141] Where P π，μ，σ (O i |z) represents the input trajectory feature z and the i-th historical driving trajectory O i The matching probability between , where:

[0142]

[0143] π k is the partition coefficient;

[0144]

[0145]

[0146] ρ k ∈[-1, 1]

[0147] φ is a bivariate Gaussian function, u k is the mean, ∑ k z is the covariance matrix, σ k is the standard deviation, ρ k is the correlation coefficient, and n is the number of moments in the sequence based on the driving information at each moment.

[0148] The output module is used to execute step 1222, that is, based on the matching probability of each historical driving trajectory and the driving intention, determine the estimated trajectory, and determine the destination of the estimated trajectory as the estimated destination. Specifically, it can be embodied in the following formula:

[0149] P(O|I)=∑w i P π，μ，σ (O i |z)

[0150] Where P(O|I) is the probability of the final output estimated trajectory, and the trajectory with the highest probability can be selected as the estimated trajectory, and the destination of the estimated trajectory can be determined as the estimated destination.

[0151] Based on any of the above embodiments, after step 120, the following steps may be further performed:

[0152] Obtain historical business data for the estimated destination;

[0153] Based on the driving information, the recommended business is selected from the historical business data, and a recommendation reminder of the recommended business is generated and sent.

[0154] Specifically, after obtaining the estimated destination, the historical business data of the estimated destination can be obtained. The historical business data here refers to the relevant data of the business performed by the user who is currently driving the vehicle to be predicted after reaching the estimated destination in the past, such as food orders, reservation records, parking locations and records, etc. Assuming that the estimated destination is a shopping mall, and the user has made a reservation at a restaurant in the mall, or the user's consumption record at a restaurant in the mall is queried, the relevant information of the restaurant can be added to the historical business data of the estimated destination.

[0155] Considering that the driving information reflects the driving conditions of the current vehicle to be predicted, the arrival time at the estimated destination can be predicted based on the driving information, thereby determining whether the various services covered in the historical business data are available when the vehicle arrives, or which services in the historical business data are more suitable for the vehicle arrival time, thereby selecting the recommended services from the historical business data and generating a recommendation reminder to send to the user.

[0156] Alternatively, the route to the estimated destination can be predicted based on driving information, so that a parking lot location that is more suitable for the route can be selected from historical business data, and the parking business corresponding to the suitable parking lot location can be recommended to the user as a to-be-recommended business.

[0157] In addition, based on the time in the driving information, it is also possible to determine whether it is a weekday or a holiday, and thus select the business at the corresponding time in the historical business data as the recommended business. For example, the journey from home to school may be to send children to school on weekdays, and to go to a nearby cram school on holidays. Therefore, the parking space recommendations for weekdays and holidays may be different.

[0158] On this basis, the recommendation reminder generated for the recommended service can also be determined based on factors such as the service itself, the surrounding conditions of the service, and the arrival time. For example, when the recommended service is a restaurant service, it can be estimated whether the restaurant is full at the arrival time. If it is possible, a recommendation reminder can be generated to prompt the user to make an appointment in advance.

[0159] Based on any of the above embodiments, a destination prediction method may include the following steps:

[0160] The driving information of the vehicle to be predicted is obtained, including driving time: Saturday afternoon, driving section: Road A, vehicle head direction: vehicle head facing north, lane direction: straight ahead. Based on the above driving information, it can be estimated that the driving intention is to turn left. On this basis, combined with the driving intention to turn left and the above driving information, the historical driving trajectory is matched. The matching result is that the matching probability of the current driving information and the historical driving trajectory to the mall is >85%. It can be determined that the estimated destination is the mall, and the user's historical business data in the mall is queried to give a business recommendation reminder.

[0161] Alternatively, the driving information of the vehicle to be predicted is obtained, including the driving time: 7 am on Monday, the people in the car: the user and the child, and the surrounding environment: the fountain passed by from home to school. Based on the above driving information, the driving intention can be estimated to be straight. On this basis, combined with the driving intention to go straight and the above driving information, the historical driving trajectory is matched. The matching result is that the current driving information has a matching probability of >95% with the historical driving trajectory to the school. It can be determined that the estimated destination is the school. The user's historical parking behavior is used to obtain easy parking spaces around the school, allowing the user to get off safely, bringing intelligence and convenience to the user's actual travel.

