Vehicle trajectory prediction method, system, device and medium
By adopting an uplink perception scheme combined with OTFS modulation and deep learning in autonomous vehicles, the LSTM neural network model is used to predict vehicle trajectory, which solves the problem of trajectory prediction accuracy in high mobility scenarios, and improves the safety of autonomous vehicles and efficiency.
Patent Information
- Application Number
- CN202411570581.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-06
- Publication Date
- 2024-12-06
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The prior art is difficult to achieve accurate trajectory prediction of autonomous vehicles in high mobility scenarios, especially when signal attenuation and interference are large under the influence of Doppler effect, which affects the safe driving and path planning of vehicles.
Using an uplink perception scheme based on OTFS modulation and deep learning, the status information of the target vehicle, including signal arrival angle and delay Doppler, is obtained through the LSTM neural network model, to predict the vehicle's trajectory, and improve the accuracy and safety of trajectory prediction. .
Through this solution, the vehicle driving trajectory can be effectively predicted, the safety and efficiency of autonomous vehicles are improved, the impact of Doppler effect is reduced, and the communication capabilities between the vehicle and the infrastructure are enhanced.
Smart Images

Figure CN119078887A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of vehicle driving technology, and in particular to a vehicle trajectory prediction method, system, device and medium. Background Art
[0002] Trajectory prediction of autonomous vehicles is one of the key technologies to ensure safe driving of vehicles. The accuracy of trajectory prediction can directly affect the vehicle's collision avoidance, path planning and system responsiveness. In order to achieve these goals, the vehicle needs to be able to accurately perceive the surrounding environment and communicate effectively with other vehicles and infrastructure.
[0003] Ultra-reliable low-latency (URLLC) services for vehicle-to-infrastructure (V2I) communications are key to the safe operation of autonomous vehicles, ensuring that vehicles receive critical information in a very short time so they can respond quickly.
[0004] Orthogonal Time-Frequency-Space (OTFS) modulation is a technology that operates signals in the Delay-Doppler (DD) domain. It is highly resistant to the Doppler effect and is very suitable for communication and perception in high mobility scenarios. Compared with Orthogonal Frequency Division Multiplexing (OFDM) technology, OTFS has obvious advantages in handling high mobility communications, because the orthogonality of OFDM in the time-frequency domain is affected by the increase in mobile speed, resulting in signal attenuation and interference between subcarriers.
[0005] The application of deep learning (DL) technology in channel estimation, target classification and angle parameter prediction provides new solutions and optimization methods for the OTFS-assisted ISAC system. The powerful performance of deep learning can help improve the accuracy of trajectory prediction, thereby promoting the development of autonomous driving technology.
[0006] Therefore, in order to realize the ISAC framework in the vehicle network, it is necessary to develop an uplink perception scheme based on OTFS modulation and deep learning. Such a scheme will be able to fully utilize the Doppler robustness of OTFS and the powerful data processing capabilities of deep learning, thereby improving the safety and efficiency of autonomous vehicles. Summary of the invention
[0007] To solve the above problems, the present invention provides a vehicle trajectory prediction method in a first aspect, comprising: Get the status information of the target vehicle, The target vehicle is a vehicle for which vehicle trajectory prediction is required; Based on the state information, determining the position information of the target vehicle at least at one time in the first time period; Based on the position information of the target vehicle at at least one time in the first time period, a first position sequence of the target vehicle in the first time period is determined, The first position sequence is a sequence obtained by arranging position information of at least one moment in time in chronological order; Based on the first position sequence, predicting a second position sequence of the target vehicle in a second time period by using a prediction model; The second time period is a time period after the first time period; The second position sequence is a sequence obtained by arranging the position information of the target vehicle at at least one moment in the second time period in chronological order; Wherein, the time sequence is a positive time sequence or a reverse time sequence; The obtaining of the state information of the target vehicle comprises: receiving a continuous time signal from a target vehicle; Determine the state information of the target vehicle based on the continuous time signal; The prediction model is an LSTM neural network model.
[0008] In some embodiments, determining the state information of the target vehicle includes: determining the signal arrival angle of each path, and the time delay and Doppler corresponding to each path.
[0009] In some embodiments, estimating the signal arrival angle comprises: Assume that the signal received by the receiving antenna is , ,in: (7) In (7), Y represents the signal matrix received by the receiving antenna, A represents the angle matrix of each signal, S represents the transmitted signal matrix, and N represents the noise matrix; to Indicates the received data from the first to the Brth antennas, to Indicates that the antenna array containing Br antennas corresponds to the target index direction to The steering vector at to is the sending signal corresponding to target index 0 to target N, to is the receiving noise corresponding to antenna index 0 to antenna Br; The covariance matrix of the received signal is expressed as ; in, , are the covariance matrices of the transmitted signal and noise respectively; through eigenvalue decomposition, we get As the basis of the signal subspace, and containing the remaining eigenvectors As a basis of the noise subspace; The spatial spectrum is expressed as: (8) (8) Indicates the direction The basis vectors of the noise subspace at , Indicates direction The corresponding steering vector, the superscript H indicates the conjugate transpose operation; when is the signal arrival angle, ; When the spatial spectrum reaches its peak, the spatial spectrum will have N+1 peaks, corresponding to N+1 signals, so as to obtain the signal arrival angle of each path.
