Method and device for training driving intention recognition model and method and device for recognizing driving intention

Abstract

The embodiment of the invention discloses a method and device for training a driving intention recognition model and recognizing a driving intention. The specific implementation mode of the method comprises the steps that a training sample is obtained, and the training sample comprises the driving intention of a vehicle, the driving intentions of other vehicles, the earnings of the vehicle and the earnings of the other vehicles; taking the driving intentions of the other vehicles and the benefits of the other vehicles as the input of a first game model, taking the driving intention of the own vehicle as the expected output of the first game model, and training the first game model; training a second game model by taking the driving intention of the own vehicle and the income of the own vehicle as the input of the second game model and taking the driving intentions of other vehicles as the expected output of the second game model; and taking the output of the first game model as the input of the second game model, serially generating driving intention recognition models of other vehicles, taking the output of the second game model as the input of the first game model, and serially generating a driving intention recognition model of the own vehicle. According to the embodiment, the driving intention of the vehicle at the future moment can be effectively predicted.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the field of computer technologies, and particularly to a method and apparatus for training a driving intention recognition model and recognizing driving intentions. Background Art

[0002] In order to improve the adaptive ability of autonomous vehicles to dynamic and complex environments, it is usually necessary to predict the future driving intentions of other vehicles based on the driving state information of other vehicles, so that the autonomous vehicle can make reasonable response actions according to the driving intention, so as to avoid collisions between the autonomous vehicle and other vehicles in the conflict area.

[0003] Currently, the decision-making method of autonomous vehicles in related technologies generally solves the equilibrium by designing a payoff matrix, and based on this, recognizes the driving intentions of other vehicles. This method is widely used in solving and analyzing single-frame small-scale problems, but it cannot solve the large-scale time-series problem of recognizing the time-series intentions of multiple interacting vehicles. Summary of the Invention

[0004] Embodiments of the present disclosure provide a method and apparatus for training a driving intention recognition model and recognizing driving intentions.

[0005] In a first aspect, an embodiment of the present disclosure provides a method for training a driving intention recognition model, including: obtaining training samples, where the training samples include: the driving intention of the host vehicle, the driving intention of other vehicles, the benefit of the host vehicle, and the benefit of other vehicles; using the driving intention of other vehicles and the benefit of other vehicles as inputs of a first game model, and using the driving intention of the host vehicle as the expected output of the first game model to train the first game model; using the driving intention of the host vehicle and the benefit of the host vehicle as inputs of a second game model, and using the driving intention of other vehicles as the expected output of the second game model to train the second game model; using the output of the first game model as the input of the second game model, and concatenating to generate a driving intention recognition model for other vehicles, and using the output of the second game model as the input of the first game model, and concatenating to generate a driving intention recognition model for the host vehicle.

[0006] In some embodiments, the obtaining of the training samples includes: obtaining the historical planned path of the host vehicle and the historical driving state information of other vehicles; calculating the conflict area between the host vehicle and other vehicles based on the historical planned path and the historical driving state information; calculating the benefit of the host vehicle and the benefit of other vehicles based on the conflict area; calculating the driving intention of the host vehicle according to the historical cutting-in frequency of the host vehicle; calculating the driving intention of other vehicles according to the historical cutting-in frequency of other vehicles.

[0007] In some embodiments, the benefits include risk perception benefits and time delay benefits, where the risk perception benefits are associated with the time to collision, and the time delay benefits are associated with the time required to pass through the conflict area.

[0008] In some embodiments, both the first game model and the second game model include a physics-informed neural network, and the loss functions used in the training processes of the first game model and the second game model both include a physical equation loss function and a data matching loss function.

[0009] In some embodiments, the physical equation loss function is generated in the following manner: a dynamic equation is established based on the driving intention of the host vehicle, the driving intention of other vehicles, the benefits of the host vehicle, and the benefits of other vehicles; a replicator dynamic equation set is generated based on the dynamic equation, and an analytical solution of the replicator dynamic equation set is solved; a partial differential equation is obtained based on the analytical solution; and the physical equation loss function is generated based on the partial differential equation.

[0010] In a second aspect, embodiments of the present disclosure provide a method for identifying driving intentions, including: obtaining the planned path of the host vehicle and the driving state information of other vehicles; calculating the conflict area between the host vehicle and other vehicles based on the planned path and the driving state information; calculating the benefits of the host vehicle and other vehicles based on the conflict area; and inputting the benefits of the host vehicle and other vehicles into the driving intention recognition model of the host vehicle and the driving intention recognition model of other vehicles trained according to any one of the first aspect, to obtain the driving intention of the host vehicle and the driving intention of other vehicles.

[0011] In some embodiments, the step of inputting the benefits of the host vehicle and other vehicles into the driving intention recognition model of the host vehicle and the driving intention recognition model of other vehicles to obtain the driving intention of the host vehicle and the driving intention of other vehicles includes: repeating the following steps until the sampling ends: sampling the probability values of the driving intention at a predetermined interval, inputting the sampled probability values, the benefits of the host vehicle, and the benefits of other vehicles into the driving intention recognition model of the host vehicle and the driving intention recognition model of other vehicles respectively, to obtain a first predicted value and a second predicted value; calculating a first distance between the sampled probability value and the first predicted value and a second distance between the sampled probability value and the second predicted value, to obtain the first distance and the second distance corresponding to different probability values; and determining the driving intention of the host vehicle based on the probability value corresponding to the minimum first distance, and determining the driving intention of other vehicles based on the probability value corresponding to the minimum second distance.