[0162] Alternatively, when there is a 4S store near the estimated destination determined based on driving information, it can be determined based on the mileage of the predicted vehicle whether the predicted vehicle needs to go to the 4S store for maintenance and repair. When there is a maintenance and repair need, a 4S store recommendation reminder is sent to the user.

[0163] Based on any of the above embodiments, Figure 8 is a schematic diagram of the structure of the closed-loop recommendation system provided by the present invention, such as Figure 8 As shown, the closed-loop recommendation system is a system for performing service closed-loop and business recommendation, including:

[0164] The driving information collection module 810 is used to realize real-time driving information collection, and specifically can realize information collection through the positioning recognition system installed in the vehicle, the image and sound sensor devices inside and outside the vehicle, etc.

[0165] The server module 820 can receive the driving information collected in real time by the driving information collection module 810, and transmit the driving information to the user intention understanding and calculation module 830.

[0166] The user intention understanding calculation module 830 is used to understand the intention of the input driving information, obtain the driving intention at the current moment, and transmit the driving intention to the user trajectory probability output module 840.

[0167] The user trajectory probability output module 840 is used to estimate the trajectory of the current trip of the vehicle to be predicted based on the input driving information and driving intention, so as to obtain the matching probability of each historical driving trajectory of the user being the estimated trajectory corresponding to the driving information.

[0168] The destination judgment analysis module 850 is used to determine the current possible estimated trajectory according to the matching probability of each historical driving trajectory of the user to the estimated trajectory corresponding to the driving information, so as to determine the current estimated destination.

[0169] The real-time calculation and updating module 860 is used to select the service to be recommended from the historical service data corresponding to the estimated destination in combination with the driving information, generate and send the recommendation reminder of the service to be recommended.

[0170] Based on any of the above embodiments, Fig. 9 is a schematic diagram of the structure of the destination prediction device provided by the present invention, such as Fig. 9 As shown, the device comprises:

[0171] The information acquisition unit 910 is used to acquire the driving information of the vehicle to be predicted;

[0172] The destination estimation unit 920 is used to estimate the driving intention of the vehicle to be predicted based on the driving information, and to perform trajectory estimation based on the driving information and the driving intention to obtain the estimated destination of the vehicle to be predicted.

[0173] The device provided by the embodiment of the present invention estimates the driving intention through the driving information of the vehicle, and thus predicts the destination by combining the driving information and the driving intention; the combination of driving intention can guide the direction of travel of the estimated trajectory, thereby ensuring that the final estimated trajectory is consistent with the user's driving habits while also being able to fit the user's driving intention, thereby greatly improving the reliability and accuracy of destination prediction in non-navigation scenarios.

[0174] Based on any of the above embodiments, the destination estimation unit 920 includes an intention recognition module, which is used to:

[0175] Determining the intention regularity information at the current moment based on the driving information at the current moment and the intention regularity information at the previous moment before the current moment;

[0176] Based on the intention regularity information at the current moment, the driving intention at the current moment is predicted.

[0177] Based on any of the above embodiments, the intention recognition module is specifically used for:

[0178] Determining the driving intention corresponding to the intention law information at the current moment based on the relationship between the historical intention law information and the driving intention;

[0179] The relationship between the historical intention regularity information and the driving intention is determined based on the historical driving information of the user who is currently driving the vehicle to be predicted.

[0180] Based on any of the above embodiments, the destination estimation unit 920 includes a trajectory matching module, which is used to:

[0181] Performing trajectory matching on the driving information to obtain a matching probability that each historical driving trajectory of the vehicle to be predicted is a corresponding estimated trajectory of the driving information;

[0182] Based on the matching probability of each historical driving trajectory and the driving intention, the estimated trajectory is determined, and the destination of the estimated trajectory is determined as the estimated destination.