[0010] In some embodiments, estimating the delay and Doppler corresponding to each path includes: Assume that the OTFS transmission signal is: (9) (9) is the element in the delay-Doppler array, It is a DD domain grid with any position The pilot symbols are 0 and They are protection symbol and information data symbol respectively. and s represent the delay and Doppler domain indexes respectively, and Respectively represent the index value of pilot placement, and Respectively represent the maximum delay index and the maximum Doppler index, Represents data symbol; The received symbol used for channel estimation is represented as: (10) Set the threshold value to ,when , , indicating that there is a path; in, Indicates delay, represents the Doppler shift, is an indicator variable of 0 or 1, 1 means there is a path, Indicated in , The estimated channel gain at express , The estimation bias at Get the gain of this path , on the contrary, , indicating that the path does not exist; Select ,in, ; get The delay of the i-th vehicle at time , Doppler , and the angle .
[0011] In some embodiments, determining the location information of the target vehicle at least at one time in the first time period includes: Assume the location information of the target vehicle , the speed of the target vehicle’s position , the speed of the target vehicle’s position .
[0012] Assume that the acceleration of the vehicle in the x and y directions is a random variable, which is subject to , ,in , is a constant. According to the kinetic equation, we can get: (11) (11) is a time step; Assume the location of the base station is , according to the geometric relationship, the position information corresponding to the i-th vehicle is obtained as: (12) (12) ; represents the straight-line distance from the base station to the receiving station, represents the propagation distance of the signal reflected by the i-th target; and They correspond to the angle between the direct connection line and the x-axis, and the angle between the connection line between the i-th target and the receiving end and the x-axis.
[0013] In some embodiments, the LSTM neural network model includes: an input layer, three LSTM layers, a fully connected layer and an output layer; Let the input layer be , indicating that it includes Vehicle location information at the moment.
[0014] In the LSTM layer, let and is the hidden state and cell state of the nth LSTM layer at the mth time step, and the two have the following relationship: (13) (13) is the output gate, which is determined by the input at the current time m and the hidden state at the previous time m-1; The hidden state of the last time step of the last layer is the output of the entire LSTM layer, which is expressed by the following formula: (14) (14) It is a parameter that summarizes the characteristics of the entire LSTM layer. After three layers of LSTM, it enters the fully connected layer; the fully connected layer is used to calculate the output of the previous layer. Perform linear operations and finally output the predicted position information ; The characteristics of the output layer are: (15) (15) is the weight, is bias; The output layer is represented as: (16) (16) is a function that summarizes the characteristics of the entire network, and the parameters .
[0015] In some embodiments, the LSTM neural network model uses a mean square error loss function to reflect the errors of multiple training samples, and the mean square error loss function is: (17) (17) , They represent the label and input of the nth training sample respectively, and M is the number of samples.
[0016] A second aspect of the present invention provides a vehicle trajectory prediction system, comprising a server, a network, a user terminal, a vehicle and a storage device, wherein the server obtains information and / or data of the vehicle via the network, the server comprises a processing device, the processing device processes information and / or data related to a service request, the network is used to facilitate the exchange of information and / or data, the storage device is used to store data obtained from the user terminal, the user terminal receives service confirmation and / or information or instructions from the server, and is used to execute a vehicle trajectory prediction method described in any one of the above schemes.
[0017] A third aspect of the present invention provides a vehicle trajectory prediction device, which includes at least one processor and at least one memory; the at least one memory is used to store computer instructions; and the at least one processor is used to execute at least part of the computer instructions to implement a vehicle trajectory prediction method described in any one of the above schemes.
[0018] A fourth aspect of the present invention is a computer-readable storage medium, which stores computer instructions. When a computer reads the computer instructions in the storage medium, the computer executes a vehicle trajectory prediction method described in any one of the above schemes.
[0019] By adopting the above technical solution, the present invention mainly has the following technical effects: By obtaining the state information of the target vehicle, the first position sequence of the target vehicle in the first time period is determined, and then the second position sequence of the target vehicle in the second time period is predicted by the prediction model. The trained LSTM network model can effectively predict the vehicle's driving trajectory, achieving the technical effect of improving the safety and efficiency of autonomous driving vehicles. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] Figure 1 is a schematic diagram of an application scenario of a vehicle trajectory prediction system according to some embodiments of this specification; Figure 2 is an exemplary flow chart of a vehicle trajectory prediction method according to some embodiments of this specification; Figure 3 is a schematic diagram of an LSTM neural network model according to some embodiments of this specification; Figure 4 is a comparison diagram of the actual trajectory and the predicted trajectory of the target vehicle according to some embodiments of this specification; Figure 5 A schematic diagram of RMSE changing over time according to some embodiments of this specification; Figure 6 Schematic diagram of RMSE of predicted trajectories under different historical time steps according to some embodiments of this specification. DETAILED DESCRIPTION
[0021] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of this specification. For ordinary technicians in this field, this specification can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.
[0022] It should be understood that the "system", "device", "unit" and / or "module" used herein are a method for distinguishing different components, elements, parts, portions or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.
