Information physical twinning probabilistic surrounding vehicle interactive prediction model identification method

Abstract

The invention relates to the technical field of automatic driving application, in particular to a cyber-physical twinborn probabilistic surrounding vehicle interactive prediction model identification method, which comprises the following steps of: in an establishment stage of a probabilistic surrounding vehicle interactive prediction model, fitting parameters of the probabilistic surrounding vehicle interactive prediction model by utilizing a preset data set to obtain a probabilistic surrounding vehicle interactive prediction model; in the operation stage of the probabilistic surrounding vehicle interaction prediction model, updating an algorithm and a preset updating trigger mechanism based on preset parameters and intentions, and updating the target vehicle to obtain the target vehicle intent parameter of the probabilistic surrounding vehicle interaction prediction model; and updating the parameters of the probabilistic surrounding vehicle interaction prediction model and the intention and uncertainty of the target vehicle by using real-time sensing data. Therefore, the problem that effective modeling and real-time updating are difficult to carry out on complex uncertainty in long-time-domain interaction prediction in the prior art is solved, surrounding vehicle interaction behavior prediction can be carried out in a longer time domain, and meanwhile a clear interaction basis is provided.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving application technology, and in particular to a method for identifying a cyber-physical twin probabilistic week-car interactive prediction model. Background Art

[0002] The integration of vehicle, road and cloud and the intelligent cyber-physical system (IVCPS) built on it are new ways to solve the above problems. The integrated control of vehicle, road and cloud can support and supplement the autonomous driving of a single vehicle, and use the advantages of the cloud control system based on wide-area perception on the roadside, the large parallel computing power that can be dispatched in a short time on the cloud server, and the ability to coordinate multiple vehicles and even the entire intelligent networked vehicle traffic flow for centralized decision-making to assist the single vehicle in making more energy-saving and efficient decisions and planning.

[0003] Under the vehicle-road-cloud integrated architecture, it is necessary to establish real-time twins of traffic conditions and all traffic participants in the cloud to implement more applications based on long-term predictive information, so as to better ensure vehicle safety and make better decisions. However, the current twins established in the cloud are generally deterministic, or only contain simple probabilistic information, and it is difficult to model and adjust the complex uncertainties and their propagation in long-term interactive predictions in real time. Therefore, how to establish a model that can concisely and effectively represent long-term interactive predictions and update the uncertainty of model parameters and intentions in real time has become an urgent problem to be solved.

[0004] With the rapid development of autonomous driving technology, behavioral decision-making has become an indispensable part of realizing autonomous driving functions. The key to behavioral decision-making lies in how to effectively handle the interaction between the ego vehicle and surrounding vehicles. The traditional rule-based solution has the logical fallacy of prediction first and then decision-making. This is because the prediction results for surrounding vehicles should be different under different decisions of the ego vehicle. At present, single-vehicle autonomous driving is mainly divided into two categories: modular autonomous driving and end-to-end autonomous driving. Among them, the latest modular autonomous driving solution handles the interaction problem between the ego vehicle and surrounding vehicles through a variety of strategies such as explicit interaction, implicit interaction, and comprehensive interaction; while end-to-end autonomous driving solves the interaction problem through overall collaborative optimization from perception to planning results, as well as multi-objective and multi-modal methods.

[0005] Although the above methods perform well in dealing with the interaction between the ego vehicle and surrounding vehicles in a short time domain, single-vehicle autonomous driving is usually limited by the perception and computing limitations of the ego vehicle, and it is still easy to fall into local optimal decisions, and it is difficult to consider the dynamic traffic conditions in a longer time domain to make better decisions and plans. The reason for this limitation is that the perception range of the ego vehicle is short, and it is difficult to obtain perception results at a longer distance; and even if a larger range of perception is obtained through cloud information platforms such as vehicle-road collaboration or map navigation, it is also limited by the small computing power of the vehicle and the difficulty in well representing the uncertainty changes in the movement of surrounding vehicles as the prediction time increases.

[0006] The integration of vehicle, road and cloud and the intelligent cyber-physical system (IVCPS) built on it are new ways to solve the above problems. The integrated control of vehicle, road and cloud can support and supplement the autonomous driving of a single vehicle, and use the advantages of the cloud control system based on wide-area perception on the roadside, the large parallel computing power that can be dispatched in a short time on the cloud server, and the ability to coordinate multiple vehicles and even the entire intelligent networked vehicle traffic flow for centralized decision-making to assist the single vehicle in making more energy-saving and efficient decisions and planning.

[0007] Under the vehicle-road-cloud integrated architecture, it is necessary to establish real-time twins of traffic conditions and all traffic participants in the cloud to implement more applications based on long-term predictive information, so as to better ensure vehicle safety and make better decisions. However, the current twins established in the cloud are generally deterministic, or only contain simple probabilistic information, and it is difficult to model and adjust the complex uncertainties and their propagation in long-term interactive predictions in real time. Therefore, how to establish a model that can concisely and effectively represent long-term interactive predictions and update the uncertainty of model parameters and intentions in real time has become an urgent problem to be solved. Summary of the invention

[0008] The present application provides a method for identifying a probabilistic week-vehicle interaction prediction model of an information-physical twin to solve the problem that the existing technology is difficult to effectively model and update in real time the complex uncertainties in long-term interaction predictions. It can predict week-vehicle interaction behaviors over a longer time domain and provide a clear basis for interaction.

[0009] The first aspect of the present application provides a method for identifying a probabilistic weekly vehicle interactive prediction model of a cyber-physical twin, including the following steps:

[0010] Obtain real-time perception data of the target vehicle;

[0011] In the establishment stage of the probabilistic vehicle-to-vehicle interaction prediction model, the parameters of the probabilistic vehicle-to-vehicle interaction prediction model are fitted using a preset data set to obtain the initialized probabilistic vehicle-to-vehicle interaction prediction model parameters and the initialized target vehicle intention parameters, wherein the probabilistic vehicle-to-vehicle interaction prediction model includes a driver model, a probabilistic lane-changing model and an uncertainty accumulation mechanism, and the probabilistic vehicle-to-vehicle interaction prediction model parameters include driver model parameters and probabilistic lane-changing model parameters;

[0012] During the operation phase of the probabilistic vehicle-to-vehicle interaction prediction model, the parameters of the probabilistic vehicle-to-vehicle interaction prediction model, the intention and uncertainty of the target vehicle are updated using the real-time perception data based on a preset parameter and intention update algorithm and a preset update trigger mechanism.

