Method for determining the most probable path of a motor vehicle

Abstract

Method for determining a most probable travel path of a motor vehicle, the method (1, 20, 30) comprising the following steps: - detecting absolute position data of a motor vehicle by means of an absolute positioning system (2,21); - mapping the absolute position data onto a road contained in digital map material (3,22); - Recording of current vehicle-specific data (4.23); - Providing information about the road contained in the digital map material and the current vehicle-specific data to an artificial intelligence module, wherein the artificial intelligence module is trained to determine a most probable driving path by artificial intelligence based on a road on which the motor vehicle is currently located and vehicle-specific data (5,24); and - Determining the most probable travel path of the motor vehicle by the artificial intelligence module based on the information about the road contained in the digital map material and the current vehicle-specific data (6, 25), wherein it is first checked whether sufficient vehicle-specific data is available so that the artificial intelligence module can determine the most probable travel path based on the information about the road contained in the digital map material and the current vehicle-specific data, wherein if sufficient vehicle-specific data is available, the most probable travel path is determined by the artificial intelligence module based on the information about the road contained in the digital map material and the current vehicle-specific data, and wherein if insufficient vehicle-specific data is available,Fleet values ​​or estimates are used to determine the most likely driving path.

Description

[0001] The invention relates to a method for determining a most probable travel path of a motor vehicle and in particular to a method for determining a most probable travel path of a motor vehicle using artificial intelligence.

[0002] To increase safety when driving a motor vehicle, driver assistance systems in motor vehicles, especially in autonomous vehicles, should consider not only information about the immediate surroundings of the vehicle, but also information about the further, likely upcoming route. For example, safety when driving an autonomous vehicle can be increased if a likely upcoming route is taken into account when controlling driver assistance functions of the autonomous vehicle, for example when controlling a speed limit assistant or Speed ​​Limit Assist, when controlling adaptive headlights, or when shifting an automatic transmission. This can significantly increase safety for the driver of the vehicle and other road users.This also results in energy savings, especially since the driver assistance system can always be adapted directly to the expected conditions and the expected route.

[0003] Usually, a most probable path, also called Most Probable Path (MPP), is determined, in particular on the basis of digital map material, i.e. a digital road network, the current position of the motor vehicle and the direction of travel of the motor vehicle by a module, for example a horizon provider or a driver horizon system.

[0004] From the publication DE 10 2014 223 331 A1, a method and a device for generating the most probable travel path are known. These methods and devices are configured to generate a travel horizon based on the most probable travel path only after the most probable travel path has been generated. In particular, the method and device are configured to reduce the frequency of generating the most probable travel path and travel horizon.

[0005] From the publication DE 10 2018 209 804 A1, a method for predicting a probable travel route for a vehicle is known. The method initially comprises a prediction step in which a travel destination is predicted using at least one data signal representing a past location of the vehicle and a position signal representing a current position of the vehicle. The method further comprises a creation step in which a reward map is created for depicting a weighting of at least one area between the current position of the vehicle and the predicted travel destination using a driving history of the vehicle and / or a driving preference of the driver.Finally, the method comprises a calculation step in which the probable driving route is calculated using a neural network and the created reward map to thereby predict the probable driving route.

[0006] From the publication DE 10 2019 114 737 A1, a device for predicting the behavior of a road user in a traffic situation is known. The device is configured to determine feature values ​​for a plurality of features at a given point in time. Furthermore, the device is configured to predict an action of the road user using at least one neural network. The neural network is configured to receive the feature values ​​for the plurality of features as input values ​​of the neural network and to provide the road user's action as the output value of the neural network.

[0007] From the document DE 10 2006 057 920 A1 a method is known for controlling the display of a navigation system in a mode in which no route and no destination are entered, wherein the most probable route for a journey with the navigation system is determined and the surroundings along the most probable route are highlighted on the map.

[0008] The object of embodiments of the invention is to provide an improved method for determining a most probable travel path of a motor vehicle.

[0009] The problem is solved by the subject matter of the subordinate claims. Further advantageous developments are the subject matter of the dependent claims.

[0010] According to one embodiment of the invention, this object is achieved by a method for determining a most probable travel path of a motor vehicle, wherein absolute position data of a motor vehicle are recorded by means of an absolute positioning system and the absolute position data are mapped onto a road contained in digital map material, and current motor vehicle-specific data are recorded, wherein information about the road contained in the digital map material and the current motor vehicle-specific data are provided or transmitted to an artificial intelligence module.The artificial intelligence module is trained to determine a most probable driving path using artificial intelligence based on a road on which the motor vehicle is currently located and vehicle-specific data, whereby the most probable driving path of the motor vehicle is determined by the artificial intelligence module based on the information about the road contained in the digital map material and the current vehicle-specific data.

[0011] Absolute position data indicates the measured position of the vehicle at a specific time in absolute values, for example, in a UTM or WGS84 reference coordinate system. Optionally, the absolute position data can also be provided with an orientation, which indicates, for example, the current direction of movement of the vehicle. A combination of position and orientation is often referred to as a pose. Such absolute positioning systems can, for example, be satellite-based technologies, such as a Global Positioning System (GPS) or a Global Navigation Satellite System (GNSS).

[0012] Digital map material is further understood to mean map material in digital form that contains geometric, topological, and semantic information about a road network and that can be used and processed, for example, by a motor vehicle's navigation system or driver assistance systems. The digital map material can be enriched with additional information, such as speed information, e.g., speed limits, or warnings. The digital map material is updateable. The mapping of the absolute position data to a road contained in the digital map material can be achieved, for example, using a map matching algorithm, i.e., a process that compares the absolute position data with location information from a digital map, i.e., the digital map material.