[0012] In a third aspect, an embodiment of the present disclosure provides an apparatus for training a driving intention recognition model, including: a sample acquisition unit configured to acquire training samples, where the training samples include: the driving intention of the host vehicle, the driving intention of other vehicles, the benefits of the host vehicle, and the benefits of other vehicles; a first training unit configured to use the driving intention of other vehicles and the benefits of other vehicles as inputs to a first game model, and use the driving intention of the host vehicle as the expected output of the first game model to train the first game model; a second training unit configured to use the driving intention of the host vehicle and the benefits of the host vehicle as inputs to a second game model, and use the driving intention of other vehicles as the expected output of the second game model to train the second game model; a combination unit configured to use the output of the first game model as the input of the second game model to serially generate a driving intention recognition model for other vehicles, and use the output of the second game model as the input of the first game model to serially generate a driving intention recognition model for the host vehicle.

[0013] In some embodiments, the acquisition unit is further configured to: acquire the historical planned path of the host vehicle and the historical driving state information of other vehicles; calculate the conflict area between the host vehicle and other vehicles based on the historical planned path and the historical driving state information; calculate the benefits of the host vehicle and other vehicles based on the conflict area; calculate the driving intention of the host vehicle according to the historical cutting-in frequency of the host vehicle; calculate the driving intention of other vehicles according to the historical cutting-in frequency of other vehicles.

[0014] In some embodiments, the benefits include a risk perception benefit and a time delay benefit, where the risk perception benefit is associated with the time to collision, and the time delay benefit is associated with the time required to pass through the conflict area.

[0015] In some embodiments, both the first game model and the second game model include a physical information neural network, and the loss functions used in the training processes of the first game model and the second game model both include a physical equation loss function and a data matching loss function.

[0016] In some embodiments, the physical equation loss function is generated in the following manner: establish a dynamic equation based on the driving intention of the host vehicle, the driving intention of other vehicles, the benefits of the host vehicle, and the benefits of other vehicles; generate a replicator dynamic equation set based on the dynamic equation and solve the analytical solution of the replicator dynamic equation set; obtain a partial differential equation based on the analytical solution; generate a physical equation loss function based on the partial differential equation.

[0017] Fourth aspect, embodiments of the present disclosure provide a device for identifying driving intentions, including: an acquisition unit configured to acquire the planned path of the host vehicle and the driving state information of other vehicles; a conflict unit configured to calculate a conflict area between the host vehicle and other vehicles based on the planned path and the driving state information; a benefit unit configured to calculate the benefit of the host vehicle and the benefit of other vehicles based on the conflict area; and an identification unit configured to input the benefit of the host vehicle and the benefit of other vehicles into the driving intention recognition model of the host vehicle and the driving intention recognition model of other vehicles to obtain the driving intention of the host vehicle and the driving intention of other vehicles.

[0018] In some embodiments, the identification unit is further configured to: repeatedly execute the following steps until the sampling ends: sample the probability value of the driving intention at a predetermined interval, input the sampled probability value, the benefit of the host vehicle, and the benefit of other vehicles into the driving intention recognition model of the host vehicle and the driving intention recognition model of other vehicles respectively to obtain a first prediction value and a second prediction value; calculate a first distance between the sampled probability value and the first prediction value and a second distance between the sampled probability value and the second prediction value to obtain the first distance and the second distance corresponding to different probability values; determine the driving intention of the host vehicle based on the probability value corresponding to the minimum first distance, and determine the driving intention of other vehicles based on the probability value corresponding to the minimum second distance.

[0019] Fifth aspect, embodiments of the present disclosure provide an electronic device, including: one or more processors; a storage device storing one or more computer programs thereon, when the one or more computer programs are executed by the one or more processors, enabling the one or more processors to implement the method according to any one of the first aspect or the second aspect.

[0020] Sixth aspect, embodiments of the present disclosure provide a computer-readable medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the method according to any one of the first aspect or the second aspect.

[0021] Seventh aspect, embodiments of the present disclosure provide an autonomous vehicle including the electronic device according to the fifth aspect.

[0022] The method and device for training a driving intention recognition model and identifying driving intentions provided by the embodiments of the present disclosure truly and efficiently retain the physical characteristics of the game model through offline training by introducing a physical neural network to solve the game model for driving intention recognition. In addition, the concept of time is introduced into the network model, and the driving intention at a future moment can be effectively learned / solved under complete data. So that the host vehicle has the ability to anticipate and handle the cutting-in behavior of other vehicles, and can then timely respond to the cutting-in behavior of other vehicles to avoid collisions between the host vehicle and other vehicles in the conflict area.

[0023] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it used to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Other features, objects, and advantages of the present disclosure will become more apparent by reading the detailed description of the non-limiting embodiments with reference to the following drawings: Figure 1 is an exemplary system architecture diagram to which an embodiment of the present disclosure can be applied; Figure 2 is a flowchart of an embodiment of a method for training a driving intention recognition model according to the present disclosure; Figure 3 is a schematic diagram of a conflict area according to an embodiment of the present disclosure; Figures 4a - 4d is a schematic diagram of the network structure of a driving intention recognition model according to the present disclosure; Figure 5 is a flowchart of an embodiment of a method for recognizing a driving intention according to the present disclosure; Figure 6 is a schematic structural diagram of an embodiment of an apparatus for training a driving intention recognition model according to the present disclosure; Figure 7 is a schematic structural diagram of an embodiment of an apparatus for recognizing a driving intention according to the present disclosure; Figure 8 is a schematic structural diagram of a computer system of an electronic device suitable for implementing the embodiments of the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] The present disclosure will be further described in detail below with reference to the drawings and embodiments. It can be understood that the specific embodiments described herein are only for explaining the relevant invention and not for limiting the invention. Additionally, it should be noted that for the sake of description, only parts related to the relevant invention are shown in the drawings.

[0026] It should be noted that, without conflict, the embodiments in the present disclosure and the features in the embodiments can be combined with each other. The present disclosure will be described in detail below with reference to the drawings and embodiments.