[0013] According to an embodiment of the present application, the method of fitting the parameters of the probabilistic vehicle-to-vehicle interaction prediction model using a preset data set to obtain the initialized probabilistic vehicle-to-vehicle interaction prediction model parameters and the initialized intention parameters of the target vehicle includes:

[0014] Based on the weighted bi-norm loss function, the driver model parameters are fitted, and based on the cross entropy loss function, the probabilistic lane-changing model parameters are fitted to obtain the initialized probabilistic vehicle-around-vehicle interaction prediction model parameters and the initialized target vehicle intention parameters.

[0015] According to an embodiment of the present application, the updating algorithm based on preset parameters and intention and the preset update trigger mechanism, using the real-time perception data to update the parameters of the probabilistic vehicle-to-vehicle interaction prediction model, the intention and uncertainty of the target vehicle, includes:

[0016] Based on a preset posterior distribution calculation method, jointly sampling the driver model parameter and the intention parameter, and calculating a first posterior probability according to a first sampling result;

[0017] Based on a preset posterior distribution calculation method, the probabilistic lane change model parameters are separately sampled, and a second posterior probability is calculated according to a second sampling result;

[0018] updating the probabilistic vehicle-to-vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle according to the first posterior probability and / or the second posterior probability;

[0019] The calculation formula of the first posterior probability is:

[0020]

[0021] Among them, N is the sampling time, X pred,i(θ, I) is the state predicted based on the driver model parameters and intention (θ, I), X obsv,i is the actual observed state, σ is the prior variance between the predicted state and the observed state;

[0022] The calculation formula of the second posterior probability is:

[0023]

[0024] Where C is the number of samples, P(l) is the probability of left lane change predicted by the model, and p L|w is the probability of changing lanes to the left calculated based on parameter w, P(K) is the probability of not changing lanes predicted by the model, and p K|w is the probability of not changing lanes calculated based on parameter w, P(R) is the probability of changing lanes right predicted by the model, and p R|w is the probability of right lane change calculated based on parameter w.

[0025] According to an embodiment of the present application, the updating algorithm based on preset parameters and intention and the preset update trigger mechanism, using the real-time perception data to update the parameters of the probabilistic vehicle-to-vehicle interaction prediction model, the intention and uncertainty of the target vehicle, also includes:

[0026] If the joint sampling of the driver model parameters and the intention parameters meets the preset time interval condition, the probabilistic vehicle-to-vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle are updated based on the time update trigger mechanism;

[0027] If the target vehicle meets the preset lane changing condition or the prediction result of the probabilistic surrounding vehicle interaction prediction model meets the preset result, the probabilistic surrounding vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle are updated based on the event update trigger mechanism.

[0028] According to one embodiment of the present application, before acquiring the real-time perception data of the target vehicle, the method further includes:

[0029] A representation of longitudinal motion, lateral motion, and uncertainty accumulation is established for each vehicle to be interactively predicted.

[0030] According to the information-physical twin probabilistic surrounding vehicle interaction prediction model identification method of the embodiment of the present application, in the establishment stage of the probabilistic surrounding vehicle interaction prediction model, the parameters of the probabilistic surrounding vehicle interaction prediction model are fitted using a preset data set to obtain the initialized probabilistic surrounding vehicle interaction prediction model parameters and the initialized target vehicle intention parameters. In the operation stage of the probabilistic surrounding vehicle interaction prediction model, based on the preset parameter and intention update algorithm and the preset update trigger mechanism, the real-time perception data is used to update the probabilistic surrounding vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle. Thus, the problem that the existing technology is difficult to effectively model and update the complex uncertainty in the long-term interaction prediction in real time is solved, and the surrounding vehicle interaction behavior prediction can be performed in a longer time domain, while providing a clear basis for interaction.

[0031] The second aspect of the present application provides a cyber-physical twin probabilistic weekly vehicle interactive prediction model recognition device, including:

[0032] An acquisition module, used to acquire real-time perception data of the target vehicle;

[0033] A fitting module is used to fit the parameters of the probabilistic vehicle interaction prediction model using a preset data set during the establishment phase of the probabilistic vehicle interaction prediction model to obtain the initialized probabilistic vehicle interaction prediction model parameters and the initialized target vehicle intention parameters, wherein the probabilistic vehicle interaction prediction model includes a driver model, a probabilistic lane change model and an uncertainty accumulation mechanism, and the probabilistic vehicle interaction prediction model parameters include driver model parameters and probabilistic lane change model parameters;

[0034] An updating module is used to update the parameters of the probabilistic vehicle interaction prediction model, the intention and uncertainty of the target vehicle using the real-time perception data based on a preset parameter and intention update algorithm and a preset update trigger mechanism during the operation phase of the probabilistic vehicle interaction prediction model.

[0035] According to one embodiment of the present application, the fitting module is used to:

[0036] Based on the weighted bi-norm loss function, the driver model parameters are fitted, and based on the cross entropy loss function, the probabilistic lane-changing model parameters are fitted to obtain the initialized probabilistic vehicle-around-vehicle interaction prediction model parameters and the initialized target vehicle intention parameters.

[0037] According to one embodiment of the present application, the update module is used to:

[0038] Based on a preset posterior distribution calculation method, jointly sampling the driver model parameter and the intention parameter, and calculating a first posterior probability according to a first sampling result;

[0039] Based on a preset posterior distribution calculation method, the probabilistic lane change model parameters are separately sampled, and a second posterior probability is calculated according to a second sampling result;

[0040] updating the probabilistic vehicle-to-vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle according to the first posterior probability and / or the second posterior probability;

[0041] The calculation formula of the first posterior probability is:

[0042]

[0043] Among them, N is the sampling time, X pred,i (θ, I) is the state predicted based on the driver model parameters and intention (θ, I), X obsv,i is the actual observed state, σ is the prior variance between the predicted state and the observed state;

[0044] The calculation formula of the second posterior probability is:

[0045]

[0046] Where C is the number of samples, P(L) is the probability of left lane change predicted by the model, and p L|w is the probability of changing lanes to the left calculated based on parameter w, P(K) is the probability of not changing lanes predicted by the model, and p K|w is the probability of not changing lanes calculated based on parameter w, P(R) is the probability of changing lanes right predicted by the model, and p R|w is the probability of right lane change calculated based on parameter w.