[0013] Motor vehicle-specific data also includes data that is specific or characteristic of the use, i.e. the users or drivers of the motor vehicle, and in particular data that characterise certain properties, such as seat occupancy, travel times and weight distribution, which may influence the actual route selection, whereby this influence may have been recorded during previous journeys with the motor vehicle.

[0014] An artificial intelligence module also refers to a module or a control unit that is designed to learn a motor vehicle's behavior and, in particular, to learn how vehicle-specific data affects the selection of a route to be traveled or of roads traveled, and thus the determination of the most probable travel path. The artificial intelligence is generated, for example, through machine learning, i.e., trained during (training) trips with the motor vehicle. In addition or alternatively, the artificial intelligence can be trained in a backend in a manner known from the prior art. Such trained artificial intelligence can then be used to easily predict or determine a most probable travel path for the motor vehicle.For example, such artificial intelligence can be based on an artificial (neural) network, whereby such a network is capable of better modeling correlations between input variables, even in the presence of nonlinearities, while simultaneously minimizing the computational effort required to determine the most probable path. Furthermore, no special modeling knowledge is required within the system itself.

[0015] Information about the road contained in the digital map material also includes data that identify the road contained in the digital map material and usually also a current position of the motor vehicle and which can be processed by the artificial intelligence module.

[0016] Overall, a method is thus specified in which, in addition to digital map material, the current position of the motor vehicle and the direction of travel of the motor vehicle, vehicle-specific data are also taken into account when determining the most probable travel path. For example, it may be advantageous and preferable to avoid routes with steep gradients in certain weather conditions or load states and to choose alternative routes instead, wherein, according to the invention, corresponding behavior learned from previous journeys with the same motor vehicle when certain vehicle-specific data is available can be taken into account when determining the most probable travel path. Overall, an improved method for determining a most probable travel path of a motor vehicle is thus specified and, in particular, a method with which the accuracy of determining the most probable travel path can be increased.

[0017] Because the determination of the most probable route is based on artificial intelligence, the determination of the most probable route can also be carried out quickly and efficiently with little computational effort, whereby the determination of the most probable route is not based on a computer-aided trajectory but on learned values ​​for individual possible routes, which reflect the actual behavior of users of the motor vehicle.

[0018] The improved or more accurate most probable driving path can then be used or processed by driver assistance functions or driver assistance systems and other electronic components of the motor vehicle in order to further increase the safety of driving the motor vehicle.

[0019] In particular, the method can further comprise providing the determined most probable path to driver assistance systems and / or electronic components of the motor vehicle. Providing here means that the improved or more precise most probable path is transmitted to the driver assistance systems, whereby the transmission can be wired or wireless. The driver assistance systems, for example a speed limit assistant, a control system for adaptive headlights, or a control system for an automatic transmission, can subsequently use and process the determined more precise most probable path to control the corresponding functions of the motor vehicle accordingly, thereby increasing the accuracy of controlling the corresponding functions of the motor vehicle.This can significantly increase safety for the driver and other road users. It also results in energy savings, especially since the driver assistance system can always be adapted directly to the anticipated conditions.

[0020] The vehicle-specific data may also include time data, seat occupancy data, loading data, charge level data, fuel data, calendar data and / or weather data.

[0021] The time data may, for example, be times of day, days of the week and / or months, so that information can be obtained about streets or routes selected at certain times of day, days of the week and / or months.

[0022] Seat occupancy data refers, in particular, to data indicating the occupancy of individual seats in a motor vehicle, which can be detected, for example, using a seat occupancy sensor, such as a pressure-sensing seat occupancy sensor. Thus, a motor vehicle user can usually choose a fast route if they are traveling alone, but a different, quieter route if there are other people in the vehicle.

[0023] Loading data, in turn, refers to data that characterizes the load of the motor vehicle, for example, what the motor vehicle is carrying and / or the total weight of the load, and which can be detected, for example, by a load sensor on the motor vehicle and / or an optical sensor. For example, a user can usually choose a different route when there is no additional load in the motor vehicle than when the motor vehicle is loaded, thereby avoiding larger increases, for example, especially if the motor vehicle is fully loaded.

[0024] Charge level data refers, in particular, to data relating to the charge level of a storage device used to store electrical energy for powering an electric or hybrid vehicle, and thus to the range of the electric or hybrid vehicle. This allows a user to preferentially choose routes along which electric charging stations are located, especially when the remaining range is limited.

[0025] Fuel data also refers to data indicating the fuel level of a motor vehicle's fuel tank, and thus data indicating how far the vehicle can likely travel with the fuel level in the tank. For example, if the fuel level in the vehicle's tank is low, a motor vehicle user can preferentially choose a route along which gas stations are located.

[0026] Based on calendar data, it can also be recorded whether the user is on the way to a scheduled appointment, for example a meeting, whereby the user would, for example, prefer to choose a faster route to the location where the appointment or meeting is taking place.

[0027] Furthermore, the user of the motor vehicle may prefer different routes depending on the weather conditions.

[0028] For example, during heavy snowfall, a user may prefer streets that are cleared more frequently than others.

[0029] However, the fact that the vehicle-specific data includes time data, seat occupancy data, load data, charge level data, fuel data, calendar data, and / or weather data is merely a non-limiting embodiment. In particular, the vehicle-specific data can also include all other data depending on which users of the vehicle typically choose different routes.

[0030] According to one embodiment, the artificial intelligence module is trained to determine the most probable travel path based on learned conditional probabilities that are dependent on vehicle-specific data and indicate how frequently users of the motor vehicle have followed a road in the past, given certain vehicle-specific data, starting from an intersection point contained in the digital map material, from which users of the motor vehicle can follow different roads. The step of determining the most probable travel path of the motor vehicle by the artificial intelligence module comprises providing corresponding learned conditional probabilities based on the current vehicle-specific data and determining the most probable travel path based on the corresponding learned conditional probabilities.

