Method, device, vehicle, program product, and system for determining a vehicle driving strategy

Abstract

The present disclosure relates to a method, an apparatus, a vehicle, a program product, and a system for determining a driving strategy of a vehicle. A method and a system for determining a driving strategy. The method includes the following steps: obtaining vehicle parameters by a driving strategy planning module of the vehicle, determining, by the driving strategy planning module based on the obtained vehicle parameters, at least one feasible driving strategy by means of at least one decision sub-module of the driving strategy planning module, transmitting one of the at least one feasible driving strategies to a motion planning module of the vehicle, obtaining an evaluation quantity from the motion planning module by the driving strategy planning module, and adjusting the at least one decision sub-module according to the obtained evaluation quantity. Also disclosed is a system for determining a driving strategy of a vehicle, a vehicle for implementing the driving strategy, a control device for a vehicle, and a computer program for implementing the method.

Description

Technical Field

[0001] The present invention relates to a method and a system for determining a driving strategy of a vehicle, a vehicle for implementing the driving strategy, a control device for controlling the vehicle for the vehicle, and a computer program for implementing the method. Background Art

[0002] One of the main challenges for the autonomous or semi-autonomous control of a vehicle is to determine the driving strategy of the vehicle. Usually, a feasible driving strategy is determined based on the data collected by the vehicle, such as the number of passable lanes of the road, the movement trajectories of adjacent vehicles, and the navigation data for reaching the desired driving destination.

[0003] In the prior art, it has been disclosed to plan the itinerary of a vehicle through a driving itinerary planning module (also called "mission planning framework"). The system uses the current location and the destination for this purpose. In addition, the system has a driving strategy planning module (also called "behavioral planning framework"), which plans the driving strategy based on sensor data. The driving strategy generated according to the driving itinerary planning module and the driving strategy planning module is transmitted to the motion planning module (also called "motion planning framework"), which implements the driving strategy.

[0004] However, the disadvantage of these systems is that the selection of the driving strategy is only based on theoretical considerations. Summary of the Invention

[0005] In view of this, an object of the present invention is to provide a method for determining a driving strategy, which takes into account the practical feedback in the process of selecting the driving strategy.

[0006] The solution of the present invention to achieve the above object is a method for determining the driving strategy of a vehicle. First, the driving strategy planning module of the vehicle obtains vehicle parameters. Based on the obtained vehicle parameters, the driving strategy planning module determines at least one feasible driving strategy by means of at least one decision sub-module of the driving strategy planning module. Subsequently, the driving strategy is selected and transmitted to the motion planning module of the vehicle. The driving strategy planning module obtains an evaluation quantity from the motion planning module. Finally, the at least one decision sub-module is adjusted according to the obtained evaluation quantity.

[0007] According to the present invention, the motion planning module is connected to the driving strategy planning module, wherein the driving strategy planning module determines at least one feasible driving strategy and transmits at least one driving strategy to the motion planning module. The driving strategy planning module obtains an evaluation quantity for the transmitted driving strategy, which includes practical feedback regarding the transmitted driving strategy. Subsequently, at least one decision sub-module of the driving strategy planning module is adjusted based on the practical feedback, so that in addition to the theoretical basis, the practical basis is also used for selecting the driving strategy.

[0008] The motion planning module can adopt an implementation scheme based on the cloud or a server, or be locally installed in the vehicle. In other words, it is not necessary to implement the motion planning module inside the vehicle.

[0009] For example, a corresponding IT infrastructure is provided in the vehicle, so that the motion planning module is provided through the IT infrastructure in the vehicle.

[0010] As an alternative, the motion planning module can also be implemented on a central server, where the server obtains vehicle parameters from the vehicle. In this case, the vehicle parameters are transmitted through a network, such as the Internet.

[0011] Generally, the following scheme can be adopted: the IT infrastructures of adjacent vehicles (i.e., vehicles within a certain maximum distance) are jointly combined, and the motion planning module is provided on the IT infrastructure of the vehicles.

[0012] This also applies to the driving strategy planning module.

[0013] The vehicle parameters obtained by the driving strategy planning module can be position data according to GNSS data (i.e., information about the current geographical location of the vehicle), acceleration and / or speed data of the vehicle, such as optical distance measurement and speed measurement through LIDAR sensors (light detection and ranging) and radar sensors, camera data, the number of passable lanes, the trajectories and / or navigation data of adjacent vehicles.

[0014] These data can be recorded by the sensors of the vehicle, sourced from adjacent vehicles, and / or transmitted from the central server to the vehicle through a wireless interface.

[0015] In a technical solution of the present invention, a driving strategy to be implemented is selected from the at least one driving strategy, and the motion planning module controls the vehicle such that the vehicle implements the driving strategy to be implemented. In this way, the evaluation module can directly evaluate the implemented driving strategy.

[0016] To take into account the uncertainties in different vehicle parameters and to evaluate the feasibility of the selected driving strategy, the motion planning module can perform reachability analysis.

[0017] The motion planning module can evaluate the quantity based on the implemented driving strategy and / or based on the determination of vehicle parameters during the implementation of the driving strategy. In this way, the evaluation quantity directly depends on the implemented driving strategy, so the decision-making sub-module can be adjusted directly according to the implementation of the driving strategy.

[0018] For example, the evaluation quantity can have: the minimum distance from at least one adjacent vehicle during the driving strategy, the maximum lateral acceleration of the vehicle during the driving strategy, the longitudinal acceleration of the vehicle, and / or the cost value of the driving strategy.