[0047] According to one embodiment of the present application, the update module is further used to:

[0048] If the joint sampling of the driver model parameters and the intention parameters meets the preset time interval condition, the probabilistic vehicle-to-vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle are updated based on the time update trigger mechanism;

[0049] If the target vehicle meets the preset lane changing condition or the prediction result of the probabilistic surrounding vehicle interaction prediction model meets the preset result, the probabilistic surrounding vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle are updated based on the event update trigger mechanism.

[0050] According to one embodiment of the present application, before acquiring the real-time perception data of the target vehicle, the acquisition module is further used to:

[0051] A representation of longitudinal motion, lateral motion, and uncertainty accumulation is established for each vehicle to be interactively predicted.

[0052] According to the information-physical twin probabilistic surrounding vehicle interaction prediction model identification device of the embodiment of the present application, in the establishment stage of the probabilistic surrounding vehicle interaction prediction model, the parameters of the probabilistic surrounding vehicle interaction prediction model are fitted using a preset data set to obtain the initialized probabilistic surrounding vehicle interaction prediction model parameters and the initialized target vehicle intention parameters. In the operation stage of the probabilistic surrounding vehicle interaction prediction model, based on the preset parameter and intention update algorithm and the preset update trigger mechanism, the real-time perception data is used to update the probabilistic surrounding vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle. Thus, the problem that the existing technology is difficult to effectively model and update the complex uncertainty in the long-term interaction prediction in real time is solved, and the surrounding vehicle interaction behavior prediction can be performed in a longer time domain, while providing a clear basis for interaction.

[0053] The third aspect of the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the information-physical twin probabilistic weekly vehicle interactive prediction model identification method as described in the above embodiment.

[0054] The fourth aspect of the present application provides a computer-readable storage medium on which a computer program is stored, and the program is executed by a processor to implement the information-physical twin probabilistic weekly vehicle interactive prediction model identification method as described in the above embodiment.

[0055] The fifth aspect of the present application proposes a computer program product, including a computer program, which, when executed by a processor, is used to implement the information-physical twin probabilistic weekly vehicle interactive prediction model identification method as described in the above embodiment.

[0056] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0058] Figure 1 A schematic diagram of the structure of a cyber-physical twin probabilistic weekly vehicle prediction model establishment and identification system according to an embodiment of the present application;

[0059] Figure 2 A schematic diagram of a method for constructing a longitudinal driver model according to an embodiment of the present application;

[0060] Figure 3 Schematic diagram of a method for constructing a lateral probabilistic lane-changing model according to an embodiment of the present application;

[0061] Figure 4 A schematic diagram illustrating the propagation of uncertainty according to an embodiment of the present application;

[0062] Figure 5 A flowchart of a cyber-physical twin probabilistic weekly vehicle interactive prediction model identification method provided according to an embodiment of the present application;

[0063] Figure 6 A schematic diagram of a method for establishing a parameter uncertainty model according to an embodiment of the present application;

[0064] Figure 7 A schematic diagram of a method for identifying intent and its uncertainty model establishment phase according to an embodiment of the present application;

[0065] Figure 8 A schematic diagram of a method for identifying a parameter and intention uncertainty model operation phase according to an embodiment of the present application;

[0066] Fig. 9 It is a block diagram of a device for identifying a probabilistic weekly vehicle interactive prediction model of a cyber-physical twin according to an embodiment of the present application;

[0067] Fig.10 Schematic diagram of the structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0068] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0069] The following describes the information-physical twin probabilistic weekly vehicle interactive prediction model identification method according to an embodiment of the present application with reference to the accompanying drawings.

[0070] Before introducing the information-physical twin probabilistic cycle-car interactive prediction model identification method of the present application, we first introduce the information-physical twin probabilistic cycle-car interactive prediction model establishment and identification system and the information-physical twin probabilistic cycle-car interactive prediction model establishment method involved in the information-physical twin probabilistic cycle-car interactive prediction model identification method of the present application. Among them, the probabilistic cycle-car interactive prediction model establishment and identification method used by the system is the probabilistic cycle-car interactive prediction model establishment and identification method proposed in the present application, and the specific implementation steps of the method will be described in detail below.

[0071] First, in order to realize the establishment and identification of the cyber-physical twin probabilistic weekly vehicle interactive prediction model, this paper proposes a cyber-physical twin probabilistic weekly vehicle interactive prediction model establishment and identification system based on the vehicle-road-cloud integrated architecture. Figure 1 As shown in the figure, the system includes information space and physical space. The physical space contains user vehicles and other traffic participants who use the service. The information space includes information mapping layer and fusion application layer, both of which are distributed on the vehicle side and the cloud side. In the information mapping layer, the cloud side contains the digital twin of the current traffic participants. The fusion application layer is distributed in the vehicle side controller and the cloud server. The cloud side fusion application layer can establish and identify the weekly vehicle probability model and use it in other applications.

[0072] Furthermore, the information-physical twin probabilistic surrounding vehicle interactive prediction model establishment and identification system of the embodiment of the present application also includes roadside equipment, intelligent networked vehicles, cloud and users. Among them, the roadside equipment includes environmental perception function and communication function, which can identify the global ID of vehicles on the road and the status information of the vehicles and transmit the information; the number of intelligent networked vehicles is arbitrary, including environmental perception function and communication function, which can perceive its surrounding environment and transmit the original perception information or perception target level results; the cloud contains real-time traffic twin and historical database, the real-time traffic twin is obtained by fusion perception processing of the perception data uploaded by the roadside equipment and intelligent networked vehicles, and the historical database is obtained by recording and storing the real-time traffic twin; users can communicate with the cloud, upload the area of ​​interest, and obtain the probabilistic surrounding vehicle interactive prediction model establishment and identification service provided by the cloud.