[0031] Conditional probability is generally understood as the probability that a specific event occurs, for example that users follow a road, under the assumption that another event has already occurred, for example that certain vehicle-specific data are available or given.

[0032] An intersection point is also a point in the digital map material at which two or more roads intersect, and in particular a point in the digital map material from which users of the motor vehicle can take or drive on different roads, i.e., follow them. Furthermore, points from which users of the motor vehicle have driven in different directions in the past are also considered intersection points.

[0033] Overall, the most probable driving path can thus be determined based on stored learned values, which can significantly increase the accuracy and speed in determining the most probable driving path, which is particularly advantageous in safety-critical conditions, for example when using autonomously driving vehicles.

[0034] The step of providing corresponding learned conditional probabilities based on the current vehicle-specific data may comprise providing one or more appropriately trained ML models, for example one or more appropriately trained artificial (neural networks).

[0035] In particular, the artificial intelligence module can thus be designed to determine the most probable path based on one or more artificial (neural) networks. An artificial network is, in particular, a network of artificial neurons simulated in a computer program. The artificial neural network is typically based on a network of several artificial neurons. The artificial neurons are typically arranged in different layers. The artificial neural network usually comprises an input layer and an output layer, whose neuron output is the only one of the artificial neural network that is visible. Layers lying between the input layer and the output layer are typically referred to as hidden layers.Typically, an architecture and / or topology of an artificial neural network is first initiated and then trained in a training phase for a specific task or for multiple tasks. Training the artificial neural network typically involves changing the weight of a connection between two artificial neurons of the artificial neural network. Training the artificial neural network can also involve developing new connections between artificial neurons, deleting existing connections between artificial neurons, adjusting thresholds of the artificial neurons, and / or adding or deleting artificial neurons.

[0036] Such an artificial network has the advantage that it can easily process even large amounts of data. Such an artificial network is capable of improved modeling of correlations between the input variables—that is, the road contained in the digital map material and the current vehicle-specific data—even in the presence of non-linearities, while simultaneously keeping the necessary computational effort low. This means that the method can also be carried out using conventional computing units and corresponding processors, for example, control units integrated into the vehicle. A separate artificial network can be trained starting from each intersection point or from each point from which users have followed different roads or set off in different directions in the past.Furthermore, a single artificial network can also be trained for the entire road network contained in the digital map material.

[0037] Furthermore, the conditional probabilities, i.e. the relative frequencies with which users of the motor vehicle have followed different roads in the past starting from an intersection point when certain vehicle-specific data are available, can be determined or learned, for example, based on first- or n-th-order Markov models, or based on clustering or cluster analysis using further attributes and using a Markov model.

[0038] When providing the corresponding learned conditional probabilities, roads with a learned conditional probability lower than a specified threshold cannot be considered. This means that roads that vehicle users have rarely traveled in the past, even when the same vehicle-specific data is available, are not considered or provided.This has the advantage of reducing the amount of data to be processed during the determination of the most probable travel path, ensuring, in particular, that the method can also be executed by a motor vehicle control unit, which typically has lower computing power and lower storage capacity than comparable control units outside the motor vehicle. Thus, no data, in particular no personal data, needs to be transmitted to an external control unit or an external server. The threshold value can be preset by a motor vehicle manufacturer and be, for example, 0.05. Furthermore, the threshold value can also be selected and set by a motor vehicle user themselves.

[0039] The artificial intelligence module can also be designed to continuously update the learned conditional probabilities, with recent motor vehicle journeys being given greater weight when learning or updating the conditional probabilities than longer-ago motor vehicle journeys. This ensures that the conditional probabilities are learned and updated in such a way that they are always adapted to the current practices and habits of the driver or user of the motor vehicle. Recent journeys could, for example, be training trips or motor vehicle journeys made within the last few months, for example, in the last six months.

[0040] In one embodiment, the step of determining the most probable travel path based on the corresponding learned conditional probabilities further comprises selecting a route or a road which has the highest conditional probability starting from a starting position corresponding to the current absolute position data of the motor vehicle as part of the most probable travel path, wherein subsequently, at intersection points along the most probable travel path, the road leading from the corresponding intersection point with the highest conditional probability is selected as part of the most probable travel path.Thus, at any given time, the road or route is selected that, based on the current vehicle-specific data, has the highest conditional probability of all roads leading from a corresponding intersection point, with the most probable route being formed by the correspondingly selected routes or roads. This has the advantage that the computing power and time required to determine the most probable route can be kept extremely low, thus further minimizing the requirements for a data processing system on which the method is executed. The method can, for example, be implemented entirely on control units integrated into the vehicle itself.

[0041] In a further embodiment, the step of determining the most probable travel path based on the corresponding learned conditional probabilities for one or more possible destination positions comprises determining all possible routes between a start position corresponding to current absolute position data of the motor vehicle and the corresponding destination position, wherein the route between the start position and a destination position of the one or more possible destination positions is subsequently selected as the most probable travel path, which route has the greatest overall probability of all the determined routes.

[0042] A route is understood as a possible path between the starting position and a destination position, whereby this path consists of a combination of individual routes or individual streets.

[0043] Overall probability also refers to the probability with which users of the motor vehicle follow exactly this combination of individual routes or roads or the relative frequency with which users of the motor vehicle in the past, when the same vehicle-specific data were available, have followed exactly this combination of individual routes or roads, i.e. have driven exactly these roads one after the other.