[0019] Generally, the following scheme can also be adopted: the evaluation quantity includes several values. The evaluation quantity is, for example, a vector that contains one or several of the values.

[0020] In addition, the evaluation quantity can also be a multi-dimensional vector and, for example, includes one or several of the aforementioned values for each time step of the driving strategy.

[0021] Wherein, the time step can be based on the time resolution of one or several sensors of the vehicle or can be a freely selected value, such as 10 ms.

[0022] To provide information about the implemented driving strategies of other vehicles, the knowledge data can be transmitted to other vehicles. The knowledge data includes at least vehicle parameters, the driving strategy to be implemented, the implemented driving strategy, and / or the evaluation quantity.

[0023] Based on the transmitted knowledge data, the decision-making sub-module of other vehicles can be adjusted.

[0024] In this way, several vehicles with corresponding decision-making sub-modules can use the knowledge data of the implemented driving strategy and adjust their decision-making sub-modules based on the implemented driving strategy. Accordingly, the vehicle no longer needs to compulsorily implement the driving strategy by itself, and the at least one decision-making sub-module can learn faster, thereby providing a better driving strategy.

[0025] In a technical solution of the present invention, the server determines the adjustments required for the decision-making sub-module of other vehicles based on the implemented driving strategy of one vehicle.

[0026] Alternatively, the following scheme can also be adopted: the IT infrastructure of the other vehicle determines the adjustments required for each at least one decision-making sub-module based on the implemented driving strategy.

[0027] Generally speaking, the following solution can also be adopted: the server processes and / or transmits vehicle parameters, selected driving strategies, implemented driving strategies, and / or evaluation metrics, that is, generally speaking, knowledge data.

[0028] In a technical solution of the present invention, the at least one decision sub-module determines feasible driving strategies according to a state machine ("state machine"). Accordingly, feasible driving strategies are determined through a model that describes the characteristics of the vehicle through states, state transitions, and actions. Driving on a highway, parking in a parking lot, waiting at a traffic light, driving through a traffic restricted area, turning at an intersection, etc. can be regarded as states, and the actions are the corresponding driving strategies that the vehicle can implement in the corresponding states. In this case, implementing a driving strategy may result in a state transition.

[0029] The at least one decision sub-module can also determine feasible driving strategies according to a cost function.

[0030] A cost function generally refers to a function that assigns a value to an action or a path, and this value indicates the cost or expense of the corresponding action or the corresponding path. Accordingly, by comparing several values, the optimal driving strategy can be determined, that is, the action or path with the smallest value and thus the smallest cost or expense.

[0031] For example, each lane is parameterized by a cost function, so that the at least one decision sub-module can determine whether the vehicle is in the optimal lane by comparing the cost functions of different lanes. If not, the corresponding lane change can be determined as a driving strategy.

[0032] In a technical solution of the present invention, the at least one decision sub-module determines feasible driving strategies according to a car-following model. A car-following model is a model that determines a driving strategy based on the driving strategies of other vehicles.

[0033] The at least one decision sub-module is, for example, a distance-related car-following model, and thus determines feasible driving strategies according to the distance from adjacent vehicles.

[0034] In a technical solution of the present invention, the car-following model is based on Wiedemann's psychophysical car-following model, which has four driving modes, namely free driving, approaching, following, and braking. According to the speed and the distance from adjacent vehicles, regions with free characteristics, following regions, approaching regions, braking regions, and collision regions are distinguished, and thus feasible driving strategies related to the driving modes are determined according to the region where the vehicle is located.

[0035] The at least one decision sub-module determines feasible driving strategies, for example, by means of a decision tree, that is, through decision rules.

[0036] Generally, combinations of the decision sub-modules may also be employed. Accordingly, the at least one decision sub-module can determine a feasible driving strategy based on a cost function, a decision tree, a car-following model, particularly a distance-related car-following model and / or based on Wiedemann's psychophysical car-following model, and / or a state machine. These models are proven methods for determining corresponding driving strategies in different driving situations.

[0037] At least one of the at least one decision sub-module can determine a feasible driving strategy by means of a machine learning decision process, particularly a reinforcement machine learning decision process. In this way, the at least one decision sub-module generates artificial experience from feasible driving strategies, driving strategies to be implemented, and / or implemented driving strategies. Thus, the at least one decision sub-module is trained by the driving strategy.

[0038] In the reinforcement machine learning decision process, the process autonomously learns strategies to maximize the obtained reward. For example, when a predetermined distance from an adjacent vehicle is followed during a driving strategy, the machine learning decision process obtains a reward.

[0039] An evaluation quantity can be transmitted as a reward for a driving strategy implemented in terms of road traffic order to the at least one decision sub-module. In this way, the decision sub-module can be modified to implement a driving strategy in terms of road traffic order.

[0040] In one technical solution, the machine learning decision process is a monitored or partially monitored machine learning decision process, so that the driver can monitor the feasible driving strategy.

[0041] For example, the feasible driving strategy determined by the machine learning decision process is compared with the driving strategy implemented by the driver of the vehicle.

[0042] The machine learning decision process can be implemented at least by a neural network, can be based on game theory, can be a Markov decision process and / or a partially observable Markov decision process. The English name of the partially observable Markov decision process is “partially observable Markov decision processes”.

[0043] In a Markov decision process, similar to a state machine, a vehicle is described by states, actions, and state transitions. Additionally, there is a reward function, i.e., an evaluation measure, which assigns a reward to a specific state transition, and a policy (in English: “policy”), which assigns an optimal action to each state. Additionally, there is a discount factor (in English: “discount factor”), based on which the reward that can still be achieved is reduced after an action. With the Markov decision process, the decision sub-module can independently determine a driving strategy.