[0073] Secondly, the method for establishing the information-physical twin probabilistic weekly vehicle interactive prediction model of this application is introduced.

[0074] Specifically, the cyber-physical twin probabilistic weekly vehicle interactive prediction model establishes representations of longitudinal motion, lateral motion and uncertainty accumulation for each vehicle that needs to be interactively predicted. The model for one of the vehicles specifically includes: for longitudinal motion, a rule-based model is used and its parameters are considered to conform to the normal distribution; for lateral motion, a model that outputs the probability of predefined semantic behavioral decisions (left lane change, right lane change, lane keeping) is used; the uncertainty accumulation mechanism takes into account parameter uncertainty, lane change uncertainty of the vehicle in front of the vehicle, and uncertainty of the vehicle's intention.

[0075] Among them, the rule-based model is the driver model, and the driver model is continuously and partially differentiable everywhere. A modified intelligent driver model (IDM) can be used, which is in the form of:

[0076]

[0077]

[0078] Among them, s * is the ideal following distance, s min is the minimum distance, v ego is the vehicle speed, T hw is the expected headway, v p is the speed of the front vehicle, a max is the maximum acceleration, b comf is the comfortable deceleration, a pred is the driver model, v des,ego is the expected speed, δ is the driver’s aggressiveness, and s is the actual distance.

[0079] Among them, the softplus(*) function is:

[0080] softplus(x)=log(1+e x )(3)

[0081] like Figure 2 As shown in Figure 1, replacing ReLU(*) with softplus(*) can make the IDM model continuous and differentiable everywhere, thereby obtaining the analytical gradient of each parameter. The parameters θ that need to be estimated include T hw Expected headway, a max Maximum acceleration, b comf Comfortable deceleration, s min Minimum distance, v des,ego The vector consisting of the desired speed and the driver’s aggressiveness δ.

[0082] Furthermore, the model that outputs the probability of predefined semantic behavior decisions is a probabilistic lane change model, which outputs the probabilities of different lane change decisions, specifically a modification of the MOBIL model. Using the widely recognized lane change model MOBIL model as a basis, the acquisition of the probability of defining a vehicle L, K, R (left lane change, lane keeping, right lane change) includes the following steps:

[0083] Step 1: Use the MOBIL model idea to obtain the lane change acceleration vector a, including Δa L,ego (acceleration gain of the vehicle changing lanes to the left), a K,ego =0 (acceleration gain of the vehicle not changing lanes), Δa R,ego (acceleration gain of the vehicle changing lanes to the right), Δa L,f (the algebraic sum of the acceleration gains of the vehicle behind the original lane and the new vehicle behind the left lane after the vehicle changes lanes to the left), Δa K,f =0 (the benefit of other vehicles when the vehicle is keeping the lane), Δa R,f(The algebraic sum of the acceleration gains of the vehicle behind the original lane and the new vehicle behind the right lane after the ego vehicle changes lanes to the right) The expected acceleration gains of the ego vehicle and the vehicle behind the lane after the lane change corresponding to the ego vehicle's LKR action (calculated using the identified driver model of each vehicle). For the ego vehicle, it is the difference between the acceleration after the lane change and the acceleration before the lane change; for the vehicle behind the lane, it is the sum of the loss of the new vehicle behind the lane after the lane change due to the ego vehicle's lane change and the gain of the original vehicle behind the lane before the lane change due to the ego vehicle's lane change.

[0084] Step 2: Design the evaluation function: The evaluation function f(a,p) needs to be able to reflect the vehicle's preference for different lane-changing options. Generally, the greater the acceleration, the more the vehicle tends to choose this behavior. It can be defined as:

[0085]

[0086] Among them, α L ,β L ,α R ,β R ,p A total of 5 parameters are model parameters to be identified. The common parameter p is the politeness coefficient corresponding to the MOBIL model, which reflects the degree of attention that this car attaches to the impact of its own car on other cars.

[0087] Step 3: Get the lane-changing probability of the vehicle. Use the softmax(*) function to normalize the output of the evaluation function to obtain the probability of changing lanes to the left or right or not changing lanes at this moment.

[0088]

[0089] Among them, P(L) is the probability of changing lanes to the left, P(R) is the probability of changing lanes to the right, and P(K) is the probability of not changing lanes.

[0090] Specifically, if Figure 3 As shown in the figure, it is equivalent to connecting the output of the MOBIL model to the fully connected layer and the Softmax layer, generating the probabilities of the three decisions of changing lanes left, right, and not changing lanes.

[0091] Furthermore, the parameter uncertainty considered by the uncertainty accumulation mechanism is the parameter uncertainty of the driver model. The lane changing uncertainty of the vehicle in front of the vehicle is output through the lateral motion model of the probabilistic vehicle-to-vehicle interactive prediction model established for the vehicle in front of the vehicle. The consideration of the vehicle's intention uncertainty is obtained by calculating the entropy value of the intention probability, and the above uncertainty is accumulated over time through the motion equation (including two state quantities: longitudinal speed and longitudinal position).

[0092] Specifically, the uncertainty accumulation mechanism of the embodiment of the present application is an additive uncertainty accumulation mechanism, and the uncertainty of the output acceleration prediction result is the weighted product of the uncertainty of the driver's parameters of the vehicle, the uncertainty of the lane change of the preceding vehicle and the uncertainty of the state parameters of the two vehicles, and the sum of the uncertainty of the vehicle's intention.

[0093] The propagation of uncertainty between vehicles is considered by Taylor expanding the influence of the driver model and lane change model parameters on the model output results. Figure 4 As shown, the specific implementation is as follows:

[0094] The driver model can be expressed as:

[0095] a pred =f(v p ,v ego ,d intvl ,θ)(6)

[0096] Among them, a pred is the output of the driver model, i.e., the predicted acceleration, f(·) is the driver model, v p is the speed of the front vehicle, v ego is the vehicle speed, d intvl is the distance between the vehicle and the front vehicle, and θ is the parameter vector to be estimated.