[0044] This allows the accuracy of determining the most probable path between a starting point identified by the current absolute position data and a possible target position to be further improved in an efficient manner and with low computing power.

[0045] The one or more possible destination positions can be one or more predetermined destination points, one or more positions reachable after a certain time based on the digital map material, or one or more positions reachable after traveling a certain distance based on the digital map material. The predetermined destination points can be specified by a user of the motor vehicle, for example, using a navigation system of the motor vehicle. Furthermore, the predetermined destination point can also be determined from previous journeys with the motor vehicle, for example by determining that users of the motor vehicle always drove to the same destination point at a certain time on a certain day of the week.The specific time and / or distance can be preset by the manufacturer of the motor vehicle or can be freely selected by the user of the motor vehicle. The specific time and / or distance can be preset in such a way that system requirements are met and the method can be executed, in particular, on a control unit within the motor vehicle, i.e., on control units fully integrated into the motor vehicle.

[0046] In one embodiment, the artificial intelligence module is further configured and trained to determine the most probable travel path based on a road onto which absolute position data of the motor vehicle are mapped, motor vehicle-specific data, and user-specific data, wherein the method further comprises determining a driver of the motor vehicle, determining user-specific data associated with the driver of the motor vehicle, and determining the most probable travel path for the motor vehicle by the artificial intelligence module based on the information about the road contained in the digital map material, the current motor vehicle-specific data, and the user-specific data.

[0047] While motor vehicle-specific data refers to information or data that is based on all journeys made with the motor vehicle, i.e. is independent of which user actually used the motor vehicle, user-specific data refers to data that has been recorded separately for each individual driver or user of the motor vehicle.

[0048] The fact that the artificial intelligence module is designed and trained to determine the most probable driving path based on a road onto which absolute position data of the motor vehicle is mapped, vehicle-specific data and user-specific data can mean, for example, that a separate artificial network is trained for each individual user or driver of the motor vehicle and stored in the artificial intelligence module.

[0049] Thus, the method can also be further designed in such a way that the artificial intelligence module is trained separately for each individual driver or user of the motor vehicle, whereby the determination of the most probable driving path can be adapted to the practices or habits of the individual driver or user and thus the accuracy in determining the most probable driving path can be further increased.

[0050] The identification of a driver or user of the motor vehicle can be carried out in all common ways, for example by a user logging into a corresponding system before using an autonomous vehicle.

[0051] A further embodiment of the invention also provides a control device for determining a most probable travel path of a motor vehicle, which control device is designed to carry out a method described above.

[0052] Such a control unit has the advantage of being designed to execute a method in which, in addition to digital map material, the current position of the motor vehicle, and the direction of travel of the motor vehicle, vehicle-specific data are also taken into account when determining the most probable route. For example, it may be advantageous and preferable to avoid routes with steep gradients under certain weather conditions or load conditions and instead select alternative routes. According to the invention, corresponding behavior learned from previous journeys with the same motor vehicle can be taken into account when determining the most probable route.Overall, an improved method for determining a most probable travel path of a motor vehicle by the control unit is thus carried out and, in particular, a method with which the accuracy in determining the most probable travel path can be increased.

[0053] Because the determination of the most probable driving path is based on artificial intelligence, the determination of the most probable driving path can also be carried out quickly and efficiently with little computational effort, whereby the determination of the most probable driving path is not based on a computer-aided trajectory but on learned values ​​for individual possible driving routes, which reflect the actual behavior of drivers of the motor vehicle.

[0054] A further embodiment of the invention also provides a motor vehicle which has such a control device.

[0055] Such a motor vehicle has the advantage of having a control unit configured to execute a method in which, in addition to digital map material, the current position of the motor vehicle, and the direction of travel of the motor vehicle, vehicle-specific data are also taken into account when determining the most probable route. For example, it may be advantageous and preferable to avoid routes with steep gradients under certain weather conditions or load conditions and instead select alternative routes. According to the invention, corresponding behavior learned from previous journeys with the same motor vehicle can be taken into account when determining the most probable route.Overall, an improved method for determining a most probable travel path of a motor vehicle by the control unit is thus carried out and, in particular, a method with which the accuracy in determining the most probable travel path can be increased.

[0056] Because the determination of the most probable driving path is based on artificial intelligence, the determination of the most probable driving path can also be carried out quickly and efficiently with little computational effort, whereby the determination of the most probable driving path is not based on a computer-aided trajectory but on learned values ​​for individual possible driving routes, which reflect the actual behavior of drivers of the motor vehicle.

[0057] The improved or more accurate most probable driving path can then be used or processed by driver assistance functions or driver assistance systems and other electronic components of the motor vehicle in order to further increase the safety of driving the motor vehicle.

[0058] A further embodiment of the invention also provides a computer program product comprising instructions executable by a computer to carry out a method described above.

[0059] Such a computer program product has the advantage that it is implemented to execute a method in which, in addition to digital map material, the current position of the motor vehicle, and the direction of travel of the motor vehicle, vehicle-specific data are also taken into account when determining the most probable route. For example, it may be advantageous and preferable to avoid routes with steep gradients under certain weather conditions or load conditions and instead select alternative routes. According to the invention, corresponding behavior learned from previous journeys with the same motor vehicle can be taken into account when determining the most probable route.Overall, an improved method for determining a most probable travel path of a motor vehicle is thus implemented and, in particular, a method with which the accuracy in determining the most probable travel path can be increased.

[0060] Because the determination of the most probable route is based on artificial intelligence, the most probable route can also be determined quickly and efficiently with little computational effort, whereby the determination of the most probable route is not based on a computer-aided trajectory but on learned values ​​for individual possible routes, which reflect the actual behavior of drivers of the commercial vehicle.