[0044] In terms of game theory, in one technical solution of the present invention, the machine learning decision process can determine the feasible driving strategies of adjacent vehicles and assign probabilities to them, so as to determine at least one feasible driving strategy based on the feasible driving strategies of adjacent vehicles.

[0045] In one technical solution of the present invention, the evaluation measure is used to train the machine learning decision process of the at least one decision sub-module. In this way, the decision sub-module can be directly trained through the evaluation measure.

[0046] In other words, instead of being trained by a person, the machine learning decision process is trained based on the evaluation measure. In this way, the resources required for training can be reduced.

[0047] In order to utilize the capabilities of the various methods for determining driving strategies, several decision sub-modules can be provided, each of which determines at least one feasible driving strategy. In this case, the decision maker selects a driving strategy to be implemented from the determined feasible driving strategies, which is transmitted to the motion planning module.

[0048] In one technical solution of the present invention, the decision maker is the driver of the vehicle, a selection module, or a combination of the above.

[0049] The selection module selects the driving strategy to be implemented, for example, according to the weights of the credibility of the corresponding decision sub-module and / or the frequency of occurrence of the feasible driving strategy.

[0050] Generally, the following scheme can be adopted: the weights of the credibility are related to the number of determined driving strategies and / or the driving situation.

[0051] That is, in a specific driving situation, for example, when turning or overtaking another vehicle, the feasible driving strategy of a specific decision sub-module may be preferred.

[0052] The weights of the credibility can also be adjusted according to the evaluation measure, so that the driving strategy of the corresponding decision sub-module obtains a higher weight in the further selection process of the driving strategy, or the corresponding decision sub-module itself obtains a higher weight.

[0053] In one aspect of the present invention, the selection module has an artificial neural network and is trained accordingly to select a driving strategy in specific driving situations.

[0054] The solution of the present invention for achieving the above object further lies in a control device for a vehicle to control the vehicle, wherein the control device is adapted to implement the foregoing method. In terms of advantages and features, the foregoing description of the method equally applies to the control device.

[0055] Furthermore, the solution of the present invention for achieving the above object further lies in a vehicle, which includes the foregoing control device and a sensor for detecting at least one vehicle parameter, wherein the vehicle is adapted to implement the above method. In terms of advantages and features, the foregoing description of the method equally applies to the vehicle here.

[0056] The solution of the present invention for achieving the above object further lies in a computer program including program coding components, which is used to implement the steps of the above method when the computer program is executed on a computing unit, especially the computing unit of the above control device. The foregoing description of advantages and features equally applies to the computer program including program coding components here.

[0057] In this regard and hereinafter, "program coding components" are computer-executable instructions in the form of program code and / or program code modules in compiled and / or uncompiled form, which can exist in any programming language and / or machine language form.

[0058] The solution of the present invention for achieving the above object further lies in a system for determining a driving strategy, including at least one vehicle as described above. The foregoing description of advantages and features equally applies to the system here.

[0059] In particular, two of the above vehicles are provided for the system.

[0060] In one aspect of the present invention, the system has a server, which is connected to the at least one vehicle for data connection, wherein the motion planning module controls the vehicle such that the vehicle implements a driving strategy, wherein the server obtains and / or is provided with knowledge data from one of the vehicles, and the server transmits the knowledge data to at least another vehicle, and / or the server determines an adjustment to the decision-making sub-module of the at least another vehicle according to the knowledge data and transmits the adjustment information to the at least another vehicle. Through this system, vehicles can access each other's knowledge data, so that the decision-making sub-module can be adjusted faster and better.

[0061] That is, in particular, the decision-making sub-module of the vehicle can be adjusted based on the implemented driving strategies of other vehicles.

[0062] The knowledge data also includes at least the knowledge data of the vehicle in which the driving strategy is implemented, that is, vehicle parameters, possible driving scenarios, driving strategies to be implemented, implemented driving strategies, and / or evaluation metrics.

[0063] The server can be accessed by the vehicle through a network such as the Internet, for example. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] For more advantages and features, refer to the following description and the accompanying drawings referred to below. Among them:

[0065] - Figure 1 Is a schematic diagram of the vehicle of the present invention, including the control device of the present invention for determining the driving strategy, which is for the system of the present invention.

[0066] - Figure 2 Is Figure 1 A schematic block diagram of the control device for determining the driving strategy in

[0067] - Figure 3 Is Figure 2 A schematic block diagram of the vehicle parameter acquisition module in

[0068] - Figure 4 Is Figure 2 A schematic block diagram of the driving strategy planning module in

[0069] - Figure 5 Is Figure 2 A schematic block diagram of the motion planning module in

[0070] - Figure 6 Is a schematic block diagram of the first embodiment of the system for determining the driving strategy.

[0071] - Figure 7 Is a schematic block diagram of the second embodiment of the system for determining the driving strategy.

[0072] - Figure 8 Is a schematic block diagram of the third embodiment of the system for determining the driving strategy. DETAILED DESCRIPTION OF THE INVENTION

[0073] Figure 1 Is a schematic diagram of vehicle 10. Vehicle 10 includes several sensors 12, a control device 14, a driving assistance system 16, and several control devices 18 for controlling vehicle 10.

[0074] Vehicle 10 is, for example, an automobile for road traffic, such as a freight vehicle or a passenger vehicle.