[0183] Based on any of the above embodiments, the trajectory matching module is specifically used for:

[0184] Determine the trajectory characteristics of the current moment based on the driving information and driving intention at the current moment, and the driving intention of the driving information at each moment before the current moment;

[0185] Based on the trajectory features of each historical driving trajectory, a matching probability between the trajectory features at the current moment and each historical driving trajectory is determined.

[0186] Based on any of the above embodiments, the trajectory matching module is specifically used for:

[0187] Based on the driving information at the current moment and the driving information at each moment before the current moment, determining the trajectory regularity information at the current moment;

[0188] Determining the trajectory code at the current moment based on the trajectory regularity information at the current moment and the driving intention;

[0189] Based on the trajectory code at the current moment and the trajectory feature at a moment before the current moment, the trajectory feature at the current moment is determined.

[0190] Based on any of the above embodiments, the destination estimation unit 920 is used to:

[0191] The driving information is input into a dedicated destination prediction model of a user who is currently driving the vehicle to be predicted, and the dedicated destination prediction model estimates the driving intention based on the driving information, and performs trajectory prediction based on the driving information and the driving intention, to obtain an estimated destination output by the dedicated destination prediction model;

[0192] The exclusive destination prediction model is obtained by performing transfer learning on the general trajectory prediction model based on the historical driving information and historical destinations of the user.

[0193] Based on any of the above embodiments, the driving information includes driving trajectory information, or includes the driving trajectory information and driving environment information.

[0194] Based on any of the above embodiments, the device further includes a service recommendation unit, configured to:

[0195] Acquiring historical business data of the estimated destination;

[0196] Based on the driving information, selecting a service to be recommended from historical service data;

[0197] Generate and send a recommendation reminder for the service to be recommended.

[0198] Fig.10 An example of a physical structure diagram of an electronic device is shown in FIG. Fig.10As shown, the electronic device may include: a processor 1010, a communication interface 1020, a memory 1030 and a communication bus 1040, wherein the processor 1010, the communication interface 1020 and the memory 1030 communicate with each other through the communication bus 1040. The processor 1010 may call the logic instructions in the memory 1030 to execute the destination prediction method, which includes: obtaining the driving information of the vehicle to be predicted; estimating the driving intention of the vehicle to be predicted based on the driving information, and performing trajectory estimation based on the driving information and the driving intention to obtain the estimated destination of the vehicle to be predicted.

[0199] In addition, the logic instructions in the above-mentioned memory 1030 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk.

[0200] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the destination prediction method provided by the above-mentioned methods, and the method includes: obtaining driving information of the vehicle to be predicted; estimating the driving intention of the vehicle to be predicted based on the driving information, and performing trajectory estimation based on the driving information and the driving intention to obtain the estimated destination of the vehicle to be predicted.

[0201] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it is implemented to execute the above-mentioned destination prediction methods, the method comprising: obtaining driving information of the vehicle to be predicted; estimating the driving intention of the vehicle to be predicted based on the driving information, and performing trajectory estimation based on the driving information and the driving intention to obtain the estimated destination of the vehicle to be predicted.

[0202] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0203] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0204] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A destination prediction method, It is characterized in that include: Acquire driving information of the vehicle to be predicted, wherein the driving information is information from the start time of the vehicle to the current time, or information calculated from the current time to a preset time period; estimating the driving intention of the vehicle to be predicted based on the driving information, and performing trajectory estimation based on the driving information and the driving intention, and determining an estimated destination of the vehicle to be predicted based on the estimated trajectory, wherein the driving intention reflects the driving direction of the vehicle to be predicted; The estimating the driving intention of the vehicle to be predicted based on the driving information includes: Based on the driving information at the current moment and the intention law information at the previous moment before the current moment, determining the intention law information at the current moment, wherein the intention law information reflects the driving law during the driving process; Based on the intention regularity information at the current moment, the driving intention at the current moment is predicted.