[0023] As shown in this specification, unless the context clearly indicates an exception, the words "a", "an", "a kind" and / or "the" do not refer to the singular, but also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.
[0024] Flowcharts are used in this specification to illustrate the operations performed by the system according to the embodiments of this specification. It should be understood that the preceding or following operations are not necessarily performed precisely in order. Instead, the steps may be processed in reverse order or simultaneously. At the same time, other operations may also be added to these processes, or one or more operations may be removed from these processes.
[0025] Figure 1 1 is a schematic diagram of an application scenario of a vehicle trajectory prediction system according to some embodiments of this specification. In some embodiments, the vehicle trajectory prediction system 100 may include a server 110, a network 120, a user terminal 130, a vehicle 140, and a storage device 150.
[0026] In some embodiments, server 110 may be a single server or a server group. The server group may be centralized or distributed (e.g., server 110 may be a distributed system). In some embodiments, server 110 may be local or remote. For example, server 110 may obtain information and / or data of vehicle 140 via network 120. For another example, server 110 may be directly connected to storage device 150 to access stored information and / or data. In some embodiments, server 110 may be implemented on a cloud platform. By way of example only, a cloud platform may include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-layer cloud, etc., or any combination thereof. In some embodiments, server 110 may obtain information and / or data of vehicle 140 via a base station.
[0027] In some embodiments, the server 110 may include a processing device 112. The processing device 112 may process information and / or data related to the service request to perform one or more functions described in the present application. For example, the processing device 112 may predict the vehicle trajectory from the service request obtained from the user terminal 130. In some embodiments, the processing device 112 may include one or more processing engines (e.g., a single-chip processing engine or a multi-chip processing engine). By way of example only, the processing device 112 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), an application-specific instruction set processor (ASIP), a graphics processing unit (GPU), a physical processing unit (PPU), a digital signal processor (DSP), a field programmable gate array (FPGA), a programmable logic device (PLD), a controller, a microcontroller unit, a reduced instruction set computer (RISC), a microprocessor, etc., or any combination thereof.
[0028] The network 120 can facilitate the exchange of information and / or data. In some embodiments, one or more components (e.g., server 110, user terminal 130, vehicle 140, and / or storage device 150) in the vehicle trajectory prediction system 100 can send information and / or data to other components of the vehicle trajectory prediction system 100 through the network 120. For example, the server 110 can obtain / get a service request from the user terminal 130 through the network 120. In some embodiments, the network 120 can be any form of wired network or wireless network, or any combination thereof. Just as an example, the network 120 may include a cable network, a cable network, an optical fiber network, a telecommunications network, an intranet, the Internet, a local area network (LAN), a wide area network (WAN), a wireless local area network (WLAN), a metropolitan area network (MAN), a public switched telephone network (PSTN), a Bluetooth network, a purple bee network, a near field communication (NFC) network, etc. or any combination thereof. In some embodiments, the network 120 may include one or more network exchange points. For example, the network 120 may include wired or wireless network exchange points, such as base stations and / or Internet exchange points 120 - 1 , 120 - 2 , . . . , through which one or more components of the vehicle trajectory prediction system 100 may be connected to the network 120 to exchange data and / or information.
[0029] In some embodiments, the user (e.g., driver and / or passenger) may be the owner of the user terminal 130. In some embodiments, the owner of the user terminal 130 may be someone other than the passenger. For example, the owner A of the user terminal 130 may use the user terminal 130 to send a service request to the passenger B, or receive a service confirmation and / or information or instructions from the server 110.
[0030] In some embodiments, the user terminal 130 may include a mobile device 130-1, a tablet computer 130-2, a laptop computer 130-3, etc., or any combination thereof. In some embodiments, the mobile device 130-1 may include a smart home device, a wearable device, a smart mobile device, a virtual reality device, an augmented reality device, etc., or any combination thereof. In some embodiments, the smart home device may include a smart lighting device, a smart appliance control device, a smart monitoring device, a smart TV, a smart camera, an intercom, etc., or any combination thereof. In some embodiments, the wearable device may include a smart bracelet, smart shoes and socks, smart glasses, a smart helmet, a smart watch, smart clothing, a smart backpack, a smart accessory, etc., or any combination thereof. In some embodiments, the smart mobile device may include a smart phone, a personal digital assistant (PDA), a gaming device, a navigation device, a point of sale (POS), etc., or any combination thereof. In some embodiments, the virtual reality device and / or the augmented reality device may include a virtual reality helmet, virtual reality glasses, virtual reality goggles, an augmented reality helmet, augmented reality glasses, augmented reality goggles, etc., or any combination thereof. For example, the virtual reality device and / or the augmented reality device may include Google TM Glass, Oculus Rift, HoloLens, Gear VR, etc. In some embodiments, the user terminal 130 may be a device with positioning technology for locating the position of the user and / or the user terminal 130 .