[0075] The sensor 12 is arranged at the front, rear and / or sides of the vehicle 10 and is adapted to detect the surroundings of the vehicle 10 and / or data related to the vehicle 10. The sensor 12 generates corresponding vehicle parameters P and transmits these parameters to the control device 14, as shown by the arrows in Figure 1 .

[0076] The sensor 12 includes a position sensor 26 ( Figure 3 ), for example a satellite-based position sensor, which determines the current geographical position of the vehicle 10. The position sensor 26 accordingly generates position data which serves as a vehicle parameter P.

[0077] A camera sensor 28 ( Figure 3 ) is also provided, which takes images and / or videos of the surroundings of the vehicle 10 with the aid of a camera (not shown). The camera sensor 28 accordingly generates camera data which serves as a vehicle parameter P. Adjacent vehicles can be detected with the aid of the camera sensor 28, lane boundaries can be recognized and / or characteristic objects can be detected.

[0078] "Characteristic objects" are, for example, traffic signs and / or scenic spots in the nearby surroundings.

[0079] In addition, an acceleration and / or speed sensor 30 ( Figure 3 ) is provided, which detects the acceleration and / or speed data of the vehicle 10 which serves as a vehicle parameter P.

[0080] In addition, an optical sensor 32 ( Figure 3 ), for example a LIDAR and / or radar sensor, which generates the distance to adjacent objects and the speed of adjacent objects (such as adjacent vehicles) as vehicle parameters P.

[0081] For the sake of illustration, a separate sensor 12 is described for each of these vehicle parameters P. Of course, not every one of these vehicle parameters P requires a separate sensor 12.

[0082] The control device 14 has a data carrier 19 and a computing unit 20, wherein a computer program is stored on the data carrier 19, which is executed on the computing unit 20 and includes program code elements for determining a driving strategy according to the method described below.

[0083] The control device 14 may also include a driving assistance system 16. However, the driving assistance system 16 can also be configured as an independent system.

[0084] Generally speaking, the vehicle parameters P are transmitted to the control device 14, the control device 14 processes the vehicle parameters P obtained by the sensor 12, and controls the vehicle 10 at least partially automatically, in particular fully automatically, with the aid of the driving assistance system 16.

[0085] The vehicle parameters P can also be transmitted from adjacent vehicles and / or a central server to the control device 14 via a network.

[0086] The driving assistance system 16 can control at least partially automatically, in particular fully automatically, the lateral movement and / or the longitudinal movement of the vehicle 10. This is Figure 1 schematically illustrated by an arrow in, which shows the signal G leading to the corresponding control device 18 of the vehicle 10.

[0087] Figure 2 is a block diagram of the control device 14. The control device 14 has a vehicle parameter acquisition module 21, a driving strategy planning module 22, and a motion planning module 24.

[0088] The vehicle parameter acquisition module 21 collects and / or generates vehicle parameters P, for example, via the sensor 12, and transmits these vehicle parameters to the driving strategy planning module 22 and the motion planning module 24.

[0089] The driving strategy planning module 22 determines at least one feasible driving strategy M taking into account the vehicle parameters P, and transmits at least one feasible driving strategy M and / or the driving strategy M to be implemented AZ to the motion planning module 24.

[0090] The motion planning module 24 evaluates the at least one feasible driving strategy M and / or the driving strategy M to be implemented AZ , that is, calculates the evaluation quantity B, and generates a control command S for controlling the vehicle 10.

[0091] The control command S is transmitted to the driving assistance system 16, which in turn transmits the signal G ( Figure 1 ) to the control device 18 of the vehicle 10.

[0092] The motion planning module 24 transmits the implemented driving strategy M AG and / or the evaluation quantity B of the driving strategy to the driving strategy planning module 22, which improves the decision-making process for selecting the driving strategy M by means of the implemented driving strategy M AG and / or the evaluation quantity B, for example, by adjusting the decision-making sub-module 38 of the driving strategy planning module 22 ( Figure 4 ).

[0093] In Figures 3 to 5 , the vehicle parameter acquisition module 21, the driving strategy planning module 22, or the motion planning module 24 is shown in detail in a block diagram respectively.

[0094] As Figure 3 shown, in addition to the above-mentioned sensor 12, the vehicle parameter acquisition module 21 also includes a driving itinerary planning module 34 and a further processing module 36.

[0095] The driving route planning module 34 plans a route for the vehicle 10 to a desired destination. To this end, the driving route planning module accesses corresponding mapping data, for example stored on the data carrier 19, and can use the position data of the position sensor 26, so as to navigate from the current geographical position to the destination. That is to say, the driving route planning module 34 generates navigation data, which is also understood as vehicle parameters P within the scope of the present invention.

[0096] The further processing module 36 is adapted to combine the data of the sensor 12 and / or the vehicle parameters P, so as to obtain processed and / or more detailed vehicle parameters P.

[0097] Figure 3 The combination of position data and camera data is shown. By taking into account the number of detected lanes, detected traffic signs and / or nearby scenic spots, the determination of the current geographical position is improved in this embodiment.

[0098] It is also possible to combine the distance data from adjacent vehicles and the camera data, so as to obtain the movement trajectories of adjacent vehicles and / or predict the characteristics of adjacent vehicles. The data generated by the further processing module 36 is also understood as vehicle parameters P.

[0099] The vehicle parameters P are transmitted to the driving strategy planning module 22 and the motion planning module 24 by the vehicle parameter acquisition module 21.

[0100] In Figure 4 a schematic block diagram shows the driving strategy planning module 22, which includes a processing module 37, a decision sub-module 38, a decision maker 40 and a knowledge module 42.