[0027] Figure 1 An exemplary system architecture 100 is shown to which embodiments of the method or apparatus for training a driving intention recognition model and recognizing a driving intention according to the present disclosure can be applied.

[0028] As Figure 1As shown, the system architecture 100 may include a first vehicle terminal 101, a second vehicle terminal 102, a network 103, and a server 104. The network 103 is used to provide a medium for communication links between the first vehicle terminal 101, the second vehicle terminal 102, and the server 104. The network 103 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0029] Users can use the first vehicle terminal 101 and the second vehicle terminal 102 to interact with the server 104 through the network 103 to receive or send messages, etc. Various communication client applications may be installed on the first vehicle terminal 101 and the second vehicle terminal 102, such as video playback applications, navigation applications, search applications, instant messaging tools, email clients, etc. The first vehicle terminal 101 and the second vehicle terminal 102 are also referred to as agents, in-vehicle computers, in-vehicle intelligent devices, intelligent vehicle terminals, vehicle dispatching and monitoring terminals, in-vehicle wireless terminals, etc.

[0030] The server 104 may provide various services. For example, the server 104 may train a driving intention recognition model for the host vehicle and a driving intention recognition model for other vehicles based on the historical planned path of the host vehicle and the historical driving state information of other vehicles. Then, use the driving intention recognition model of the host vehicle and the driving intention recognition model of other vehicles to recognize the driving intention of the host vehicle and the driving intention of other vehicles at the current moment. So that the autonomous vehicle can make reasonable response actions according to the driving intention.

[0031] It should be noted that the server 104 may be hardware or software. When the server 104 is hardware, it can be implemented as a distributed server cluster composed of multiple servers, or as a single server. When the server 104 is software, it can be implemented as multiple software or software modules (such as those used to provide distributed services), or as a single software or software module. No specific limitation is made here.

[0032] It should be noted that the vehicle control method provided by the embodiments of the present application is generally executed by the server 104. Correspondingly, the vehicle control device is generally arranged in the server 104.

[0033] It should be understood that Figure 1 the numbers of the first vehicle terminal, the second vehicle terminal, the network, and the server in

[0034] are merely illustrative. According to the implementation requirements, there may be any number of first vehicle terminals, second vehicle terminals, networks, and servers. Figure 2 Continuing to refer to Step 201, obtain training samples.

[0035] In this embodiment, the execution subject of the method for training a driving intention recognition model (such as Figure 1 the server shown) can obtain training samples from a third-party server through a wired connection or a wireless connection. Among them, the training samples include: the driving intention of the host vehicle, the driving intention of other vehicles, the benefits of the host vehicle, and the benefits of other vehicles.

[0036] The driving intention of the host vehicle may include the probability of the host vehicle cutting in line or the probability of the host vehicle yielding. The driving intention of other vehicles may include the probability of other vehicles cutting in line or the probability of other vehicles yielding.

[0037] Here, other vehicles refer to obstacle vehicles whose driving paths conflict with the driving path of the host vehicle. Both the host vehicle and other vehicles can be autonomous vehicles.

[0038] The benefits of the host vehicle refer to the time delay benefits obtained by the host vehicle being able to quickly pass through the conflict area due to the cutting-in-line or yielding behavior of the host vehicle and the cutting-in-line or yielding behavior of other vehicles. There will also be risk perception benefits obtained by avoiding vehicle collisions by predicting the cutting-in-line behavior of both parties in advance.

[0039] The benefits of the host vehicle and other vehicles can be calculated based on the vehicle driving speed and the conflict area.

[0040] In addition, the number of other vehicles can be multiple.

[0041] Step 202, use the driving intention of other vehicles and the benefits of other vehicles as the input of the first game model, and use the driving intention of the host vehicle as the expected output of the first game model to train the first game model.

[0042] In this embodiment, the first game model can adopt a general MLP (Multi-Layer Perceptron) model, and the network structure is as Figure 4a shown. The model input includes the benefits of other vehicles ( representing the risk perception benefits of other vehicles, representing the time delay benefits of other vehicles), the driving intention y of other vehicles, and the time frame t, and outputs the predicted driving intention x of the host vehicle. Using supervised training, taking the driving intention of the host vehicle as the expected output of the first game model, continuously adjusting the network parameters of the first game model so that the difference between the predicted x and the driving intention of the host vehicle in the training samples becomes smaller and smaller, and converges to a predetermined value, then the training is completed, and the trained first game model is obtained.

[0043] Step 203: Use the driving intention of the host vehicle and the benefits of the host vehicle as the input of the second game model, and use the driving intentions of other vehicles as the expected output of the second game model to train the second game model.

[0044] In this embodiment, the second game model can adopt a general MLP (Multi-Layer Perceptron) model, and the network structure is as Figure 4b shown. The model input includes the benefits of the host vehicle ( indicating the risk perception benefit of the host vehicle, indicating the time delay benefit of the host vehicle), the driving intention x of the host vehicle, and the time frame t, and the output is to predict the driving intention y of other vehicles. Using supervised training, with the driving intentions of other vehicles as the expected output of the second game model, continuously adjust the network parameters of the second game model so that the difference between the predicted y and the driving intentions of other vehicles in the training samples becomes smaller and smaller, converging to a predetermined value, then the training is completed, and the trained second game model is obtained.

[0045] Step 204: Use the output of the first game model as the input of the second game model to serially generate a driving intention recognition model for other vehicles, and use the output of the second game model as the input of the first game model to serially generate a driving intention recognition model for the host vehicle.

[0046] In this embodiment, in the order as Figure 4d shown, use the output of the first game model as the input of the second game model to serially generate a driving intention recognition model for other vehicles. In the order as Figure 4c shown, use the output of the second game model as the input of the first game model to serially generate a driving intention recognition model for the host vehicle.