[0097] Optionally, it can be simplified to free flow driving when there is no vehicle ahead:

[0098] a pred =f(θ)(7)

[0099] The parameters of the driver model will have variance when identified Through Taylor expansion, the influence of the uncertainty of the leading vehicle state and the uncertainty of the driver parameter θ of the following vehicle on the speed can be approximated as:

[0100]

[0101] in, is the variance of the predicted acceleration, is the variance of the preceding vehicle’s speed, is the spacing variance, is the variance of parameter i, θ i is the ith parameter.

[0102] Among them, the uncertainty of the spacing variance is the sum of the uncertainty of the position states of the two vehicles:

[0103]

[0104] in, is the variance of the vehicle position, is the variance of the preceding vehicle’s position.

[0105] Furthermore, the longitudinal uncertainty including lane change uncertainty can be calculated as follows. Since the vehicle ahead may change lanes and leave (1-P(L)) or stay in the lane (P(K)), the longitudinal uncertainty is accumulated and calculated as:

[0106]

[0107] Among them, the subscript pp is the preceding vehicle of the preceding vehicle, For the front car, is the distance variance of the preceding vehicle.

[0108] When considering intention uncertainty, since there is sufficient reason to assume that the intentions of different vehicles are independent of each other, the calculation of uncertainty increases the product of intention entropy. Let:

[0109]

[0110] in, is the corrected front vehicle speed variance, is the corrected inter-vehicle distance variance, σ ego is the ego vehicle parameter uncertainty.

[0111] Then the following formula holds:

[0112]

[0113] Among them, H p is the lane-changing decision entropy of the preceding vehicle, H ego is the lane-changing decision entropy of the ego vehicle.

[0114] The intention entropy H of a certain car is calculated as follows: the intention includes N possible intentions of this car I = {I 1 ,I 2 ,…,I N}

[0115]

[0116] It can be assumed that the intentions of different vehicles are independent of each other because the macro intention depends on the path planning of the vehicle to the destination, and there will be basically no mutual influence between the vehicles.

[0117] Furthermore, in the absence of state updates, the propagation value of uncertainty depends on the prediction step, which is achieved through the following steps: In each time step Δt, the state of the vehicle is predicted by the acceleration model, and the state vector x includes position and velocity. The state transition is as follows:

[0118]

[0119] Right now:

[0120] x=Ax+Ba pred (15)

[0121] Perform ZOH discretization to obtain the discrete state equation:

[0122] x k+1 =Gx k +Ha pred,k (16)

[0123] Among them, x k+1 is the state quantity at time k+1, G is the state transfer matrix, x k is the state quantity at time k, a pred,k is the predicted acceleration at time k.

[0124] At each time step, the uncertainty of the state is propagated, and this uncertainty affects the uncertainty of position and velocity through the state transfer equation. Therefore, the uncertainty propagation of the leading vehicle is described by the following equation:

[0125] P k+1 =GP k G+Q (17)

[0126] Among them, P k+1 is the uncertainty at time k+1, P k is the uncertainty at time k.

[0127] Where Q is the process noise covariance matrix, defined as:

[0128]

[0129] in, It represents the effect of acceleration on position variance. Since position is twice the integral of acceleration, the variance is proportional to the fourth power of the sampling time. It represents the effect of acceleration on the unknown and velocity covariance. Since velocity is the first integral of acceleration and position is the second integral of acceleration, it leads to a cubic time dependence. 2 This indicates that velocity is dependent on the square of acceleration.

[0130] The following introduces the information-physical twin probabilistic week-car interactive prediction model identification method proposed in this application.

[0131] Specifically, Figure 5 A flow chart of a method for identifying a probabilistic weekly vehicle interactive prediction model for an information-physics twin provided in an embodiment of the present application.

[0132] like Figure 5As shown, the cyber-physical twin probabilistic weekly vehicle interactive prediction model identification method includes the following steps:

[0133] In step S501, real-time perception data of the target vehicle is obtained.

[0134] Specifically, the embodiments of the present application can obtain real-time perception data of the target vehicle through vehicle sensors, which is not specifically limited here.

[0135] In step S502, during the establishment phase of the probabilistic vehicle-to-vehicle interaction prediction model, the parameters of the probabilistic vehicle-to-vehicle interaction prediction model are fitted using a preset data set to obtain the initialized probabilistic vehicle-to-vehicle interaction prediction model parameters and the initialized target vehicle intention parameters, wherein the probabilistic vehicle-to-vehicle interaction prediction model includes a driver model, a probabilistic lane change model and an uncertainty accumulation mechanism, and the probabilistic vehicle-to-vehicle interaction prediction model parameters include driver model parameters and probabilistic lane change model parameters.

[0136] The preset data set includes the historical driving status data of one or more traffic vehicles and the records of their interaction with surrounding vehicles. The initialization intention parameters of the target vehicle include the intention and its probability.

[0137] Further, in some embodiments, the parameters of the probabilistic vehicle-to-vehicle interaction prediction model are fitted using a preset data set to obtain the initialized probabilistic vehicle-to-vehicle interaction prediction model parameters and the initialized intention parameters of the target vehicle, including: fitting the driver model parameters based on a weighted bi-norm loss function, and fitting the probabilistic lane-changing model parameters based on a cross-entropy loss function to obtain the initialized probabilistic vehicle-to-vehicle interaction prediction model parameters and the initialized intention parameters of the target vehicle;

[0138] Specifically, if Figure 6 As shown in Figure 2, parameter identification in the model building phase is achieved by fitting the average behavior in the dataset. Weighted bi-norm loss and cross entropy loss are used for the driver model and lane changing model, respectively.

[0139] Among them, the weighted two-norm loss function is:

[0140]

[0141] in, is the weighted two-norm loss, w 1 is the weight of the error between the predicted acceleration and the actual acceleration, a pred is the acceleration output by the driver model, a is the actual observed acceleration, and w 2 is the weight of the error between the predicted velocity obtained by integrating the predicted acceleration and the actual observed velocity, v predis the predicted velocity obtained by integrating the predicted acceleration, v is the actually observed velocity, and w 3 is the weight of the error between the predicted position and the actual position obtained by integrating the predicted speed, s pred is the predicted position obtained by integrating the predicted speed, and s is the actual position.