[0101] The processing module 37 receives the vehicle parameters P from the vehicle parameter acquisition module 21, and is adapted to preprocess the vehicle parameters P for further use.

[0102] For example, the processing module 37 converts the vehicle parameters P to the same coordinate system. It is also possible to adopt the following scheme: the processing module 37 generates a model for the characteristics of adjacent vehicles.

[0103] The processing module 37 transmits the processed vehicle parameters P and / or the unprocessed vehicle parameters P to the decision sub-module 38.

[0104] The decision sub-module 38 respectively determines at least one feasible driving strategy M based on the vehicle parameters P. Different decision sub-modules 38 select a feasible driving strategy M based on the vehicle parameters P by different methods or models, such as according to a cost function, a decision tree, a car-following model, especially a distance-related car-following model and / or based on Wiedemann's psychophysical car-following model, and / or a state machine.

[0105] In the illustrated embodiment, there are three different decision sub-modules 38, namely the first, second, and third decision sub-modules 38a, 38b, or 38c.

[0106] The first decision sub-module 38a determines a feasible driving strategy M based on all or several of the vehicle parameters P with the aid of a state machine a .

[0107] The second decision sub-module 38b is the occupant and / or driver of the vehicle 10, who determines the feasible driving strategy M b .

[0108] In the illustrated embodiment, the third decision sub-module 38c determines the feasible driving strategy M with the aid of a reinforcement machine learning decision process c .

[0109] For example, the third decision sub-module 38c has an artificial neuron module that implements the machine learning decision process. However, the machine learning decision process can also be a Markov decision process, a partially observable Markov decision process, and / or based on game theory.

[0110] Generally, the following scheme can be adopted: the decision sub-module 38 determines more than one feasible driving strategy M, and / or different decision sub-modules 38 select the same driving strategy M. For example, the feasible driving strategy of the decision sub-module 38a (i.e., the driving strategy M a ) and the feasible driving strategy of the decision sub-module 38b (i.e., the driving strategy M b ) can be the same.

[0111] The decision sub-modules 38a, 38b, and 38c transmit the feasible driving strategies M a , M b and M c to the decision maker 40.

[0112] Among them, a weighting coefficient a corresponding to the credibility of the corresponding decision sub-module 38 is specified for each decision sub-module 38. That is, the weighting coefficient a gives the possibility that the feasible driving strategy M determined by the corresponding decision sub-module 38 is the best or optimal driving strategy.

[0113] The weighting coefficient a is not the confidence determined by each decision sub-module 38 or the decision process used during the determination of the feasible driving strategy M, but can be based on this.

[0114] The weighting coefficient a can be adjusted over time.

[0115] For example, the credibility a of the first decision sub-module 38a aThe weight of and the credibility a of the second decision sub-module 38b b The weight is initially higher than the credibility a of the third decision sub-module 38c c The weight, and the third decision sub-module determines a feasible driving strategy M through a reinforcement machine learning decision process c .

[0116] The weighting coefficient a is known to the decision maker 40, for example, stored in the memory of the decision maker 40.

[0117] The decision maker 40 mainly based on the weighting coefficient a a , a b , a c Selects a feasible driving strategy M a , M b and M c from among them, and then determines the driving strategy M to be implemented AZ , which is transmitted to the motion planning module 24.

[0118] In the illustrated embodiment, the decision maker 40 is a selection module 44, that is, a module that determines the driving strategy M to be implemented through a computer-aided selection process AZ .

[0119] The selection module 44 is in particular an artificial neuron network, where at least the feasible driving strategies M a , M b , M c , vehicle parameters P and / or the weighting coefficient a a , a b , a c are input quantities of the artificial neuron network, and the driving strategy M to be implemented AZ is at least one output quantity of the artificial neuron network.

[0120] Generally, the following scheme can be adopted: The decision-making module 38 that determines several feasible driving strategies M defines a preference to assist the decision maker 40 in the process of selecting the driving strategy M to be implemented. The decision sub-module 38a, for example, selects the preferred feasible driving strategy M and / or defines a hierarchy under the selected driving strategy M. AZ In addition, the following scheme can be adopted: The selection module 44 checks whether there already exists a similar driving situation and selects the feasible driving strategy M of the decision sub-module 38 that makes the best decision in that similar driving situation.

[0121] It can also adopt the following scheme: The decision maker 40 is an occupant and / or driver of the vehicle 10, who views the feasible driving strategies M on a display screen and makes a decision by selecting and / or implementing the driving strategy M.

[0122] ​

[0123] Subsequently, the driving strategy M to be implemented AZ is transmitted to the motion planning module 24.

[0124] Figure 5 FIG. 24 is a schematic block diagram of the motion planning module 24, which includes an analysis module 46 and an evaluation module 48 in the illustrated embodiment.

[0125] As described above, the motion planning module 24 obtains the vehicle parameter P from the vehicle parameter acquisition module 21 and the driving strategy M to be implemented from the driving strategy planning module 22 AZ .

[0126] The analysis module 46 analyzes the driving strategy M to be implemented for feasibility mainly based on the vehicle parameter P AZ . In particular, the analysis module is capable of estimating whether the vehicle 10 will collide with an object, such as another vehicle, during the implementation of the driving strategy.

[0127] The analysis module 46 estimates the feasibility of the driving strategy, for example, using the uncertainty in the trajectories of adjacent vehicles, the uncertainty in the speed of the vehicle 10, and the uncertainty in the direction of the lane.

[0128] The analysis module 46 is also capable of determining whether the end of the driving strategy can be reached, for example, due to physical limits or comfort limits for longitudinal and lateral accelerations.