[0047] The method provided by the above embodiment of the present disclosure introduces a neural network to solve the game model, introduces the time concept into the network model, and can effectively learn / solve the driving intentions at future moments under complete data.

[0048] In some alternative implementation manners of this embodiment, obtaining training samples includes: obtaining the historical planned path of the host vehicle and the historical driving state information of other vehicles; based on the historical planned path and the historical driving state information, calculating the conflict area between the host vehicle and other vehicles; calculating the benefits of the host vehicle and other vehicles based on the conflict area; calculating the driving intention of the host vehicle according to the historical cutting-in frequency of the host vehicle; calculating the driving intention of other vehicles according to the historical cutting-in frequency of other vehicles.

[0049] Multiple historical moments can be obtained from the driving records of each vehicle to acquire the planned path of the ego vehicle (which may include trajectory points and vehicle width) and the driving state information of other vehicles (which may include position, speed, and acceleration) when making longitudinal actions (such as whether to cut in) for the ego vehicle. Based on the driving state information of other vehicles, the planned paths of other vehicles can be obtained. Then, based on the planned path of the ego vehicle and the planned paths of other vehicles, the path overlapping regions between the ego vehicle and other vehicles can be obtained. Based on the path overlapping regions between the ego vehicle and other vehicles, the conflict regions can be calculated.

[0050] The conflict regions are determined by the shapes and paths of the ego vehicle and other vehicles. As Figure 3 shown, the boundaries of the conflict regions are the projected positions on their respective paths of the overlapping regions between the path of the ego vehicle and the predicted trajectories of other vehicles. Among them, ABCD is the interaction conflict region between the ego vehicle and other vehicles. A is the front boundary of the ego vehicle reaching the conflict region, B is the rear boundary of the ego vehicle leaving the conflict region, C is the front boundary of other vehicles reaching the conflict region, and D is the rear boundary of other vehicles leaving the conflict region.

[0051] Both other vehicles and the ego vehicle can pass through the conflict region prior to the other, or choose to wait and let the other pass through the conflict region first.

[0052] The benefit is decoupled into the sum of the following two benefits: (1) Risk perception benefit, which is used to characterize the unpleasant experience generated when two vehicles collide; (2) Time delay benefit, which is used to characterize the time loss generated when a vehicle chooses to yield, and is equal to the time required for the other vehicle to pass through the conflict region in the current state.

[0053] The benefits of the ego vehicle and other vehicles should satisfy the following principles: (1) If both the ego vehicle and other vehicles choose to cut in, both the ego vehicle and other vehicles will lose the risk perception benefit and part of the time delay benefit; (2) When other vehicles choose to cut in and the ego vehicle chooses to yield, other vehicles will obtain the time delay benefit and the risk perception benefit, and the ego vehicle will obtain the risk perception benefit. The ego vehicle loses the time delay benefit due to choosing to wait; (3) When other vehicles choose to yield and the ego vehicle chooses to cut in, the ego vehicle will obtain the time delay benefit and the risk perception benefit, and other vehicles will obtain the risk perception benefit. Other vehicles lose the time delay benefit due to choosing to wait; (4) When both the ego vehicle and other vehicles choose to yield, they will both obtain the risk perception benefit, but they will both lose the time delay benefit due to choosing to wait.

[0054] Specifically, the benefits of the ego vehicle and other vehicles can be represented in the form of a payoff matrix as shown in the following table. It should be noted that the specific form of the payoff matrix can be adjusted according to the requirements of the business scenario.

[0055]

[0056] Among them, represents the risk perception benefit of the host vehicle, represents the risk perception benefit of other vehicles. is the time required for the host vehicle to pass through the conflict area normally, representing the time delay benefit of the host vehicle. is the time required for other vehicles to pass through the conflict area normally, representing the time delay benefit of other vehicles. and are parameters of the driving intention recognition model, which vary with the cumulative waiting time of other vehicles; and are parameters of the driving intention recognition model, used to characterize the coefficient of the time delay benefit when the host vehicle and other vehicles both choose to cut in.

[0057] Optionally, , , , , is used to characterize the speed of the host vehicle, is used to characterize the speed of other vehicles, is used to characterize the distance between the host vehicle and the front boundary of the conflict area, is used to characterize the distance between other vehicles and the front boundary of the conflict area, is used to characterize the collision time of the host vehicle, is used to characterize the collision time of other vehicles, is used to characterize the distance between the front boundary and the rear boundary of the conflict area corresponding to the host vehicle, is used to characterize the distance between the front boundary and the rear boundary of the conflict area corresponding to other vehicles.

[0058] Here, in the case where both the host vehicle and other vehicles cut in, can represent the cut-in benefit of the host vehicle, can represent the cut-in benefit of other vehicles. In the case where the host vehicle cuts in and other vehicles yield, can represent the cut-in benefit of the host vehicle, can represent the yield benefit of other vehicles. In the case where the host vehicle yields and other vehicles cut in, can represent the yield benefit of the host vehicle, can represent the cut-in benefit of other vehicles. In the case where both the host vehicle and other vehicles yield, can represent the yield benefit of the host vehicle, can represent the yield benefit of other vehicles.

[0059] Calculate the driving intention of the host vehicle based on the historical cutting-in frequency of the host vehicle. By statistically counting the behavior results of the host vehicle passing through intersections multiple times to obtain the cutting-in frequency, the average cutting-in probability of the host vehicle is obtained as the driving intention of the host vehicle. For example, if the host vehicle chooses to cut in 6 times and yield 4 times out of 10 times passing through an intersection, the average cutting-in probability of the host vehicle is 60%.