[0061] In summary, the present invention provides an improved method for determining a most probable travel path of a motor vehicle and, in particular, a method for determining a most probable travel path of a motor vehicle using artificial intelligence.

[0062] By using an appropriately trained artificial intelligence module, even large amounts of data can be processed in a simple manner and with little computational effort, so that the process can also be carried out by conventional computing units or data processing systems.

[0063] In particular, the method can also be designed in such a way that the method can be carried out entirely on a control unit of a motor vehicle without data having to be sent externally for processing.

[0064] The invention will now be explained in more detail with reference to the attached figures. Fig. 1 shows a flowchart of a method for determining a most probable travel path of a motor vehicle according to embodiments of the invention; Fig. 2 shows a flowchart of a method for determining a most probable travel path of a motor vehicle according to an example; Fig. 3 illustrates a method for determining a most probable travel path of a motor vehicle according to an example; Fig. 4 shows a block diagram of a control unit configured to determine a most probable travel path of a motor vehicle according to embodiments of the invention.

[0065] Fig. 1 shows a flowchart of a method 1 for determining a most probable travel path of a motor vehicle according to embodiments of the invention.

[0066] How Fig. 1 shows, the method comprises a step 2 of detecting absolute position data of a motor vehicle by means of an absolute positioning system and a step 3 of mapping the absolute position data onto a road contained in digital map material based on a map matching algorithm.

[0067] To increase safety when driving a motor vehicle, driver assistance systems in motor vehicles, especially in autonomous vehicles, should consider not only information about the immediate surroundings of the vehicle, but also information about the further, likely upcoming route. For example, safety when driving an autonomous vehicle can be increased if a likely upcoming route is taken into account when controlling driver assistance functions of the autonomous vehicle, for example, when controlling a speed limit assistant, adaptive headlights, or the shifting behavior of an automatic transmission.This can significantly increase safety for the driver and other road users. It also results in energy savings, especially since the driver assistance system can always be adapted directly to the anticipated conditions.

[0068] Usually, a most probable path, also called Most Probable Path (MPP), is determined, in particular on the basis of digital map material, i.e. a digital road network, the current position of the motor vehicle and the direction of travel of the motor vehicle by a module, for example a horizon provider or a driver horizon system.

[0069] The method 1 according to the first embodiment further comprises a step 4 of capturing current motor vehicle-specific data, wherein in a subsequent step 5 information about the road contained in the digital map material and the current motor vehicle-specific data are provided to an artificial intelligence module, wherein the artificial intelligence module is trained to determine a most probable travel path by artificial intelligence based on a road onto which absolute position data of the motor vehicle are mapped and motor vehicle-specific data, and wherein in a subsequent step 6 a most probable travel path of the motor vehicle is determined by the artificial intelligence module based on the information about the road contained in the digital map material and the current motor vehicle-specific data.

[0070] Motor vehicle-specific data is understood to mean data which is specific or characteristic of the use, i.e. the users or drivers of the motor vehicle, and in particular data which characterises certain properties, such as seat occupancy, travel times and weight distribution, which may influence the actual route selection, and where this influence may have been recorded during previous journeys with the motor vehicle.

[0071] According to the embodiments of the Fig. 1 thus specifies a method 1 in which, in addition to digital map material, the current position of the motor vehicle and the direction of travel of the motor vehicle, vehicle-specific data are also taken into account when determining the most probable travel path. For example, it may be advantageous and preferred to avoid routes with steep gradients in certain weather conditions or load states and to select alternative routes instead, wherein, according to the invention, corresponding behavior learned from previous journeys with the same motor vehicle can be taken into account when determining the most probable travel path. Overall, an improved method 1 for determining a most probable travel path of a motor vehicle is thus specified and, in particular, a method with which the accuracy in determining the most probable travel path can be increased.

[0072] Because the determination of the most probable route is based on artificial intelligence, the most probable route can also be determined quickly and efficiently with little computational effort, whereby the determination of the most probable route is not based on a computer-aided trajectory but on learned values ​​for individual possible routes, which reflect the actual behavior of drivers of the commercial vehicle.

[0073] The absolute position data can be recorded continuously, for example every 10 seconds, and mapped onto a road contained in the digital map material, whereby the recorded absolute position data can form the basis for determining a most probable driving path.

[0074] The artificial intelligence module can, for example, be a module developed on an external server or host, or software installed on it, to which the absolute position data of the motor vehicle and / or information about the road contained in the digital map material, as well as current vehicle-specific data, are transmitted. Furthermore, the artificial intelligence module can also be implemented in a control unit integrated into the motor vehicle, provided the corresponding control unit has the necessary computing capacity.

[0075] According to the invention, it is also first checked whether sufficient vehicle-specific data is available for the artificial intelligence module to determine the most probable route based on the information about the road contained in the digital map material and the current vehicle-specific data. 7 If this is the case, the most probable route is determined accordingly by artificial intelligence based on the information about the road contained in the digital map material and the current vehicle-specific data. If this is not the case, however, fleet models, i.e., data stored by the manufacturer, are used to determine the most probable route, or this is determined based on estimated values.

[0076] How Fig. As further shown in Figure 1, the method 1 further comprises a step 9 of providing the determined most probable driving path to driver assistance systems and / or electronic components of the motor vehicle. This has the advantage that driver assistance systems, for example a speed limit assistant, a control system for adaptive headlights, or a control system for an automatic transmission, can directly process the more precisely determined probable driving path in order to control the corresponding functions of the motor vehicle accordingly, thereby increasing the safety and accuracy when controlling the corresponding functions of the motor vehicle. The determined most probable driving path can be transmitted to the driver assistance systems wirelessly or by wire, for example via a vehicle bus or a CAN bus.