[0129] Based on the feasibility analysis and / or reachability analysis of the analysis module 46, the driving strategy M implemented by the vehicle 10 is then determined AG . For this purpose, the analysis module 46 provides a control command S, which is transmitted to the driving assistance system 16.

[0130] The evaluation module 48 then generates an evaluation quantity B for the driving strategy.

[0131] The evaluation module 48 evaluates the feasible driving strategy M, the driving strategy M to be implemented AZ and / or the implemented driving strategy M AG . For this purpose, the evaluation module 48 mainly uses the vehicle parameter P.

[0132] To evaluate the driving strategy M to be implemented AZ or the implemented driving strategy M AG , the evaluation module 48 stores and / or analyzes the vehicle parameter P determined during the driving strategy M to be implemented AZ or the implemented driving strategy M AG , and / or retrospectively analyzes the feasibility analysis result of the analysis module 46.

[0133] For example, the evaluation quantity B includes the minimum distance to an adjacent vehicle, the maximum lateral acceleration of the vehicle 10 during the driving strategy, and / or a value indicating whether the maximum speed of the road traffic order is followed.

[0134] Of course, other values or parameters can also be part of the evaluation quantity B.

[0135] The evaluation quantity B thus contains several values. The evaluation quantity B is, for example, a vector containing the said values.

[0136] The evaluation quantity B can also be a multi-dimensional vector and, for example, includes one or several of the aforementioned values for each time step of the driving strategy.

[0137] After the evaluation module 48 creates the evaluation quantity B, the motion planning module 24 transmits the implemented driving strategy M AG and the evaluation quantity B back to the driving strategy planning module 22.

[0138] Subsequently, the evaluation quantity B is transmitted to the decision sub-module 38c, and the decision sub-module 38c is adjusted taking the evaluation quantity B into account. More precisely, the machine learning decision process, in particular the artificial neural network that determines the feasible driving strategy M c is trained with the evaluation quantity B.

[0139] The selection module 44, in particular the artificial neural network of the selection module 44, can also be trained and adjusted with the evaluation quantity B.

[0140] The weighting coefficients a a 、a b 、a c of the credibility of the decision sub-modules 38a, 38b or 38c can also be adjusted according to the evaluation quantity B, for example by the selection module 44 or the decision maker 40.

[0141] As Figure 4 shown, the knowledge module 42 of the driving strategy planning module 22 is adapted to provide knowledge data E. The knowledge data E includes, for example, the feasible driving strategies M a 、M b 、M c , the driving strategy M to be implemented, the vehicle parameters P, the implemented driving strategy M AZ , and the evaluation quantity B. AG

[0142] The knowledge module 42 stores and analyzes, for example, the vehicle parameters P measured during the implemented strategy M AG so as to transmit only the relevant data of the vehicle parameters P as the knowledge data E.

[0143] The knowledge data E can also have more precise data for the selection process of the decision maker 40.

[0144] As shown by the arrow, the knowledge data E is provided by the driving strategy planning module 22, so that, for example, other vehicles can trace the knowledge data E of the vehicle 10.

[0145] In addition, the knowledge data E can be used to adjust one of the decision sub-modules 38, for example, to train the machine learning decision process.

[0146] The method described above will be described in detail below in connection with a highly simplified example of a lane change of the vehicle 10 on a highway.

[0147] On the lane on which the vehicle 10 is traveling, the sensor 12 of the vehicle 10, for example, identifies another vehicle. The other vehicle moves in front of the vehicle 10 at a certain distance and at a speed lower than that of the vehicle 10. If the vehicle 10 follows the lane at a constant speed, a collision will occur between the vehicle 10 and the other vehicle.

[0148] That is, mainly the speed of the other vehicle and the distance from the other vehicle are transmitted as vehicle parameters P to the driving strategy planning module 22. In addition, the adjacent lane (overtaking lane) is also identified by the sensor 12 as not being occupied.

[0149] Based on these vehicle parameters P, the driving strategy planning module 22, in particular the decision sub-modules 38a, 38b, 38c, determines feasible driving strategies M a , M b and M c . A feasible driving strategy M a can be: the vehicle 10 significantly reduces its speed and adjusts according to the speed of the other vehicle. Another feasible driving strategy M b is: the vehicle 10 makes a lane change to overtake the other vehicle.

[0150] The feasible driving strategies M a , M b and M c are transmitted to the decision maker 40, and the decision maker selects one of the driving strategies M a , M b , M c , for example, the first feasible driving strategy M a . This driving strategy then becomes the driving strategy M AZ to be implemented.

[0151] The driving strategy planning module 22 transmits at least the driving strategy M AZ to be implemented to the motion planning module 24, which transmits the corresponding control command S to the driving assistance system 16.

[0152] In the described embodiment, the driving strategy planning module 22 transmits a very unspecific driving strategy M to be implemented, AZ i.e., transmits the driving strategy for slowing down, and the motion planning module 24 converts this instruction or target into a specific control command S.

[0153] Generally, precise instructions can also be used. For example, first, the vehicle 10 is made to pass by another vehicle in an adjacent lane, then the speed is increased by a certain value, and at the same time, it is switched to the adjacent lane.

[0154] In addition, the driving strategy planning module 22 can define a section, such as a time section or a distance section, in which the motion planning module 24 needs to implement the driving strategy.

[0155] For example, a stationary vehicle in the lane is identified by the sensor 12, and a request is sent to the motion planning module 24: the vehicle 10 must switch lanes or stop within the next 5 seconds and / or within the next 200 m.