[0031] Vehicle 140 can be any type of vehicle. Vehicle 140 can include automatic vehicles and non-automatic vehicles. Automatic vehicles are capable of sensing environmental information and navigating without human manipulation. Vehicle 140 may include the structure of a traditional vehicle. For example, vehicle 140 may include at least two control components, which are configured to control the operation of vehicle 140. At least two control components may include a steering device (e.g., a steering wheel), a braking device (e.g., a brake pedal), an accelerator, etc. The steering device may be configured to adjust the orientation and / or direction of vehicle 140. The braking device may be configured to perform a braking operation to stop vehicle 140. The accelerator may be configured to control the speed and / or acceleration of vehicle 140. Vehicle 140 may also include a global positioning system (GPS) module. In some embodiments, vehicle 140 may report operating information (e.g., positioning information and / or acceleration information, etc.) to processing device 112.
[0032] The storage device 150 may store data and / or instructions. In some embodiments, the storage device 150 may store data obtained from the user terminal 130. In some embodiments, the storage device 150 may store data and / or instructions used by the server 110 to execute or use to complete the exemplary methods described in this application. In some embodiments, the storage device 150 may include a mass storage, a removable memory, a volatile read-write memory, a read-only memory (ROM), etc., or any combination thereof. Exemplary mass storage may include a disk, an optical disk, a solid-state drive, etc. Exemplary removable memory may include a flash drive, a floppy disk, an optical disk, a memory card, a compressed disk, a tape, etc. Exemplary volatile read-write memory may include a random access memory (RAM). Exemplary RAM may include a dynamic random access memory (DRAM), a double data rate synchronous dynamic random access memory (DDRSDRAM), a static random access memory (SRAM), a thyristor random access memory (T-RAM), and a zero capacitance random access memory (Z-RAM), etc. Exemplary read-only memories may include mask read-only memories (MROMs), programmable read-only memories (PROMs), erasable programmable read-only memories (EPROMs), electrically erasable programmable read-only memories (EEPROMs), compact disk read-only memories (CD-ROMs), and digital versatile disk read-only memories, etc. In some embodiments, the storage device 150 may be implemented on a cloud platform. By way of example only, the cloud platform may include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-layer cloud, etc., or any combination thereof.
[0033] In some embodiments, the storage device 150 may be connected to the network 120 to communicate with at least one component of the vehicle trajectory prediction system 100 (e.g., the server 110, the user terminal 130, or the vehicle 140). At least one component of the vehicle trajectory prediction system 100 may access the data or instructions stored in the storage device 150 via the network 120. In some embodiments, the storage device 150 may be directly connected to or communicate with at least one component of the vehicle trajectory prediction system 100 (e.g., the server 110, the user terminal 130, or the vehicle 140). In some embodiments, the storage device 150 may be part of the server 110. In some embodiments, the storage device 150 may be deployed on the server 110, the user terminal 130, and / or the vehicle 140. In some embodiments, the storage device 150 may be deployed independently, and the server 110, the user terminal 130, and / or the vehicle 140 may directly access or access the storage device 150 through the network 120 to obtain relevant data and / or instructions.
[0034] It should be noted that the above description of the vehicle trajectory prediction system is only for convenience of description and does not limit the present specification to the scope of the embodiments. It is understandable that, after understanding the principle of the system, those skilled in the art may arbitrarily combine the components or form a subsystem connected with other components without deviating from the principle.
[0035] Figure 2 is an exemplary flow chart of a vehicle trajectory prediction method according to some embodiments of this specification. Figure 2 As shown, the process 200 includes the following steps. In some embodiments, the process 200 may be executed by the processing device 112 .
[0036] Step 210: Obtaining status information of the target vehicle; In some embodiments, the target vehicle is a vehicle for which vehicle trajectory prediction is required, and the predicted vehicle trajectory is the predicted route that the target vehicle will travel.
[0037] In some embodiments, the target vehicle may be specified by a user, or may be determined by a machine, or may be determined by a combination of manual and machine methods. For example, a user may query a vehicle through the user terminal 130, and the processing device 112 may determine the vehicle queried by the user as the target vehicle.
[0038] In some embodiments, the server 110 may obtain the status information of the target vehicle through a base station.
[0039] In some embodiments, the base station is provided with transmit antennas and The transmitting antenna and the receiving antenna are arranged into a uniform linear array (ULA). By independently setting the transmitting array and the receiving array, the signal sent by the transmitting antenna and the signal received by the receiving antenna will not interfere with each other.
[0040] In some embodiments, obtaining the status information of the target vehicle includes: S2101, receiving a continuous time signal of a target vehicle; In some embodiments, the base station receives continuous-time signals by converting the delay-Doppler array of the target vehicle's transmission into a signal in the time-frequency (TF) domain through an inverse sigmoid fast Fourier transform (ISFFT), and finally converting the discrete time-frequency domain signal into a continuous-time signal suitable for wireless transmission through an impulse shaping filter.
[0041] In some embodiments, when information is sent from the vehicle 140 to the server 110, there is a direct (LoS) propagation path and N non-direct (NLoS) propagation paths between the vehicle 140 and the server 110. The power of the direct path is much greater than the power of the non-direct path.
[0042] In some embodiments, the transmitted delay Doppler array is assumed to be , where each element in the delay-Doppler array is , Represents delay and Doppler respectively; through the inverse symplectic fast Fourier transform (ISFFT), the signal in the delay-Doppler domain is transformed The signal converted to the time-frequency (TF) domain is ,in Represents the number of subcarriers and time slots respectively; finally, the impulse shaping filter , obtain the continuous time signal; The specific method is: (1) (1) In is the frequency interval, is the symbol duration.