[0077] According to the embodiments of the Fig. 1, the vehicle-specific data further includes time data, seat occupancy data, loading data, charge level data, fuel data, calendar data and / or weather data.

[0078] According to the embodiments of the Fig. 1, the artificial intelligence module is also trained to determine the most probable driving path based on a road onto which absolute position data of a motor vehicle are mapped, vehicle-specific data and user-specific data, whereby the method, as Fig. 1 shows, comprises a step 10 of determining a driver of the motor vehicle, wherein in a following step 11 user-specific data associated with the driver of the motor vehicle are determined, and wherein in a subsequent step 12 the most probable travel path for the motor vehicle is determined by the artificial intelligence module based on the information about the road contained in the digital map material, the current motor vehicle-specific data and the user-specific data.

[0079] While vehicle-specific data refers to information or data about specific data based on all journeys with the motor vehicle, i.e., regardless of which driver actually drove the motor vehicle or which user actually used the motor vehicle, user-specific data refers to data that was recorded separately for each individual driver or user of the motor vehicle. Thus, the method can also be further designed such that the artificial intelligence module is trained separately for each individual driver or user of the motor vehicle, whereby the determination of the most probable route can be adapted to the practices or habits of the individual driver or user, thus further increasing the accuracy of determining the most probable route.The user-specific data can be stored in individual driver profiles.

[0080] The identification of a driver or user of the motor vehicle can be carried out according to the embodiments of the Fig. 1 can be done in all common ways, for example by a user logging into the system before using an autonomous vehicle.

[0081] If the motor vehicle deviates from the determined most probable path during the journey, according to the embodiments of the Fig. 1 a new most probable travel path is determined based on absolute position data based on the current position of the motor vehicle.

[0082] Fig. 2 shows a flowchart of a method 20 for determining a most probable travel path of a motor vehicle according to an example.

[0083] How Fig. 2 shows, the method 20 again comprises capturing absolute position data of a motor vehicle 21, mapping the absolute position data onto a road contained in digital map material based 22, capturing current motor vehicle-specific data 23, providing information about the first road and the current motor vehicle-specific data to an artificial intelligence module 24, wherein the artificial intelligence module is trained to determine a most probable travel path based on a road onto which absolute position data of a motor vehicle is mapped and motor vehicle-specific data, and determining a most probable travel path of the motor vehicle by the artificial intelligence module based on the information about the first road and the current motor vehicle-specific data 25.

[0084] Following the example of Fig. 2, the artificial intelligence module is trained to determine the most probable driving path based on learned conditional probabilities, which are dependent on vehicle-specific data and which indicate how often users of the motor vehicle have followed a road in the past, given certain vehicle-specific data, starting from an intersection point contained in the digital map material, starting from which users of the motor vehicle can follow different roads.

[0085] The step 25 of determining the most probable travel path of the motor vehicle by the artificial intelligence module comprises providing corresponding learned conditional probabilities based on the current motor vehicle-specific data 26 and determining the most probable travel path based on the corresponding learned conditional probabilities 27.

[0086] The provision of corresponding learned conditional probabilities based on the current motor vehicle-specific data 26 comprises providing one or more correspondingly trained ML models, wherein according to the embodiment of the Fig. 2 one or more artificial (neural) networks are provided.

[0087] Such an artificial network is modeled on biological neural networks and allows an unknown system behavior to be learned from existing training data and then applied to unknown input variables. The neural network consists of layers with idealized neurons that are connected to each other in different ways according to a network topology. The first layer, also called the input layer, captures and transmits the input values, with the number of neurons in the input layer corresponding to the number of input signals to be processed. The last layer, also called the output layer, usually has the same number of neurons as the number of output values ​​to be provided.Between the input layer and the output layer there is also at least one intermediate layer, which is often also referred to as a hidden layer, whereby the number of intermediate layers and the number of neurons in these layers depends on the specific task that is to be solved by the artificial network.

[0088] An idealized neuron can be defined by its weighted connections, which serve as inputs, and a transfer function that describes how the excitations from the inputs are to be processed in the neuron. The transfer forms can be provided, for example, in the form of sigmoid functions. Furthermore, a constant value of the neuron can be used to adjust how the corresponding inputs are transferred to a desired reference value. Such constant values ​​represent an additional degree of freedom and have a positive influence on the ability to approximate system behavior.

[0089] As an alternative to neural networks, other ML algorithms, such as a K-Nearest Neighbor algorithm or a Support Vector Machine, can also be used to provide classified data corresponding to a learned behavior. Such a classification algorithm can be trained separately for each decision point or for a group of different starting points from which the vehicle has departed in the past.

[0090] Furthermore, the conditional probabilities can also be determined based on Markov chains, for example, first-order Markov chains, i.e., determined only by the currently traveled road segment, without being influenced by past road segments. However, the conditional probabilities can also be determined using n-order Markov chains, for example, where the n most recently traveled road segments are also taken into account.

[0091] In addition, the conditional probabilities can also be determined by clustering or cluster analysis, whereby individual clusters can, for example, correspond to different times of day, whereby a separate Markov chain can be determined for each cluster, and whereby a density-based algorithm for cluster analysis, for example DSBCAN, OPTICS, HDBSCAN, or a mean-shift clustering can be used.

[0092] Following the example of Fig. 2, roads that have a learned conditional probability that is smaller than a predefined threshold value or that are assigned a corresponding probability are not taken into account when providing the corresponding learned conditional probabilities 25. The threshold value can be preset by a manufacturer of the motor vehicle and be, for example, 0.05. Furthermore, the threshold value can also be selected and set by a user of the motor vehicle themselves.