[0060] Calculate the driving intention of other vehicles based on the historical cutting-in frequency of other vehicles. By statistically counting the behavior results of multiple other vehicles to obtain the cutting-in frequency, the average cutting-in probability of other vehicles is obtained as the driving intention of other vehicles. For example, if 6 out of 10 other vehicles choose to cut in and 4 choose to yield, the average cutting-in probability of other vehicles is 60%.

[0061] Generate different training samples based on the planned paths of the host vehicle and the driving state information of other vehicles at different historical moments. The driving intention recognition model is supervised trained with a training sample set composed of a large number of training samples.

[0062] In some optional implementation manners of this embodiment, the benefits include risk perception benefits and time delay benefits. Among them, the risk perception benefits are associated with the time to collision, and the time delay benefits are associated with the time required to pass through the conflict area. The risk perception benefits are inversely proportional to the time to collision. The time delay benefits are proportional to the time required to pass through the conflict area.

[0063] In some optional implementation manners of this embodiment, both the first game model and the second game model include a physics-informed neural network, and the loss functions used in the training processes of the first game model and the second game model both include a physical equation loss function and a data matching loss function.

[0064] A physics-informed neural network (PINN for short) is a machine learning model that combines deep learning and physics knowledge. The PINN model is usually composed of a deep neural network, and its characteristic is that a physical information term, that is, the physical law followed, is added to the loss function. When training the model, not only the data error needs to be minimized, but also the physical information error needs to be minimized to ensure that the prediction results conform to the physical laws.

[0065] The physical equation loss function is unique to PINN, and it considers whether the network prediction results satisfy the physical laws. The residuals calculated by substituting the physical quantities predicted by the network into the corresponding physical laws (usually differential equations) constitute this part of the loss function, thus ensuring physical consistency. Partial differential equations can be created according to the benefits of the host vehicle and the benefits of other vehicles. In some alternative implementation manners of this embodiment, the physical equation loss function is generated in the following manner: establish a dynamic equation based on the driving intention of the host vehicle, the driving intention of other vehicles, the benefit of the host vehicle, and the benefit of other vehicles; generate a replicator dynamic equation set based on the dynamic equation, and solve the analytical solution of the replicator dynamic equation set; obtain a partial differential equation based on the analytical solution; generate a physical equation loss function based on the partial differential equation.

[0066] According to the evolutionary game theory, establish a dynamic equation based on the driving intention of the host vehicle, the driving intention of other vehicles, the benefit of the host vehicle, and the benefit of other vehicles, as shown in formulas (1)-(6): Formula (1) Formula (2) Formula (3) Formula (4) Formula (5) Formula (6) Among them, x can represent the driving intention of the host vehicle, y can represent the driving intention of other vehicles, represents the expected benefit of the host vehicle cutting in, represents the expected benefit of the host vehicle yielding, and represents the average benefit of the host vehicle; represents the expected benefit of other vehicles cutting in, represents the expected benefit of other vehicles yielding, and represents the average benefit of other vehicles.

[0067] Generate a replicator dynamic equation set, as shown in formulas (7)(8): Formula (7) Formula (8) Solve the analytical solution of the replicator dynamic equation set:

[0068] Formula (9)

[0069] Formula (10)

[0070] Among them, represents the frame time, used as a superscript to represent the difference in benefits at different frame times. is a function of is a function of The function, i.e.,

[0071] Generate the partial differential equation 1 according to formula (9) Formula (11) The constraint conditions are: the probability of the ego vehicle cutting in at time t = 0 is between [0, 1], and the probability of the ego vehicle cutting in at any subsequent time is between [0, 1], which is expressed by the following formula:

[0072] The loss function 1 of the first game model includes a physical equation loss function and a data matching loss function.

[0073] Train the first game model to directly approximate the solution of the partial differential equation 1

[0074] However, here we assume that is unknown and regard it as an additional learnable parameter when training the first game model.

[0075] The first game model is trained using the following loss function:

[0076] where represents the data matching loss function The weight coefficient of

[0077] The physical equation loss function :

[0078] where N represents the time t. When the constraint conditions described above are satisfied according to the Lagrange multiplier method, the difference between the predicted value of the output of the first game model and the expected probability of the ego vehicle cutting in is used as the objective function, and the network parameters of the first game model when the objective function is minimized, as well as the coefficient of the payoff in the payoff matrix , , and respectively represent the weight coefficients.

[0079] The data matching loss function :

[0080] Among them, M represents the moment t, represents the expected output of the model, that is, the actual probability of the ego vehicle cutting in, represents the predicted probability of the ego vehicle cutting in.

[0081] The partial differential equation 2 is obtained according to formula (10)

[0082] The constraint conditions are: the probability of other vehicles cutting in at the moment t = 0 is between [0, 1], and the probability of other vehicles cutting in at any subsequent moment is between [0, 1], which is expressed by the following formula:

[0083] The loss function 2 of the second game model includes a physical equation loss function and a data matching loss function.

[0084] Train the second game model to directly approximate the solution of the partial differential equation 2

[0085] Here we also assume that is unknown and is regarded as an additional learnable parameter when training the second game model.

[0086] The second game model is trained using the following loss function:

[0087] Among them, represents the data matching loss function of the weight coefficient.

[0088] The physical equation loss function :

[0089] Among them, N represents the moment t. Under the condition of satisfying the constraint conditions described above according to the Lagrange multiplier method, the predicted value of the output of the second game model and the expected probability of other vehicles cutting in The difference is used as the objective function to solve the network parameters of the first game model when the objective function is minimized, and the coefficient of the payoff in the payoff matrix , , and respectively represent the weight coefficients.

[0090] Data matching loss function :

[0091] Where M represents time t, represents the expected output of the model, that is, the actual probability of other vehicles cutting in, represents the predicted probability of other vehicles cutting in.