[0043] In some embodiments, in order to maintain orthogonality so that different subcarriers do not interfere with each other, , the uplink communication and sensing channels are: (2) (2) In , , Respectively represent The delay, Doppler and communication coefficient of each path, δ is the impulse function, and N is the total number of paths; Receive the direction vector for the i-th path; when When the receiving direction vector for: (3) (3) In is the cardinality of the receiving antenna array, For the The angle of arrival of the signal path, Represents an imaginary unit.
[0044] In some embodiments, the signal is transmitted through a channel, and the continuous signal passes through a receiving impulse shaping filter and OTFS demodulation at the receiving end, and finally a received signal is obtained in the DD domain.
[0045] The specific method is: (4) (4) is the received signal in the Delay-Doppler (DD) domain, is the receive beamformer, The value of the signal sent in the DD domain. For the The delay of the signal path; For the The Doppler of the signal path, The frequency spectral density is Gaussian white noise.
[0046] In some embodiments, assuming that the delay and Doppler shift are integers, then , .
[0047] Furthermore, in order to more accurately perceive each vehicle, it is necessary to distinguish signals from different paths. Due to the asymptotic orthogonality of different beam pointing vectors at different angles of the massive MIMO receiving antenna array, different targets can be detected by the angle of arrival (AoA). In some embodiments, from The sensing signal reflected by the target can be transmitted through the receiving beamformer Obtained, in the delay-Doppler (DD) domain: (5) (5) represents the composite antenna array gain, Indicates The path delay is , the Doppler shift is The gain, Indicates Noise on the path.
[0048] In some embodiments, all received symbols Stacked into a vector , which can be expressed as: (6) (6) No. The elements are , by estimating , get the first The delay and Doppler of the path corresponding to each target.
[0049] S2102, determining the state information of the target vehicle based on the continuous time signal; In some embodiments, the state information of the target vehicle includes: signal arrival angle, delay and Doppler.
[0050] The determining of the state information of the target vehicle includes: determining the signal arrival angle of each path, and the time delay and Doppler corresponding to each path.
[0051] In some embodiments, the signal arrival angle is a parameter that characterizes the angle at which the signal arrives at the receiving antenna array; the time delay is a parameter that characterizes the time required for the signal to propagate from the transmitting source to the receiver; and the Doppler is a parameter that characterizes the change in the frequency of the received signal caused by the relative motion between the transmitting source and the receiver.
[0052] In some embodiments, the signal arrival angle may be estimated by a MUSIC algorithm, and the estimated signal arrival angle includes: Assume that the signal received by the receiving antenna is , in: (7) In (7), Y represents the signal matrix received by the receiving antenna, A represents the angle matrix of each signal, S (7), Y represents the signal matrix received by the receiving antenna, A represents the angle matrix of each signal, S line received data, to Indicates that the antenna array containing Br antennas corresponds to the target index direction to The steering vector at to is the sending signal corresponding to target index 0 to target N, to is the receiving noise corresponding to antenna index 0 to antenna Br.
[0053] In some embodiments, the covariance matrix can be used to perform eigenvalue decomposition to obtain the signal subspace noise subcontrol.
[0054] In some embodiments, the covariance matrix of the received signal is expressed as ;in, , are the covariance matrices of the transmitted signal and noise respectively; through eigenvalue decomposition, we get As the basis of the signal subspace, and containing the remaining eigenvectors as the basis of the noise subspace.
[0055] In some embodiments, the spatial spectrum is represented as: (8) (8) Indicates the direction The basis vectors of the noise subspace at , Indicates direction The corresponding steering vector, the superscript H represents the conjugate transpose operation.
[0056] In some embodiments, when is the signal arrival angle, ; When the spatial spectrum reaches its peak, the spatial spectrum will have N+1 peaks, corresponding to N+1 signals, so as to obtain the signal arrival angle of each path.
[0057] In some embodiments, a threshold-based method may be used to estimate the delay and Doppler corresponding to each path. Estimating the delay and Doppler corresponding to each path includes: Assume that the OTFS transmission signal is: (9) (9) is the element in the delay-Doppler array, It is a DD domain grid with any position The pilot symbols are 0 and They are protection symbol and information data symbol respectively. and s represent the delay and Doppler domain indexes respectively, and Respectively represent the index value of pilot placement, and Respectively represent the maximum delay index and the maximum Doppler index, Indicates data symbol.
[0058] In some embodiments, the received symbol used for channel estimation is represented as: (10) In some embodiments, the threshold is set to ,when , , indicates that there is a path where Indicates delay, represents the Doppler shift, is an indicator variable of 0 or 1, 1 means there is a path, Indicated in , The estimated channel gain at express , The estimation bias at Get the gain of this path ; on the contrary, , indicating that the path does not exist; Select ,in, ; get The delay of the i-th vehicle at time , Doppler , and the angle .
[0059] Step 220: Based on the state information, determine the location information of the target vehicle at at least one time in the first time period; In some embodiments, the location information includes location coordinates of the target vehicle.