2. The destination prediction method according to claim 1, It is characterized in that The predicting the driving intention at the current moment based on the intention regularity information at the current moment includes: Determining the driving intention corresponding to the intention law information at the current moment based on the relationship between the historical intention law information and the driving intention; The relationship between the historical intention regularity information and the driving intention is determined based on the historical driving information of the user who is currently driving the vehicle to be predicted.

3. The destination prediction method according to claim 2, It is characterized in that The performing trajectory estimation based on the driving information and the driving intention, and determining the estimated destination of the vehicle to be estimated based on the estimated trajectory, includes: Performing trajectory matching on the driving information to obtain a matching probability that each historical driving trajectory of the vehicle to be predicted is a corresponding estimated trajectory of the driving information; Based on the matching probability of each historical driving trajectory and the driving intention, the estimated trajectory is determined, and the destination of the estimated trajectory is determined as the estimated destination.

4. The destination prediction method according to claim 3, It is characterized in that The performing trajectory matching on the driving information to obtain the matching probability of each historical driving trajectory of the vehicle to be predicted being the estimated trajectory corresponding to the driving information includes: Determine the trajectory characteristics of the current moment based on the driving information and driving intention at the current moment, and the driving intention of the driving information at each moment before the current moment; Based on the trajectory features of each historical driving trajectory, a matching probability between the trajectory features at the current moment and each historical driving trajectory is determined.

5. The destination prediction method according to claim 4, It is characterized in that The determining of the trajectory feature at the current moment based on the driving information and driving intention at the current moment, and the driving intention of the driving information at each moment before the current moment, includes: Based on the driving information at the current moment and the driving information at each moment before the current moment, determining the trajectory regularity information at the current moment; Determining the trajectory code at the current moment based on the trajectory regularity information at the current moment and the driving intention; Based on the trajectory code at the current moment and the trajectory feature at a moment before the current moment, the trajectory feature at the current moment is determined.

6. The destination prediction method according to any one of claims 1 to 5, It is characterized in that The estimating the driving intention of the vehicle to be predicted based on the driving information, performing trajectory estimation based on the driving information and the driving intention, and determining the estimated destination of the vehicle to be predicted based on the estimated trajectory includes: The driving information is input into a dedicated destination prediction model of a user who is currently driving the vehicle to be predicted, and the dedicated destination prediction model estimates the driving intention based on the driving information, and performs trajectory prediction based on the driving information and the driving intention, to obtain an estimated destination output by the dedicated destination prediction model; The exclusive destination prediction model is obtained by performing transfer learning on the general trajectory prediction model based on the historical driving information and historical destinations of the user.

7. The destination prediction method according to any one of claims 1 to 5, It is characterized in that The driving information includes driving track information, or includes the driving track information and driving environment information.

8. The destination prediction method according to any one of claims 1 to 5, It is characterized in that The step of obtaining the estimated destination of the vehicle to be predicted further includes: Acquiring historical business data of the estimated destination; Based on the driving information, selecting a service to be recommended from historical service data; Generate and send a recommendation reminder for the service to be recommended.

9. A destination prediction device, It is characterized in that include: An information acquisition unit, used to acquire driving information of the vehicle to be predicted, wherein the driving information is information from the start time of the vehicle to the current time, or information calculated from the current time forward for a preset time period; a destination estimation unit, configured to estimate the driving intention of the vehicle to be estimated based on the driving information, and to estimate a trajectory based on the driving information and the driving intention, and to determine an estimated destination of the vehicle to be estimated based on the estimated trajectory, wherein the driving intention reflects the driving direction of the vehicle to be estimated; The destination estimation unit is specifically used for: Based on the driving information at the current moment and the intention law information at the previous moment before the current moment, determining the intention law information at the current moment, wherein the intention law information reflects the driving law during the driving process; Based on the intention regularity information at the current moment, the driving intention at the current moment is predicted.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the program, the steps of the destination prediction method according to any one of claims 1 to 8 are implemented.

11. A non-transitory computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the destination prediction method according to any one of claims 1 to 8 are implemented.

Images