[0093] The artificial intelligence module is also trained to continuously update the learned conditional probabilities, with recent vehicle journeys being given greater weight when updating the conditional probabilities than longer vehicle journeys. Recent vehicle journeys could, for example, be training journeys or vehicle journeys made in the last few months, for example, in the last six months.

[0094] The step of determining the most probable path based on the corresponding learned conditional probabilities 26 further comprises selecting a route that has the highest conditional probability, starting from a starting position in the digital map material corresponding to the current absolute position data of the motor vehicle, as part of the most probable path. Subsequently, at intersection points along the most probable path, the road leading from the corresponding intersection point with the highest conditional probability is selected as part of the most probable path. In particular, a greedy algorithm can be used to determine the probable path.

[0095] In addition, the artificial intelligence module is further configured to determine the most probable path based on the corresponding learned conditional probabilities, optionally also to determine all possible routes between a starting position in the digital map material corresponding to the current absolute position data of the motor vehicle and the corresponding target position for one or more target positions, and then to select the route between the starting position and a target position of the one or more possible target positions as the most probable path, which has the highest overall probability of all determined routes. This can be achieved in particular by using Dijkstra's algorithm, whereby negative logarithmic conditional transition probabilities can be used.

[0096] The one or more possible destination points are, in particular, one or more predetermined destination points. Furthermore, the one or more destination points can also be one or more destination points that can be reached after a certain time based on the digital map material, or one or more destination points that can be reached after traveling a certain distance based on the digital map material.

[0097] Fig. 3 illustrates a method 30 for determining a most probable travel path of a motor vehicle according to an example.

[0098] In particular, Fig. 3 This refers to several intersections connected by correspondingly directed roads, with the direction of the road being based on the direction in which users of the motor vehicle have traveled along the corresponding road in the past. Roads leading from an intersection thus indicate roads that users of the motor vehicle have traveled along in the past, starting from the corresponding intersection. Roads leading into an intersection, in turn, indicate roads that users have used in the past to reach the corresponding intersection.

[0099] An intersection point is a point in the digital map material at which two or more roads intersect, and in particular a point in the digital map material from which users of the motor vehicle can take or drive on different roads, i.e. follow them. In general, a distinction can be made between split points, from which users of the motor vehicle can follow different roads, merge points, at which several roads enter but only one road branches off in the direction of travel, i.e. from which users of the motor vehicle can only follow one road when continuing in the same direction, and split-merge points, i.e. combinations of split and merge points and in particular points at which several roads enter and several roads branch off in the direction of travel.According to the embodiment of the . Fig. 3 The distribution points also include starting points from which users of the motor vehicle have set off in different directions or directions of travel in the past.

[0100] In addition, current absolute position data are recorded and mapped onto digital data, whereby it is determined that the motor vehicle is currently located at the starting position designated by reference number 31.

[0101] Current vehicle-specific data is also recorded.

[0102] It can also be seen that a conditional probability is assigned to each individual road or route, whereby the conditional probabilities are dependent on the current vehicle-specific data, and whereby the conditional probabilities indicate how often users of the motor vehicle have followed a corresponding road in the past, given certain vehicle-specific data, starting from an intersection point contained in the digital map material, starting from which users of the motor vehicle can follow different roads.

[0103] Starting from the starting position 31, the most probable route expected to be travelled in the next 15 km is now to be determined, whereby the possible destination points marked with reference numbers 32, 33 and 34 can each be reached after a 15 km journey based on the road network contained in the digital map material.

[0104] The most probable route can now be determined by selecting a route which has the highest conditional probability starting from the starting position 31 as part of the most probable route, whereby at intersection points along the most probable route the road leading from the corresponding intersection point with the highest conditional probability is then selected as part of the most probable route.

[0105] How Fig. 3 shows, starting from the starting position 31, a route leading to the left would initially be selected as part of the most probable path, since this has the highest conditional probability of all routes leading from the starting position. Following this route now leads to an intersection point marked with reference numeral 35. Here, the road leading from the intersection point 35 would now be selected as part of the most probable path, which road has the highest conditional probability of all roads leading from the intersection point. According to the embodiment of the Fig. 3, a road leading upwards, designated by reference numeral 36, would be selected as part of the most probable path. This process is then repeated until one of the possible target positions 32, 33, 34, reachable after 15 km, is reached. As soon as one of the possible target positions 32, 33, 34, reachable after 15 km, is reached, the process terminates, and a path composed of the individually selected routes and roads is output as the most probable path.

[0106] It should be noted that if multiple roads leading from an intersection have the same conditional probability, certain rules are defined that can be used to select one of these roads as part of the most probable path. For example, it can be specified that in such a case, the road with the fewest curves and thus the longest straight line is selected as part of the most probable path.

[0107] Alternatively, the most probable route can also be determined by determining all possible routes between the starting position 31 and one of the possible destination positions 32, 33, 34 and then determining an overall probability for each of the determined routes based on the individual conditional probabilities assigned to the roads. The overall probability of a route represents the relative frequency with which users have followed this exact route in the past when the same vehicle-specific data were available. The route with the highest overall probability of all determined routes is then selected as the most probable route.

[0108] Fig. 4 shows a block diagram of a control unit 40 which is designed to determine a most probable travel path of a motor vehicle, according to embodiments of the invention.

[0109] How Fig. As shown in Figure 4, the control unit 40 has a receiver 41 configured to receive absolute position data of the motor vehicle acquired by an absolute positioning system and vehicle-specific data acquired by sensors or other control units. The receiver 41 may, in particular, be a transceiver.