[0142] The cross entropy loss function is:

[0143]

[0144] in, is the cross entropy loss, P(L) is the probability of left lane change predicted by the model, and p L is the probability of actually changing lanes to the left (because the actual behavior has been determined by observation, it is a quantity that is either 0 or 1, p K 、p P Same as above), P(K) is the probability of not changing lanes predicted by the model, p K is the probability of not changing lanes in reality, P(R) is the probability of changing lanes right predicted by the model, and p R is the probability of actually changing lane right.

[0145] like Figure 7 As shown, for the intention recognition in the model building stage, the intention bifurcation point of the vehicle within a given long distance in the future is determined through map information and traffic rules (K possible intentions are identified, for example, if the vehicle is currently driving on a highway and there is a ramp 500m ahead, then the possible intentions include going straight and going off the ramp; and based on the identified driver model, the vehicle is set to different intentions (each intention has its own prior information, which can be reasonably inferred through the road structure and vehicle position. For example, at the ramp, P(going straight) = 0.25, P(going off the ramp) = 0.75 can be given based on the usual traffic flow + the current lane position of the vehicle + the current acceleration trend of the vehicle).

[0146] In step S503, during the operation phase of the probabilistic vehicle-to-vehicle interaction prediction model, the parameters of the probabilistic vehicle-to-vehicle interaction prediction model, the intention and uncertainty of the target vehicle are updated using real-time perception data based on a preset parameter and intention update algorithm and a preset update trigger mechanism.

[0147] Furthermore, in some embodiments, based on a preset parameter and intention update algorithm and a preset update trigger mechanism, real-time perception data is used to update the probabilistic surrounding vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle, including: based on a preset posterior distribution calculation method, the driver model parameters and the intention parameters are jointly sampled, and a first posterior probability is calculated based on the first sampling result; based on a preset posterior distribution calculation method, the probabilistic lane change model parameters are separately sampled, and a second posterior probability is calculated based on the second sampling result; and the probabilistic surrounding vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle are updated according to the first posterior probability and / or the second posterior probability.

[0148] Specifically, in the model operation phase of the cyber-physical twin probabilistic vehicle-to-vehicle interactive prediction model identification method, the parameter and intention update algorithm is Metropolis Hastings sampling, which samples the driver model parameters and intention parameters jointly according to the defined posterior distribution calculation method, and samples and updates the lane change probability model parameters separately. Figure 8 As shown in Figure 1, the MCMC (Markov Chain Monte Carlo) method in Bayesian estimation is used to sample and update the uncertainty of the driver model parameters and lane change model parameters, i.e., the intention. The MCMC method constructs a Markov chain and samples a series of θ from the posterior distribution to construct the probability distribution of the parameters. The θ obtained by sampling 1 ,θ 2 ,…,θ M , the distribution of the parameters can be estimated. hw Sample value of The sample mean can be used as the estimated value of the parameter, and the sample variance can be used as the uncertainty of the parameter. The remaining question is how to obtain the posterior probability.

[0149] Furthermore, if Figure 8 As shown in the figure, the preset posterior distribution calculation method is: for the joint sampling of the driver model parameters and the intention parameters, the posterior probability π(θ,I) of a set of parameters and intention (θ,I) is calculated as the state X predicted based on the parameters and intention (θ,I) at N sampling moments (N≥1). pred,i The error normal distribution probability product of (θ,I), where X represents the vehicle state, including at least the lateral and longitudinal position and speed, and σ is the prior value.

[0150]

[0151] Among them, N is the sampling time, X pred,i (θ, I) is the state predicted based on the driver model parameters and intention (θ, I), X obsv,i is the actual observed state, and σ is the prior variance between the predicted state and the observed state.

[0152] Furthermore, for the separate sampling of the lane-changing model parameters, the posterior probability calculation method of the parameter w is the probability corresponding to the decision actually made in the probability of the lane-changing decision calculated based on the parameter w.

[0153]

[0154] Where C is the number of samples, P(L) is the probability of left lane change predicted by the model, and p L|w is the probability of changing lanes to the left calculated based on parameter w, P(K) is the probability of not changing lanes predicted by the model, and p K|w is the probability of not changing lanes calculated based on parameter w, P(R) is the probability of changing lanes right predicted by the model, and p R|w is the probability of right lane change calculated based on parameter w.

[0155] Furthermore, in some embodiments, based on preset parameter and intention update algorithm and preset update trigger mechanism, real-time perception data is used to update the probabilistic vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle, and also includes: if the joint sampling of the driver model parameters and the intention parameters meets the preset time interval condition, then the probabilistic vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle are updated based on the time update trigger mechanism; if the target vehicle meets the preset lane change condition or the prediction result of the probabilistic vehicle interaction prediction model meets the preset result, then the probabilistic vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle are updated based on the event update trigger mechanism.

[0156] Specifically, the update trigger mechanism is a time-time joint trigger mechanism, the update mechanism for the joint sampling of the driver model parameters and intentions is a time trigger mechanism that meets certain time interval conditions, and the parameter update mechanism for the probabilistic lane changing model is an event trigger mechanism. The update is performed when the vehicle changes lanes or the model predicts that the vehicle will change lanes but does not actually change lanes.

[0157] Therefore, in the information-physical twin probabilistic weekly vehicle interactive prediction model identification method of the present application, the model establishment method adopts rule-based horizontal and vertical model building, and assigns probabilistic attributes to it to consider the uncertainty of model parameters and driver intentions. The uncertainty accumulation mechanism considers model parameters, lane change of the preceding vehicle, and intention uncertainty. The model is used for long-term interactive deduction; in the identification method, the model establishment phase initializes the model parameters by fitting the average value of the data set, and the model is updated by the parameter and intention update algorithm during the model operation phase. In the system, the cloud establishes a real-time twin and historical database of traffic through collaborative perception with roadside sensing equipment and intelligent connected vehicles, and establishes a prediction model and updates parameters for vehicles in a specific area according to user applications. The method and system proposed in the present invention provide users with basic services for long-term interactive deduction in the cloud.