[0060] In some embodiments, determining the location information of the target vehicle at least at one time in the first time period includes: Assume the location information of the target vehicle , the speed of the target vehicle’s position , the speed of the target vehicle’s position .
[0061] In some embodiments, since the acceleration of the vehicle is unstable during driving, the acceleration of the vehicle in the x and y directions is assumed to be a random variable, respectively obeying , ,in , is a constant. According to the kinetic equation, we can get: (11) (11) is a time step; In some embodiments, due to a is relatively small, so the acceleration within a time step is constant.
[0062] In some embodiments, the location of the base station is , according to the geometric relationship, the position information corresponding to the i-th vehicle is obtained as (12) (12) , represents the straight-line distance from the base station to the receiving station, represents the propagation distance of the signal reflected by the i-th target; and They correspond to the angle between the direct connection line and the x-axis, and the angle between the connection line between the i-th target and the receiving end and the x-axis.
[0063] Step 230: determining a first position sequence of the target vehicle in the first time period based on the position information of the target vehicle at at least one time point in the first time period; In some embodiments, the first position sequence is a sequence obtained by arranging the position information of the target vehicle at least at one time in the first time period in time in a time sequence. The time sequence may be a time sequence in positive order or a time sequence in reverse order. For example, the first position sequence may include position information 1, position information 2, position information N, etc. in the first time period.
[0064] Step 240: Based on the first position sequence, predict a second position sequence of the target vehicle in a second time period by using a prediction model; In some embodiments, the second time period may be a certain time period in the future. For example, the second time period may be a time period in the future 10S. In some embodiments, the second time period may be a certain time period after the first time period.
[0065] In some embodiments, the second position sequence is a sequence obtained by arranging the position information of at least one moment in the second time period in chronological order. For example, the second position sequence may include position information 1, position information 2, position information N, etc. in a certain time period in the future. In some embodiments, the second position sequence includes position information of at least one future moment.
[0066] The prediction model may be any model for predicting the vehicle position. In some embodiments, the prediction model may include a machine learning model. Specifically, the prediction model may include one or more of a regression model, a decision tree, a neural network, etc.
[0067] In some embodiments, the processing device 112 may input the position of the target vehicle at least one time in the first time period into the prediction model. Furthermore, the processing device 112 may predict the position sequence of the target vehicle in the second time period through the prediction model.
[0068] In some embodiments, the prediction model may be an LSTM neural network model; Figure 3 is a schematic diagram of an LSTM neural network model according to some embodiments of this specification; like Figure 3 As shown in the figure, the LSTM neural network consists of an input layer, three LSTM layers, a fully connected layer, and an output layer.
[0069] In some embodiments, assume that the mini-batch size is , the tensor shape of each layer is shown in the following table: , Let the input layer be , indicating that it includes Vehicle location information at the moment.
[0070] In some embodiments, since the position sequence has long-term dependencies, an LSTM neural network can be used for trajectory prediction.
[0071] In some embodiments, in the LSTM layer, and is the hidden state and cell state of the nth LSTM layer at the mth time step, and the two have the following relationship: (13) (13) is the output gate, which is determined by the input at the current time m and the hidden state at the previous time m-1. More information about input information is as follows Figure 3 shown.
[0072] In some embodiments, the hidden state of the last time step of the last layer is the output of the entire LSTM layer, which is expressed by the following formula: (14) (14) It is a parameter that summarizes the characteristics of the entire LSTM layer. After three layers of LSTM, it enters the fully connected layer; the fully connected layer is used to calculate the output of the previous layer. Perform linear operations and finally output the predicted position information ; In some embodiments, the output layer is characterized by: (15) (15) is the weight, is bias; The output layer is represented as: (16) (16) is a function that summarizes the characteristics of the entire network, and the parameters .
[0073] In some embodiments, the processing device 112 may obtain a training data set, and then use the training data set to train an initial model to obtain a trained prediction model. The training data set may include features and / or labels of multiple training samples. In some embodiments, the features of each training sample may include sample location information, and the label of each training sample may include the actual trajectory of the target vehicle.
[0074] For network training, the error of multiple training samples can be reflected by the loss function. In some embodiments, the loss function can be constructed by using any one of the prediction evaluation indicators such as Mean Absolute Error (MAE), Mean Square Error (MSE), and Root Mean Square Error (RMSE).
[0075] In some embodiments, the LSTM neural network model uses a mean square error (MSE) loss function to reflect the errors of multiple training samples. In some embodiments, the mean square error (MSE) loss function is: (17) (17) , They represent the label and input of the nth training sample respectively, and M is the number of samples.
[0076] In some embodiments, the parameters are updated by applying a back-propagation algorithm and Make After minimization and multiple training iterations, we finally obtain a well-trained LSTM neural network model that can perform location prediction well.
[0077] For example: , receiving antenna , transmitting antenna , the size of the OTFS frame is , carrier frequency GHZ. Variance =1, =0.1, initial velocity =0.5m / s, =0. The root mean square error (RMSE) is used as an indicator to evaluate the performance of the method proposed in this section. For the training of the neural network, the Adam optimizer is used, the batch size is 64, the learning rate is 0.01 and the training is performed 20 times. A time step is set to 0.03s and 1000 trajectories are generated, each containing 1000 time steps. 80% is set as the training set and 20% is set as the test set. Finally, the data is normalized.