[0156] In addition, the driving strategy planning module 22 can also plan the feasible driving strategy M in great detail and, for example, have the corresponding control command S and / or signal G for the control device 18 of the vehicle 10.

[0157] The motion planning module 24 measures an evaluation quantity B, which is negative in this example. The motion planning module 24 penalizes the selection of the driving strategy M a because the driving strategy M a requires more time to reach the driving destination and must be braked.

[0158] Therefore, the driving strategy planning module 22 obtains the negative evaluation quantity B through the motion planning module 24 and adjusts the decision-making sub-module 38 accordingly, so that the decision-making sub-module 38 or the decision-maker 40 (for example, by adjusting the weighting factors a a 、a b 、a c ) selects the driving strategy M b .

[0159] In this way, the actual experience is taken into account in the process of selecting the driving strategy.

[0160] Next, three embodiments of the system 50 for determining the driving strategy are presented in conjunction with Figures 6 to 8 the display.

[0161] System 50 has several vehicles 10, which are generally constructed as described above, so only the differences will be described below. The same and equivalent components and modules are denoted by the same reference numerals.

[0162] Figure 6 is a schematic block diagram of a first embodiment of system 50. Figure 6 The system 50 in AG has three vehicles 10, namely a vehicle 10a implementing a driving strategy M

[0163] Similar to the embodiment of the control device 14 described in Figures 1 to 5 The control device 14 of vehicle 10a, more precisely, determines at least one feasible driving strategy and / or the driving strategy to be implemented, implements the driving strategy with the aid of the driving assistance system 16, and generates corresponding knowledge data E.

[0164] Subsequently, the knowledge data E is transmitted to the driving strategy planning module 22 of vehicle 10b, so that the driving strategy planning module 22 of vehicle 10b can adjust its corresponding decision sub-module 38.

[0165] In other words, the decision sub-module 38 of vehicle 10b is adjusted according to the driving strategy implemented by vehicle 10a.

[0166] The data exchange between vehicles 10, that is, the transmission of knowledge data, can be carried out through a wireless interface.

[0167] Figure 7 is a schematic block diagram of a second embodiment of system 50 for determining a driving strategy. The illustrated embodiment is generally equivalent to the embodiment as Figure 6 shown, so only the differences will be described below.

[0168] In the embodiment as Figure 7 shown, only the vehicle parameter acquisition module 21 and the driving assistance system 16 are implemented in vehicles 10a and 10b.

[0169] There is an independent stationary server 60 on which the driving strategy planning module 22a of vehicle 10a, the driving strategy planning module 22b of vehicle 10b and a shared motion planning module 24 are implemented.

[0170] The server 60 communicates with the vehicles 10, for example, via a wireless connection such as a mobile radio network. Corresponding transmitting and receiving devices (not shown in Figure 7 ) are provided for the server 60 and vehicles 10a and 10b.

[0171] With the help of a wireless connection, the driving strategy planning module 22a obtains the vehicle parameters P of the vehicle 10a a , and the driving strategy planning module 22b obtains the vehicle parameters P of the vehicle 10b b .

[0172] As described above, the driving strategy planning module 22a or 22b provides the driving strategy M to be implemented for the vehicle 10a AZ,a or provides the driving strategy M to be implemented for the vehicle 10b AZ,b . The motion planning module 24 determines the driving strategy to be implemented

[0173] In the embodiment as Figure 7 shown, the motion planning module 24 determines the driving strategy M for the vehicle 10a AG,a , and does not determine the driving strategy for the vehicle 10b. The motion planning module 24 transmits the corresponding control command S a to the driving assistance system 16 of the vehicle 10a

[0174] In addition, only the knowledge module 42 implemented on the server side is provided. The knowledge module 42 is based on the driving strategy M AZ,a and M AZ,b , the vehicle parameters P a and P b , the evaluation quantity B and / or the implemented driving strategy M AG,a to generate the knowledge data E

[0175] The knowledge data E is transmitted to the adjustment module 62 that provides the adjustment information A, so as to adjust the driving strategy planning module 22a of the vehicle 10a with the help of the adjustment information A a , and to adjust the driving strategy planning module 22b of the vehicle 10b with the help of the adjustment information A b

[0176] In other words, the server 60 makes corresponding adjustments to the decision sub-modules of the driving strategy planning modules 22a and 22b

[0177] Figure 8 is a schematic block diagram of the third embodiment of the system 50

[0178] Differing from the embodiment as Figure 7 shown, the driving strategy planning modules 22 of the vehicle 10 are all implemented in the vehicle 10, rather than on the server 60

[0179] Correspondingly, the server 60 transmits the adjustment information A a and A b to each driving strategy planning module 22 of the vehicle 10, so as to adjust with the help of the adjustment information A a ​At least one decision sub-module of the driving strategy planning module 22 of vehicle 10a and with the aid of adjustment information A b Adjust at least one decision sub-module of the driving strategy planning module 22 of vehicle 10ba.

[0180] For the sake of simplicity, only two vehicles 10 are shown respectively for the Figure 6 、 Figure 7 and Figure 8 embodiments. Vehicle 10a implements the driving strategy, and the driving strategy planning module 22b of vehicle 10b is adjusted with the aid of knowledge data E. Generally, any number of vehicles 10a and 10b can be used.

[0181] "Fleet learning" can be achieved through system 50, that is, based on the experience of other vehicles (i.e., the driving strategies planned and / or implemented in specific traffic situations), a vehicle can adjust its driving strategy planning module 22, especially its decision sub-module 38, without having experienced some of these traffic situations.