[0158] According to the information-physical twin probabilistic surrounding vehicle interactive prediction model identification method of the embodiment of the present application, in the establishment stage of the probabilistic surrounding vehicle interactive prediction model, the parameters of the probabilistic surrounding vehicle interactive prediction model are fitted using a preset data set to obtain the initialized probabilistic surrounding vehicle interactive prediction model parameters and the intention parameters of the target vehicle after initialization. In the operation stage of the probabilistic surrounding vehicle interactive prediction model, based on the preset parameter and intention update algorithm and the preset update trigger mechanism, the real-time perception data is used to update the probabilistic surrounding vehicle interactive prediction model parameters, the intention and uncertainty of the target vehicle. Thus, the problem of update time and update method for the long-term interactive prediction model is solved. Compared with the interaction scheme of single-vehicle autonomous driving, the method proposed in this application can predict the uncertain surrounding vehicle interactive behavior in a longer time domain, and can provide a clear basis for interaction, and can be used as one of the basic functions of more cloud-based long-term predictive decision-making tasks.

[0159] Secondly, the information-physical twin probabilistic weekly vehicle interactive prediction model identification device proposed in accordance with the embodiment of the present application is described with reference to the accompanying drawings.

[0160] Fig. 9 It is a block diagram of the information-physical twin probabilistic weekly vehicle interactive prediction model identification device of an embodiment of the present application.

[0161] like Fig. 9 As shown, the information-physical twin probabilistic weekly vehicle interactive prediction model identification device 10 includes: an acquisition module 100, a fitting module 200 and an update module 300.

[0162] Among them, the acquisition module 100 is used to obtain real-time perception data of the target vehicle; the fitting module 200 is used to fit the parameters of the probabilistic vehicle interaction prediction model using a preset data set during the establishment stage of the probabilistic vehicle interaction prediction model, and obtain the initialized probabilistic vehicle interaction prediction model parameters and the initialized intention parameters of the target vehicle, wherein the probabilistic vehicle interaction prediction model includes a driver model, a probabilistic lane change model and an uncertainty accumulation mechanism, and the probabilistic vehicle interaction prediction model parameters include driver model parameters and probabilistic lane change model parameters; the updating module 300 is used to update the probabilistic vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle using real-time perception data based on preset parameters and intention update algorithm and preset update trigger mechanism during the operation stage of the probabilistic vehicle interaction prediction model.

[0163] Furthermore, in some embodiments, the fitting module 200 is used to fit the driver model parameters based on the weighted bi-norm loss function, and to fit the probabilistic lane change model parameters based on the cross entropy loss function, so as to obtain the initialized probabilistic vehicle-around-vehicle interaction prediction model parameters and the initialized target vehicle intention parameters.

[0164] Further, in some embodiments, the updating module 300 is used to: jointly sample the driver model parameters and the intention parameters based on a preset posterior distribution calculation method, and calculate a first posterior probability according to a first sampling result; separately sample the probabilistic lane change model parameters based on a preset posterior distribution calculation method, and calculate a second posterior probability according to a second sampling result; update the probabilistic vehicle-around-vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle according to the first posterior probability and / or the second posterior probability;

[0165] Among them, the calculation formula of the first posterior probability is:

[0166]

[0167] Among them, N is the sampling time, X pred,i (θ, I) is the state predicted based on the driver model parameters and intention (θ, I), X obsv,i is the actual observed state, σ is the prior variance between the predicted state and the observed state;

[0168] The calculation formula for the second posterior probability is:

[0169]

[0170] Where C is the number of samples, P(L) is the probability of left lane change predicted by the model, and p L|w is the probability of changing lanes to the left calculated based on parameter w, P(K) is the probability of not changing lanes predicted by the model, and pK|w is the probability of not changing lanes calculated based on parameter w, P(R) is the probability of changing lanes right predicted by the model, and p R|w is the probability of right lane change calculated based on parameter w.

[0171] Furthermore, in some embodiments, the update module 300 is also used for: if the joint sampling of the driver model parameters and the intention parameters meets the preset time interval condition, then the probabilistic vehicle-to-vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle are updated based on the time update trigger mechanism; if the target vehicle meets the preset lane change condition or the prediction result of the probabilistic vehicle-to-vehicle interaction prediction model meets the preset result, then the probabilistic vehicle-to-vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle are updated based on the event update trigger mechanism.

[0172] Furthermore, in some embodiments, before acquiring the real-time perception data of the target vehicle, the acquisition module 100 is further used to: establish a representation of the longitudinal motion, lateral motion, and uncertainty accumulation for each vehicle to be interactively predicted.

[0173] It should be noted that the aforementioned explanation of the embodiment of the information-physical twin probabilistic weekly vehicle interactive prediction model identification method is also applicable to the information-physical twin probabilistic weekly vehicle interactive prediction model identification device of this embodiment, and will not be repeated here.

[0174] According to the information-physical twin probabilistic surrounding vehicle interaction prediction model identification device of the embodiment of the present application, in the establishment stage of the probabilistic surrounding vehicle interaction prediction model, the parameters of the probabilistic surrounding vehicle interaction prediction model are fitted using a preset data set to obtain the initialized probabilistic surrounding vehicle interaction prediction model parameters and the initialized target vehicle intention parameters. In the operation stage of the probabilistic surrounding vehicle interaction prediction model, based on the preset parameter and intention update algorithm and the preset update trigger mechanism, the real-time perception data is used to update the probabilistic surrounding vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle. Thus, the problem that the existing technology is difficult to effectively model and update the complex uncertainty in the long-term interaction prediction in real time is solved, and the surrounding vehicle interaction behavior prediction can be performed in a longer time domain, while providing a clear basis for interaction.

[0175] Fig.10 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device may include:

[0176] A memory 1001 , a processor 1002 , and a computer program stored in the memory 1001 and executable on the processor 1002 .

[0177] When the processor 1002 executes the program, it implements the information-physics twin probabilistic week-car interactive prediction model identification method provided in the above-mentioned embodiment.

[0178] Furthermore, the electronic device further comprises:

[0179] The communication interface 1003 is used for communication between the memory 1001 and the processor 1002 .

[0180] The memory 1001 is used to store computer programs that can be executed on the processor 1002 .