[0110] In addition, the control unit 40 has an artificial intelligence module 42, which is trained to determine a most probable travel path based on a current position of the motor vehicle and motor vehicle-specific data.

[0111] In particular, the artificial intelligence module 42 is designed to determine a most probable travel path based on the data received by the receiver 41.

[0112] According to the embodiments of the Fig. 4, the control unit also has a transmitter 43, which transmits the most probable driving path determined by the artificial intelligence module to driver assistance systems and / or electronic components of the motor vehicle, so that they can subsequently use and process the determined most probable driving path accordingly.

[0113] How Fig. 4 shows, the control unit 40 further comprises a memory 44 in which digital map material is stored.

[0114] The artificial intelligence module 42 further comprises a determination unit 45, which is configured to determine a most probable travel path for the motor vehicle based on the data received by the receiver 41 and the data stored in the memory 44. The determination unit 45 can be, for example, a computing unit on which corresponding code is stored, such as a server or a control unit of the motor vehicle, i.e., a computing unit integrated into the motor vehicle itself.

Claims

[1] Method for determining a most probable travel path of a motor vehicle, the method (1, 20, 30) comprising the following steps: - detecting absolute position data of a motor vehicle by means of an absolute positioning system (2,21); - mapping the absolute position data onto a road contained in digital map material (3,22); - Recording of current vehicle-specific data (4.23); - Providing information about the road contained in the digital map material and the current vehicle-specific data to an artificial intelligence module, wherein the artificial intelligence module is trained to determine a most probable driving path by artificial intelligence based on a road on which the motor vehicle is currently located and vehicle-specific data (5,24); and - Determining the most probable travel path of the motor vehicle by the artificial intelligence module based on the information about the road contained in the digital map material and the current vehicle-specific data (6, 25), wherein it is first checked whether sufficient vehicle-specific data is available so that the artificial intelligence module can determine the most probable travel path based on the information about the road contained in the digital map material and the current vehicle-specific data, wherein if sufficient vehicle-specific data is available, the most probable travel path is determined by the artificial intelligence module based on the information about the road contained in the digital map material and the current vehicle-specific data, and wherein if insufficient vehicle-specific data is available,Fleet values ​​or estimates are used to determine the most likely driving path. [2] The method according to claim 1, wherein the method (1) further comprises the following step: - Providing the determined most probable driving path to driver assistance systems and / or electronic components of the motor vehicle (9). [3] Method according to claim 1 or 2, wherein the motor vehicle-specific data comprise time data, seat occupancy data, loading data, charge level data, fuel data, calendar data and / or weather data. [4] Method according to one of claims 1 to 3, wherein the artificial intelligence module is trained to determine the most probable travel path based on learned conditional probabilities which are dependent on motor vehicle-specific data and which indicate how often users of the motor vehicle have followed a road in the past, given certain motor vehicle-specific data, starting from an intersection point contained in the digital map material, starting from which users of the motor vehicle can follow different roads, and wherein the step of determining the most probable travel path of the motor vehicle by the artificial intelligence module (25) comprises the following steps: - providing corresponding learned conditional probabilities based on the current vehicle-specific data (26); and - Determine the most probable path based on the corresponding learned conditional probabilities (27). [5] The method of claim 4, wherein the step of providing corresponding learned conditional probabilities based on the current vehicle-specific data (26) comprises providing one or more correspondingly trained ML models. [6] Method according to claim 4 or 5, wherein, in providing the corresponding learned conditional probabilities (26), roads which have a learned conditional probability which is smaller than a predetermined threshold value are not taken into account. [7] Method according to one of claims 4 to 6, wherein the artificial intelligence module is designed to continuously update the learned conditional probabilities, and wherein more recent journeys with the motor vehicle are given greater weight when updating the conditional probabilities than more recent journeys with the motor vehicle. [8] Method according to one of claims 4 to 7, wherein the step of determining the most probable travel path based on the corresponding learned conditional probabilities (27) further comprises the following step: - Selecting a route which, starting from a starting position in the digital map material corresponding to the current absolute position data of the motor vehicle, has the highest conditional probability as part of the most probable travel path, wherein subsequently, at intersection points along the most probable travel path, the road leading from the corresponding intersection point with the greatest conditional probability based on the current vehicle-specific data is selected as part of the most probable travel path. [9] A method according to any one of claims 4 to 7, wherein the step of determining the most probable travel path based on the corresponding learned conditional probabilities further comprises the following steps: - For one or more possible destination positions, determining all possible routes between a starting position in the digital map material corresponding to the current absolute position data of the motor vehicle and the corresponding destination position; - Selecting the route between the starting position and a destination position of the one or more possible destination positions as the most probable travel path, which overall has the greatest overall probability of all determined routes. [10] Method according to claim 9, wherein the one or more possible target points are one or more predetermined target points, one or more target points reachable based on the digital map material after a certain time, or one or more target points reachable based on the digital map material after traveling a certain distance. [11] Method according to one of claims 1 to 10, wherein the artificial intelligence module is trained to determine the most probable travel path based on a road on which the motor vehicle is currently located, the current vehicle-specific data and user-specific data, and wherein the method further comprises the following steps: - identifying a driver of the motor vehicle (10); - determining user-specific data (11) associated with the driver of the motor vehicle; and - Determination of a most likely path for the motor vehicle by the artificial intelligence module based on the information about the road contained in the digital map material, the current vehicle-specific data and the user-specific data (12). [12] Control unit for determining a most probable travel path of a motor vehicle, wherein the control unit (40) is designed to carry out a method according to one of claims 1 to 11. [13] Motor vehicle having a control device according to claim 12. [14] A computer program product comprising instructions executable by a computer to perform a method according to any one of claims 1 to 11.

Images