[0078] Figure 4 is a comparison diagram of the actual trajectory and predicted trajectory of the target vehicle according to some embodiments of this specification; please refer to Figure 4 ,It can be observed that the trajectory prediction of the target is effectively achieved through the LSTM network model.
[0079] Figure 5A schematic diagram of RMSE changing over time according to some embodiments of this specification; Figure 6 It is a schematic diagram of RMSE of predicted trajectories under different historical time steps according to some embodiments of this specification; See also Figure 5 and Figure 6 ,exist Figure 5 The trajectory prediction performance of the proposed scheme and the baseline scheme are compared in . The proposed scheme uses the historical time step Figure 4 is a comparison diagram of the actual trajectory and predicted trajectory of the target vehicle according to some embodiments of this specification; please refer to Figure 4 ,It can be observed that the trajectory prediction of the target is effectively achieved through the LSTM network model.
[0080] Figure 5 A schematic diagram of RMSE changing over time according to some embodiments of this specification; Figure 6 It is a schematic diagram of RMSE of predicted trajectories under different historical time steps according to some embodiments of this specification; See also Figure 5 and Figure 6 ,exist Figure 5 The trajectory prediction performance of the proposed scheme and the baseline scheme are compared in . The proposed scheme uses the historical time step , the baseline scheme uses the historical time step It can be observed that the proposed scheme is able to accurately predict the trajectory with low RMSE. This is because LSTM has excellent performance in predicting long-term sequences.
[0081] Furthermore, the discussion is about the The RMSE of the predicted trajectory is The relevant circumstances, Is the error The variance of , which is a Gaussian distribution variable with mean zero. Figure 6 It can be seen that the performance of the proposed scheme increases with The reason is that as As the historical trajectory information increases, the accuracy increases. However, it can also be seen that when >6 o'clock, with As increases, the rate of decrease of RMSE slows down. This is because when Up to a certain point, as the historical time step increases, the additional historical information obtained becomes redundant and contributes little to improving the accuracy of trajectory prediction.
[0082] Obviously, the error includes not only the noise error, but also the estimation error from MUSIC and threshold-based algorithms. However, the RMSE can still reach , which proves the effectiveness of the proposed scheme.
[0083] The present specification provides a vehicle trajectory prediction device, which includes at least one processor and at least one memory; the at least one memory is used to store computer instructions; and the at least one processor is used to execute at least part of the computer instructions to implement the aforementioned vehicle trajectory prediction method.
[0084] This specification provides a computer-readable storage medium, which stores computer instructions. When a computer reads the computer instructions in the storage medium, the computer executes the vehicle trajectory prediction method as described above.
[0085] The basic concepts have been described above. Obviously, for those skilled in the art, the above detailed disclosure is only for example and does not constitute a limitation of this specification. Although not explicitly stated here, those skilled in the art may make various modifications, improvements and corrections to this specification. Such modifications, improvements and corrections are suggested in this specification, so such modifications, improvements and corrections still belong to the spirit and scope of the exemplary embodiments of this specification.
[0086] At the same time, this specification uses specific words to describe the embodiments of this specification. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of this specification. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures or characteristics in one or more embodiments of this specification can be appropriately combined.
[0087] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.
Claims
1. A vehicle trajectory prediction method, characterized in that: include: Get the status information of the target vehicle, The target vehicle is a vehicle for which vehicle trajectory prediction is required; Based on the state information, determining the position information of the target vehicle at least at one time in a first time period; Based on the position information of the target vehicle at at least one time in the first time period, a first position sequence of the target vehicle in the first time period is determined, The first position sequence is a sequence obtained by arranging the position information of the target vehicle at least at one moment in the first time period in chronological order; Based on the first position sequence, predicting a second position sequence of the target vehicle in a second time period by using a prediction model; The second time period is a time period after the first time period; The second position sequence is a sequence obtained by arranging the position information of the target vehicle at at least one moment in the second time period in chronological order; Wherein, the time sequence is a positive time sequence or a reverse time sequence; The obtaining of the state information of the target vehicle comprises: receiving a continuous time signal from a target vehicle; Determine the state information of the target vehicle based on the continuous time signal; The prediction model is an LSTM neural network model.
2. A vehicle trajectory prediction method according to claim 1, characterized in that: Get the status information of the target vehicle The determining of the state information of the target vehicle includes: determining the signal arrival angle of each path, and the time delay and Doppler corresponding to each path.
3. A vehicle trajectory prediction method according to claim 2, characterized in that: Estimating the signal arrival angle comprises: Assume that the signal received by the receiving antenna is , ,in: (7) In (7), Y represents the signal matrix received by the receiving antenna, A represents the angle matrix of each signal, S represents the transmitted signal matrix, and N represents the noise matrix; to Indicates the received data from the first to the Brth antennas, to Indicates that the antenna array containing Br antennas corresponds to the target index direction to The steering vector at to is the sending signal corresponding to target index 0 to target N, to is the receiving noise corresponding to antenna index 0 to antenna Br; The covariance matrix of the received signal is expressed as ; in, , are the covariance matrices of the transmitted signal and noise respectively; through eigenvalue decomposition, we get As the basis of the signal subspace, and containing the remaining eigenvectors As a basis of the noise subspace; The spatial spectrum is expressed as: (8) (8) Indicates the direction The basis vectors of the noise subspace at , Indicates direction The corresponding steering vector, the superscript H indicates the conjugate transpose operation; when is the signal arrival angle, ; When the spatial spectrum reaches its peak, the spatial spectrum will have N+1 peaks, corresponding to N+1 signals, so as to obtain the signal arrival angle of each path.