[0181] The memory 1001 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0182] If the memory 1001, the processor 1002 and the communication interface 1003 are implemented independently, the communication interface 1003, the memory 1001 and the processor 1002 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig.10 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0183] Optionally, in a specific implementation, if the memory 1001, the processor 1002 and the communication interface 1003 are integrated on a chip, the memory 1001, the processor 1002 and the communication interface 1003 can communicate with each other through an internal interface.

[0184] The processor 1002 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0185] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned information-physical twin probabilistic weekly vehicle interactive prediction model identification method.

[0186] An embodiment of the present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements the above-mentioned information-physical twin probabilistic weekly vehicle interactive prediction model identification method.

[0187] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0188] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0189] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A cyber-physical twin probabilistic week-car interactive prediction model identification method, characterized in that: The following steps are involved: Obtain real-time perception data of the target vehicle; In the establishment stage of the probabilistic vehicle-to-vehicle interaction prediction model, the parameters of the probabilistic vehicle-to-vehicle interaction prediction model are fitted using a preset data set to obtain the initialized probabilistic vehicle-to-vehicle interaction prediction model parameters and the initialized target vehicle intention parameters, wherein the probabilistic vehicle-to-vehicle interaction prediction model includes a driver model, a probabilistic lane-changing model and an uncertainty accumulation mechanism, and the probabilistic vehicle-to-vehicle interaction prediction model parameters include driver model parameters and probabilistic lane-changing model parameters; During the operation phase of the probabilistic vehicle-to-vehicle interaction prediction model, the parameters of the probabilistic vehicle-to-vehicle interaction prediction model, the intention and uncertainty of the target vehicle are updated using the real-time perception data based on a preset parameter and intention update algorithm and a preset update trigger mechanism.

2. The method according to claim 1, characterized in that The method of fitting the parameters of the probabilistic vehicle-to-vehicle interaction prediction model using the preset data set to obtain the initialized probabilistic vehicle-to-vehicle interaction prediction model parameters and the initialized intention parameters of the target vehicle includes: Based on the weighted bi-norm loss function, the driver model parameters are fitted, and based on the cross entropy loss function, the probabilistic lane-changing model parameters are fitted to obtain the initialized probabilistic vehicle-around-vehicle interaction prediction model parameters and the initialized target vehicle intention parameters.

3. The method according to claim 1, characterized in that The method of updating the parameters of the probabilistic vehicle-to-vehicle interaction prediction model, the intention and uncertainty of the target vehicle based on the preset parameter and intention updating algorithm and the preset update triggering mechanism using the real-time perception data includes: Based on a preset posterior distribution calculation method, jointly sampling the driver model parameter and the intention parameter, and calculating a first posterior probability according to a first sampling result; Based on a preset posterior distribution calculation method, the probabilistic lane change model parameters are separately sampled, and a second posterior probability is calculated according to a second sampling result; updating the probabilistic vehicle-to-vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle according to the first posterior probability and / or the second posterior probability; The calculation formula of the first posterior probability is: Among them, N is the sampling time, X pred,i (θ, I) is the state predicted based on the driver model parameters and intention (θ, I), X obsv,i is the actual observed state, σ is the prior variance between the predicted state and the observed state; The calculation formula of the second posterior probability is: Where C is the number of samples, P(L) is the probability of left lane change predicted by the model, and p L|w is the probability of changing lanes to the left calculated based on parameter w, P(K) is the probability of not changing lanes predicted by the model, and p K|w is the probability of not changing lanes calculated based on parameter w, P(R) is the probability of changing lanes right predicted by the model, and p R|w is the probability of right lane change calculated based on parameter w.

4. The method according to claim 3, characterized in that The method of updating the parameters of the probabilistic vehicle-to-vehicle interaction prediction model, the intention and uncertainty of the target vehicle based on the preset parameter and intention updating algorithm and the preset update triggering mechanism using the real-time perception data also includes: If the joint sampling of the driver model parameters and the intention parameters meets the preset time interval condition, the probabilistic vehicle-to-vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle are updated based on the time update trigger mechanism; If the target vehicle meets the preset lane changing condition or the prediction result of the probabilistic surrounding vehicle interaction prediction model meets the preset result, the probabilistic surrounding vehicle interaction prediction model parameters, the intention and uncertainty of the target vehicle are updated based on the event update trigger mechanism.

5. The method according to claim 1, characterized in that Before obtaining the real-time perception data of the target vehicle, it also includes: A representation of longitudinal motion, lateral motion, and uncertainty accumulation is established for each vehicle to be interactively predicted.

6. A cyber-physical twin probabilistic cycle-car interactive prediction model recognition device, characterized in that: include: An acquisition module, used to acquire real-time perception data of the target vehicle; A fitting module is used to fit the parameters of the probabilistic vehicle interaction prediction model using a preset data set during the establishment phase of the probabilistic vehicle interaction prediction model to obtain the initialized probabilistic vehicle interaction prediction model parameters and the initialized target vehicle intention parameters, wherein the probabilistic vehicle interaction prediction model includes a driver model, a probabilistic lane change model and an uncertainty accumulation mechanism, and the probabilistic vehicle interaction prediction model parameters include driver model parameters and probabilistic lane change model parameters; An updating module is used to update the parameters of the probabilistic vehicle interaction prediction model, the intention and uncertainty of the target vehicle using the real-time perception data based on a preset parameter and intention update algorithm and a preset update trigger mechanism during the operation phase of the probabilistic vehicle interaction prediction model.

7. The device according to claim 6, characterized in that The fitting module is used for: Based on the weighted bi-norm loss function, the driver model parameters are fitted, and based on the cross entropy loss function, the probabilistic lane-changing model parameters are fitted to obtain the initialized probabilistic vehicle-around-vehicle interaction prediction model parameters and the initialized target vehicle intention parameters.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the information-physical twin probabilistic weekly vehicle interactive prediction model identification method as described in any one of claims 1 to 5.

9. A computer storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the information-physical twin probabilistic weekly vehicle interactive prediction model identification method as described in any one of claims 1 to 5.

10. A computer program product, characterized in that It includes a computer program, which, when executed by a processor, is used to implement the information-physical twin probabilistic weekly vehicle interactive prediction model identification method described in any one of claims 1 to 5.

Images