[0092] Further referring to Figure 5 , which shows the flow 500 of an embodiment of the method for identifying driving intention. The flow 500 of the method for identifying driving intention includes the following steps: Step 501, obtaining the planned path of the host vehicle and the driving state information of other vehicles.

[0093] In this embodiment, the planned path of the host vehicle at the current moment (which may include trajectory points and vehicle width) and the driving state information of other vehicles (which may include position, speed, and acceleration) are obtained.

[0094] Step 502, calculating the conflict area between the host vehicle and other vehicles based on the planned path and driving state information.

[0095] In this embodiment, based on the driving state information of other vehicles, the planned path of other vehicles is obtained. Then, based on the planned path of the host vehicle and the planned path of other vehicles, the path overlapping area between the host vehicle and other vehicles can be obtained. Based on the path overlapping area between the host vehicle and other vehicles, the conflict area is calculated.

[0096] Step 503, calculating the benefit of the host vehicle and the benefit of other vehicles based on the conflict area.

[0097] In this embodiment, the benefit of the host vehicle and the benefit of other vehicles are calculated according to the method described above.

[0098] Step 504, inputting the benefit of the host vehicle and the benefit of other vehicles into the driving intention recognition model of the host vehicle and the driving intention recognition model of other vehicles to obtain the driving intention of the host vehicle and the driving intention of other vehicles.

[0099] In this embodiment, the driving intention recognition model of the host vehicle and the driving intention recognition model of other vehicles are trained by the method described in flow 200. The Nash equilibrium can be solved by the numerical sampling method.

[0100] In a game, if each participant has chosen a strategy and no participant can obtain a better result by changing their own strategy, then this state is called the Nash equilibrium.

[0101] In this application, the understanding of equilibrium is as follows: when the actions of one party are minimally affected by the other party (or multiple parties), it is an equilibrium solution. That is, we consider the two parties in the game to have a completely cooperative relationship. We use distance to evaluate this equilibrium. Based on the numerical sampling method, we calculate the distance two-norm. The smaller the distance two-norm, the closer it is to the equilibrium solution.

[0102] As Figure 4c shown, find the when the following formula is minimized as the equilibrium solution of the self-vehicle's cutting-in probability:

[0103] As Figure 4d shown, find the when the following formula is minimized as the equilibrium solution of the self-vehicle's cutting-in probability:

[0104] In some alternative implementation manners of this embodiment, repeat the following steps until the sampling ends: sample the probability values of the driving intention at a predetermined interval, input the sampled probability values, the benefits of the self-vehicle, and the benefits of other vehicles into the driving intention recognition model of the self-vehicle and the driving intention recognition model of other vehicles respectively to obtain the first prediction value and the second prediction value; calculate the first distance between the sampled probability value and the first prediction value and the second distance between the sampled probability value and the second prediction value to obtain the first distance and the second distance corresponding to different probability values; Determine the driving intention of the self-vehicle based on the probability value corresponding to the minimum first distance, and determine the driving intention of other vehicles based on the probability value corresponding to the minimum second distance.

[0105] As Figure 4c shown, take the sampled probability as and the benefit of the self-vehicle at time t ( and ) and input them into the second game model to obtain the predicted cutting-in probability y of other vehicles, and then input y and the benefit of other vehicles at time t ( and ) into the first game model to obtain the predicted cutting-in probability of the self-vehicle. Calculate the distance between . Traverse all the sampled probabilities and find the with the minimum distance as the cutting-in probability of the self-vehicle.

[0106] As Figure 4d shown, take the sampled probability as and the benefit of other vehicles at time t ( and ) and input them into the first game model to obtain the predicted cutting-in probability x of the self-vehicle, and then input x and the benefit of the self-vehicle at time t ( and Input the second game model to obtain the predicted probability of other vehicles cutting in . Calculate the distance between. Traverse all the sampling probabilities to find the one with the minimum distance as the probability of other vehicles cutting in.

[0107] For further reference Figure 6 , as an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of an apparatus for training a driving intention recognition model. This apparatus embodiment corresponds to Figure 2 the method embodiment shown, and this apparatus can be specifically applied to various electronic devices.

[0108] As Figure 6 shown, the apparatus 600 for training a driving intention recognition model in this embodiment includes: a sample acquisition unit 601, a first training unit 602, a second training unit 603, and a combination unit 604. Among them, the sample acquisition unit 601 is configured to acquire training samples, where the training samples include: the driving intention of the host vehicle, the driving intention of other vehicles, the benefit of the host vehicle, and the benefit of other vehicles; the first training unit 602 is configured to use the driving intention of other vehicles and the benefit of other vehicles as the input of the first game model, and use the driving intention of the host vehicle as the expected output of the first game model to train the first game model; the second training unit 603 is configured to use the driving intention of the host vehicle and the benefit of the host vehicle as the input of the second game model, and use the driving intention of other vehicles as the expected output of the second game model to train the second game model; the combination unit 604 is configured to use the output of the first game model as the input of the second game model to serially generate a driving intention recognition model for other vehicles, and use the output of the second game model as the input of the first game model to serially generate a driving intention recognition model for the host vehicle.

[0109] In this embodiment, the specific processing of the sample acquisition unit 601, the first training unit 602, the second training unit 603, and the combination unit 604 of the apparatus 600 for training a driving intention recognition model can refer to Figure 2 steps 201, 202, 203, and 204 in the corresponding embodiment.