4. A vehicle trajectory prediction method according to claim 2, characterized in that: Estimating the delay and Doppler corresponding to each path includes: Assume that the OTFS transmission signal is: (9) (9) is the element in the delay-Doppler array, It is a DD domain grid with any position The pilot symbols are 0 and They are protection symbol and information data symbol respectively. and s represent the delay and Doppler domain indices respectively, and Respectively represent the index value of pilot placement, and Respectively represent the maximum delay index and the maximum Doppler index, Represents data symbol; The received symbol used for channel estimation is represented as: (10) Set the threshold value to ,when , , indicating that there is a path; in, Indicates delay, represents the Doppler shift, is an indicator variable of 0 or 1, 1 means there is a path, Indicated in , The estimated channel gain at express , The estimation bias at Get the gain of this path , on the contrary, , indicating that the path does not exist; Select ,in, , get The delay of the i-th vehicle at time , Doppler , and the angle .
5. A vehicle trajectory prediction method according to claim 1, characterized in that: Determining the location information of the target vehicle at least at one time in the first time period includes: Assume the location information of the target vehicle, , the speed of the target vehicle’s position , the speed of the target vehicle’s position ; Assume that the acceleration of the vehicle in the x and y directions is a random variable, which is subject to , ,in , is a constant; According to the kinetic equation, we can get: (11) (11) is a time step; Assume the location of the base station is , according to the geometric relationship, the position information corresponding to the i-th vehicle is obtained as (12) (12) ; represents the straight-line distance from the base station to the receiving station, represents the propagation distance of the signal reflected by the i-th target; and They correspond to the angle between the direct connection line and the x-axis, and the angle between the connection line between the i-th target and the receiving end and the x-axis.
6. A vehicle trajectory prediction method according to claim 1, characterized in that: The LSTM neural network model includes: an input layer, three LSTM layers, a fully connected layer and an output layer; Let the input layer be , indicating that it includes Vehicle location information at the moment; In the LSTM layer, let and is the hidden state and cell state of the nth LSTM layer at the mth time step, and the two have the following relationship: (13) (13) is the output gate, which is determined by the input at the current time m and the hidden state at the previous time m-1; The hidden state of the last time step of the last layer is the output of the entire LSTM layer, which is expressed by the following formula: (14) (14) It is a parameter that summarizes the characteristics of the entire LSTM layer. A nonlinear function; after three layers of LSTM, it enters the fully connected layer; the fully connected layer outputs the previous layer Perform linear operations and finally output the predicted position information ; The characteristics of the output layer are: (15) (15) is the weight, is bias; The output layer is represented as: (16) (16) is a function that summarizes the characteristics of the entire network, and the parameters .
7. A vehicle trajectory prediction method according to claim 6, characterized in that: The LSTM neural network model uses a mean square error loss function to reflect the errors of multiple training samples. The mean square error loss function is: (17) (17) , They represent the label and input of the nth training sample respectively, and M is the number of samples.
8. A vehicle trajectory prediction system, characterized in that: The method comprises a server, a network, a user terminal, a vehicle and a storage device, wherein the server obtains information and / or data of the vehicle via the network, the server comprises a processing device, the processing device processes information and / or data related to a service request, the network is used to facilitate the exchange of information and / or data, the storage device is used to store data obtained from the user terminal, and the user terminal receives service confirmation and / or information or instructions from the server, so as to execute a vehicle trajectory prediction method as described in any one of claims 1 to 7.
9. A vehicle trajectory prediction device, characterized in that: The device includes at least one processor and at least one memory; the at least one memory is used to store computer instructions; and the at least one processor is used to execute at least part of the computer instructions to implement a vehicle trajectory prediction method as described in any one of claims 1-7.
10. A computer-readable storage medium, characterized in that: The storage medium stores computer instructions. When a computer reads the computer instructions in the storage medium, the computer executes a vehicle trajectory prediction method as described in any one of claims 1 to 7.
Citation Information
Patent Citations
Methods for trajectory prediction and methods for the automated operation of a vehicle
DE102021005625B3
Integrated trajectory forecasting, error estimation, and vehicle handling when detecting an observed scenario
US20230202470A1
Neural network trajectory prediction
US20230391374A1
Access node, user equipment, apparatus, method and computer program for determining a delay-doppler-resolution for a transceiver of a mobile communication system
WO2022096163A1
Methods and apparatus for selecting a base station in a heterogeneous network
WO2022217618A1
Cited By
Vehicle track prediction method, electronic equipment, chip system and storage medium
CN120416776A
A vehicle trajectory prediction method, electronic device, chip system and storage medium
CN120416776B