[0182] Figure 7 and Figure 8 are used as illustrative diagrams. Server 60 is used to illustrate that the motion planning module 24 and / or the driving strategy planning module 22 of system 50 can also be implemented on server 60, and this server serves as the central control center for traffic.

[0183] Generally, each module of system 50, such as analysis module 46, driving itinerary planning 34, knowledge module 42, and / or evaluation module 48, can also be implemented only on server 60.

[0184] In other words, it is not necessary to implement all modules of system 50 in vehicle 10.

[0185] For the sake of explanation, in the description of the embodiments, a distinction is made between the feasible driving strategy M, the driving strategy M to be implemented AZ and the implemented driving strategy M AG . Generally, these driving strategies do not need to be different, but can be the same.

[0186] Analysis module 46 can in particular only determine whether a driving strategy can be implemented. In the case where a driving strategy cannot be implemented, the motion planning module 24 and / or the analysis module 46 issue an instruction to the driving strategy planning module 22, instructing it to provide a new driving strategy and then analyze this driving strategy.

Claims

1. A method for determining a driving strategy of a vehicle, the method comprises: receiving vehicle parameters generated by a vehicle parameter acquisition module of a first vehicle; determining, via a driving strategy planning module of the first vehicle, at least one feasible driving strategy based on the received vehicle parameters, an evaluation measure, and second knowledge data of one or more second vehicles, the second knowledge data including data characterizing driving strategies implemented by the one or more second vehicles, the at least one feasible driving strategy being determined by at least one decision sub-module of the driving strategy planning module; transmitting a driving strategy to be implemented selected from the at least one feasible driving strategy from the driving strategy planning module to a motion planning module of the first vehicle; generating the evaluation measure via the motion planning module based on at least one driving strategy previously implemented by the first vehicle; controlling, via the motion planning module, the first vehicle to implement a driving strategy related to the driving strategy to be implemented; and transmitting first knowledge data including at least the vehicle parameters, the driving strategy to be implemented, the implemented driving strategy, and / or the evaluation measure from the first vehicle to a driving strategy planning module of the one or more second vehicles to allow decision sub-modules of the one or more second vehicles to adjust and implement driving strategies of the one or more second vehicles according to the driving strategy implemented by the first vehicle.

2. The method according to claim 1, wherein, the driving strategy to be implemented is selected from the at least one feasible driving strategy determined by the driving strategy planning module based on the vehicle parameters, the evaluation measure, and the first knowledge data and / or the second knowledge data.

3. The method according to claim 2, wherein, the evaluation measure is determined by the motion planning module based on a previously implemented driving strategy and / or vehicle parameters during implementation of the previously implemented driving strategy, wherein the evaluation measure has a minimum distance from at least one adjacent vehicle, a maximum lateral acceleration, a maximum longitudinal acceleration, and / or a cost value of the driving strategy.

4. The method according to any one of claims 1 or 3, wherein, the one or more second vehicles transmit the second knowledge data to the first vehicle.

5. The method according to any one of claims 1 to 3, wherein, the at least one decision sub-module determines the feasible driving strategy according to a cost function, a decision tree, a car-following model, and / or a state machine.

6. The method according to claim 5, wherein, the car-following model is a distance-related car-following model and / or a psychophysical car-following model according to Wiedemann.

7. The method according to any one of claims 1 to 3, wherein, at least one of the at least one decision sub-module determines the feasible driving strategy by means of a machine learning decision process.

8. The method according to claim 7, wherein, the machine learning decision process is a reinforcement machine learning decision process.

9. The method according to claim 7, wherein, The machine learning decision-making process is implemented at least through a neural network and is a Markov decision process and / or a partially observable Markov decision process based on game theory.

10. The method according to any one of claims 1 to 3, wherein, the evaluation measure is used to train the machine learning decision-making process of the at least one decision sub-module.

11. The method according to claim 1, wherein, a plurality of decision sub-modules are provided, and the plurality of decision sub-modules respectively determine at least one feasible driving strategy, wherein the decision maker selects the driving strategy to be implemented from the at least one feasible driving strategy determined, and the driving strategy to be implemented is transmitted to the motion planning module.

12. The method according to claim 11, wherein, the decision maker is a driver of the vehicle, a selection module or a combination of the above, wherein the selection module selects the driving strategy to be implemented according to the weight of the credibility of the corresponding decision sub-module and / or the occurrence frequency of the feasible driving strategy.

13. The method according to claim 12, wherein, the selection module has an artificial neural network.

14. A control device for a vehicle for controlling the vehicle, wherein, the control device is adapted to implement the method according to any one of claims 1 to 13.

15. A vehicle comprising the control device according to claim 14 and at least one sensor for detecting at least one vehicle parameter, wherein, the vehicle is adapted to implement the method according to any one of claims 1 to 13.

16. A computer program product comprising program coding components for implementing the method according to any one of claims 1 to 13 when the computer program product is executed on a computing unit of the control device according to claim 14.

17. A system for determining a driving strategy, comprising at least one vehicle according to claim 15, wherein, the at least one vehicle includes the first vehicle.

18. The system according to claim 17, wherein, the system has a server which is connected to the at least one vehicle for data connection, wherein the motion planning module controls the at least one vehicle such that the at least one vehicle implements a driving strategy, wherein the server obtains and / or provides the knowledge data, and the server transmits the knowledge data to at least another vehicle, and / or the server adjusts the decision sub-module of the at least another vehicle according to the knowledge data and transmits the adjustment information to the at least another vehicle.

Images