[0110] In some alternative implementation manners of this embodiment, the sample acquisition unit 601 is further configured to: acquire the historical planned path of the host vehicle and the historical driving state information of other vehicles; calculate the conflict area between the host vehicle and other vehicles based on the historical planned path and the historical driving state information; calculate the benefits of the host vehicle and other vehicles based on the conflict area; calculate the driving intention of the host vehicle according to the historical cutting-in frequency of the host vehicle; calculate the driving intention of other vehicles according to the historical cutting-in frequency of other vehicles.

[0111] In some alternative implementation manners of this embodiment, the benefits include risk perception benefits and time delay benefits, wherein the risk perception benefits are associated with the time to collision, and the time delay benefits are associated with the time required to pass through the conflict area.

[0112] In some alternative implementation manners of this embodiment, both the first game model and the second game model include a physical information neural network, and the loss functions used in the training processes of the first game model and the second game model both include a physical equation loss function and a data matching loss function.

[0113] In some alternative implementation manners of this embodiment, the physical equation loss function is generated in the following manner: establish a dynamic equation based on the driving intention of the host vehicle, the driving intention of other vehicles, the benefits of the host vehicle, and the benefits of other vehicles; generate a replicator dynamic equation set based on the dynamic equation, and solve the analytical solution of the replicator dynamic equation set; obtain a partial differential equation based on the analytical solution; generate a physical equation loss function based on the partial differential equation.

[0114] Further referring to Figure 7 , as an implementation of the methods shown in the above figures, an embodiment of a device for identifying driving intention is provided by the present disclosure. This device embodiment corresponds to Figure 5 the method embodiment shown, and this device can be specifically applied to various electronic devices.

[0115] As Figure 7 shown, the device 700 for identifying driving intention in this embodiment includes: an acquisition unit 701, a conflict unit 702, a benefit unit 703, and an identification unit 704. Among them, the acquisition unit 701 is configured to acquire the planned path of the host vehicle and the driving state information of other vehicles; the conflict unit 702 is configured to calculate the conflict area between the host vehicle and other vehicles based on the planned path and the driving state information; the benefit unit 703 is configured to calculate the benefits of the host vehicle and other vehicles based on the conflict area; the identification unit 704 is configured to input the benefits of the host vehicle and other vehicles into the driving intention recognition models of the host vehicle and other vehicles trained by the device 600 to obtain the driving intention of the host vehicle and the driving intention of other vehicles.

[0116] In this embodiment, for the specific processing of the acquisition unit 701, conflict unit 702, benefit unit 703, and recognition unit 704 of the driving intention recognition device 700, reference can be made to Figure 5 Steps 501, 502, 503, and 504 in the corresponding embodiment.

[0117] In some optional implementation manners of this embodiment, the recognition unit 704 is further configured to: repeatedly execute the following steps until the sampling ends: sample the probability value of the driving intention at a predetermined interval, and input the sampled probability value, the benefit of the host vehicle, and the benefits of other vehicles into the driving intention recognition model of the host vehicle and the driving intention recognition models of other vehicles respectively to obtain a first prediction value and a second prediction value; calculate a first distance between the sampled probability value and the first prediction value and a second distance between the sampled probability value and the second prediction value to obtain the first distance and the second distance corresponding to different probability values; determine the driving intention of the host vehicle based on the probability value corresponding to the minimum first distance, and determine the driving intention of other vehicles based on the probability value corresponding to the minimum second distance.

[0118] It should be noted that in the technical solution of the present disclosure, for the aspects of collection, gathering, updating, analysis, processing, use, transmission, storage, etc. of the user's personal information, all comply with the provisions of relevant laws and regulations, are used for legal purposes, and do not violate public order and good customs. Necessary measures are taken for the user's personal information to prevent illegal access to the user's personal information data, and to safeguard the user's personal information security, network security, and national security.

[0119] According to the embodiments of the present disclosure, the present disclosure also provides an electronic device and a readable storage medium.

[0120] An electronic device includes: one or more processors; a storage device on which one or more computer programs are stored, and when the one or more computer programs are executed by the one or more processors, the one or more processors implement the method described in process 200 or 500.

[0121] A computer-readable medium has a computer program stored thereon, wherein when the computer program is executed by a processor, the method described in process 200 or 500 is implemented.

[0122] Figure 8FIG. shows a schematic block diagram of an exemplary electronic device 800 that can be used to implement embodiments of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, personal digital processors, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely exemplary and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0123] As Figure 8 shown, the device 800 includes a computing unit 801 that can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0124] A plurality of components in the device 800 are connected to the I / O interface 805, including: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a magnetic disk, an optical disk, etc.; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 809 allows the device 800 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0125] The computing unit 801 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 executes the various methods and processes described above, such as the road area planning method. For example, in some embodiments, the road area planning method can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 808. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the road area planning method described above can be executed. Alternatively, in other embodiments, the computing unit 801 can be configured to execute the road area planning method by any other suitable means (e.g., by means of firmware).

[0126] Various embodiments of the systems and techniques described above in this document can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGA), application-specific integrated circuits (ASIC), application-specific standard products (ASSP), systems-on-a-chip (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a dedicated or general-purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0127] The program code for implementing the methods of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the program codes are executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program codes can be executed entirely on the machine, partially on the machine, executed partially on the machine as an independent software package and partially on a remote machine, or executed entirely on a remote machine or server.

[0128] In the context of this disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0129] In order to provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0130] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0131] A computer system may include a client and a server. The client and the server are generally far from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the respective computers and having a client-server relationship with each other. The server may be a server of a distributed system or a server combined with a blockchain. The server may also be a cloud server or an intelligent cloud computing server or an intelligent cloud host with artificial intelligence technology. The server may be a server of a distributed system or a server combined with a blockchain. The server may also be a cloud server or an intelligent cloud computing server or an intelligent cloud host with artificial intelligence technology.

[0132] It should be understood that various forms of processes shown above can be used, steps can be reordered, added or deleted. For example, the steps described in the present disclosure can be executed in parallel, sequentially or in a different order, as long as the desired results of the technical solutions disclosed in the present disclosure can be achieved, and no limitation is imposed herein.

[0133] The above specific embodiments do not constitute a limitation on the protection scope of the present disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principle of the present disclosure shall be included within the protection scope of the present disclosure.

Claims

1. A method for training a driving intention recognition model, comprising: Acquire training samples, wherein the training samples include: the driving intention of the vehicle, the driving intention of other vehicles, the benefit of the vehicle, and the benefit of other vehicles; The driving intention of the other vehicles and the benefits of the other vehicles are used as inputs of the first game model, and the driving intention of the own vehicle is used as the expected output of the first game model to train the first game model; The driving intention of the vehicle and the benefit of the vehicle are used as inputs of a second game model, and the driving intentions of other vehicles are used as expected outputs of the second game model to train the second game model; The output of the first game model is used as the input of the second game model to generate driving intention recognition models of other vehicles in series, and the output of the second game model is used as the input of the first game model to generate the driving intention recognition model of the vehicle itself in series.

2. The method according to claim 1, wherein: The obtaining of training samples comprises: Obtain the historical planned path of the vehicle and the historical driving status information of other vehicles; Calculating a conflict area between the vehicle and other vehicles based on the historical planned path and the historical driving state information; Calculating the benefit of the vehicle and the benefits of other vehicles based on the conflict area; Calculate the driving intention of the vehicle based on the historical frequency of overtaking; The driving intention of other vehicles is calculated based on the historical frequency of overtaking by other vehicles.

3. The method according to claim 2, wherein: The benefit includes a risk perception benefit and a time delay benefit, wherein the risk perception benefit is associated with the collision time, and the time delay benefit is associated with the time required to pass through the conflict area.

4. The method according to claim 1, wherein: The first game model and the second game model both include physical information neural networks, and the loss functions used in the training process of the first game model and the second game model both include physical equation loss functions and data matching loss functions.

5. The method according to claim 4, wherein: The physical equation loss function is generated in the following way: A dynamic equation is established based on the driving intention of the own vehicle, the driving intention of other vehicles, the benefit of the own vehicle and the benefits of other vehicles; Generate a duplicate dynamic equation group based on the dynamic equation, and solve the duplicate dynamic equation group for an analytical solution; Obtaining a partial differential equation based on the analytical solution; A physical equation loss function is generated based on the partial differential equation.

6. A method for identifying driving intention, comprising: Obtain the planned path of the vehicle and the driving status information of other vehicles; Calculating a conflict area between the vehicle and other vehicles based on the planned path and the driving state information; Calculating the benefit of the vehicle and the benefits of other vehicles based on the conflict area; The benefits of the own vehicle and the benefits of other vehicles are input into the driving intention recognition model of the own vehicle and the driving intention recognition model of other vehicles trained according to any one of the methods described in claims 1-5 to obtain the driving intention of the own vehicle and the driving intention of other vehicles.

7. The method according to claim 6, wherein: The step of inputting the benefits of the vehicle and the benefits of other vehicles into the driving intention recognition model of the vehicle and the driving intention recognition model of other vehicles trained according to any one of the methods of claims 1 to 5 to obtain the driving intention of the vehicle and the driving intention of other vehicles comprises: Repeat the following steps until sampling is completed: sampling the probability value of driving intention at a predetermined interval, inputting the sampled probability value, the benefit of the own vehicle and the benefit of other vehicles into the driving intention recognition model of the own vehicle and the driving intention recognition model of other vehicles trained according to any one of the methods of claims 1 to 5, respectively, to obtain a first prediction value and a second prediction value; calculating a first distance between the sampled probability value and the first prediction value and a second distance between the sampled probability value and the second prediction value, to obtain a first distance and a second distance corresponding to different probability values; The driving intention of the own vehicle is determined based on the probability value corresponding to the minimum first distance, and the driving intention of the other vehicles is determined based on the probability value corresponding to the minimum second distance.

8. A device for training a driving intention recognition model, comprising: A sample acquisition unit is configured to acquire training samples, wherein the training samples include: the driving intention of the own vehicle, the driving intention of other vehicles, the benefit of the own vehicle and the benefit of other vehicles; A first training unit is configured to use the driving intention of the other vehicles and the benefits of the other vehicles as inputs of a first game model, and use the driving intention of the self-vehicle as an expected output of the first game model to train the first game model; A second training unit is configured to use the driving intention of the self-vehicle and the benefit of the self-vehicle as inputs of a second game model, and use the driving intentions of other vehicles as expected outputs of the second game model to train the second game model; The combination unit is configured to use the output of the first game model as the input of the second game model to generate driving intention recognition models of other vehicles in series, and use the output of the second game model as the input of the first game model to generate the driving intention recognition model of the vehicle itself in series.

9. A device for identifying driving intention, comprising: An acquisition unit, configured to acquire the planned path of the vehicle and the driving status information of other vehicles; a conflict unit, configured to calculate a conflict area between the vehicle and other vehicles based on the planned path and the driving state information; A benefit unit, configured to calculate a benefit of the vehicle and benefits of other vehicles based on the conflict area; The recognition unit is configured to input the benefits of the own vehicle and the benefits of other vehicles into the driving intention recognition model of the own vehicle and the driving intention recognition model of other vehicles trained according to any method of claims 1-5, and obtain the driving intention of the own vehicle and the driving intention of other vehicles.

10. An electronic device, comprising: one or more processors; a storage device having one or more computer programs stored thereon, When the one or more computer programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

11. A computer readable medium having a computer program stored thereon, wherein: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

12. An autonomous driving vehicle comprising the electronic device as claimed in claim 10.

Images