Security control of a vehicle combination

Abstract

The invention relates to safety control of a vehicle combination. Computer systems and methods are provided for training a machine learning model to determine a safe torque distribution for a vehicle combination and for determining a safe torque distribution. A system is configured to perform a plurality of simulations on motion of a vehicle combination based on a plurality of torque distributions and operating points, classify each simulation as safe or unsafe, and train a machine learning model using the classified simulations. A system is configured to receive torque distributions and operating points of a vehicle combination and to output one or more safe torque distributions using a trained machine learning model.

Description

Technical Field

[0001] The present disclosure generally relates to vehicle control. In particular aspects, the present disclosure relates to safety control of vehicle combinations. The present disclosure may be applicable to heavy vehicles such as trucks, buses, and construction equipment, as well as other vehicle types. Although the present disclosure may be described with respect to specific vehicles, the present disclosure is not limited to any particular vehicle. Background Art

[0002] Vehicle combinations including a tractor unit and one or more trailer units are often over-actuated, meaning that there are more actuators than there are controlled motions. This means that torque (i.e., braking or accelerating torque) can be distributed in different ways between the tractor and the trailer. This can be used to optimize torque distribution in some way, for example by maximizing energy efficiency. However, in order to do this, it must first be known which distributions are safe. There is currently no fast and efficient method for performing such an online assessment for different vehicle operating points.

[0003] Accordingly, it is desirable to provide systems, methods, and other means that attempt to solve or at least mitigate one or more of these problems. Summary of the Invention

[0004] The present disclosure provides systems, methods, and other means for determining safe torque distributions for vehicle combinations. Specifically, a large number of simulations of the motion of the vehicle combination are performed. The simulations are performed for different combinations of torque distributions and vehicle states, and each simulation is classified as safe or unsafe. The results of the simulations can then be used to train a machine learning model. The trained model can then be used to output which torque distributions are safe and which are unsafe for a given operating point of the vehicle combination.

[0005] According to a first aspect of the present disclosure, there is provided a computer system for training a machine learning model to determine safe torque distributions for a vehicle combination including a tractor unit and at least one trailer unit, the computer system including processing circuitry configured to perform a plurality of simulations of the motion of the vehicle combination based on a plurality of torque distributions and a plurality of operating points of the vehicle combination; classify each simulation as safe or unsafe; and use the plurality of simulations to train a machine learning model to determine safe torque distributions for the vehicle combination based on the operating points of the vehicle combination.

[0006] A first aspect of the present disclosure may seek to provide a machine learning model that can quickly and reliably output safe torque distributions based on the state of a vehicle combination. This can be implemented in a vehicle combination such that control can be performed in a safe manner based on the parameters of the vehicle combination and / or its environment.

[0007] Optionally, in some examples, including in at least one preferred example, each torque distribution includes a corresponding torque distribution for each unit of the vehicle combination, and the sum of the corresponding torque distributions satisfies the total torque distribution for the vehicle combination. Technical benefits can include that a machine learning model can provide a safe torque distribution for vehicle combinations with different unit configurations, and the selected distribution can be implemented at the unit level.

[0008] Optionally, in some examples, including in at least one preferred example, each operating point is defined based on one or more of the following: the speed v of the vehicle combination, the lateral acceleration a of the vehicle combination y , one or more axle loads F of one or more units of the vehicle combination z , the yaw rate of one or more units of the vehicle combination , the slip angle α of one or more units of the vehicle combination, the rate of change of the slip angle of one or more units of the vehicle combination , the steer angle δ of the drive wheels of one or more units of the vehicle combination, the surface friction coefficient μ, the radius of curvature r, and / or the path gradient. Technical benefits can include that a machine learning model can provide a safe torque distribution for the operating points of vehicle combinations defined in different ways, enabling the model to be used in a variety of different scenarios.

[0009] Optionally, in some examples, including in at least one preferred example, the processing circuit is configured to classify the simulation as unsafe if the motion simulated by a certain simulation includes instability of the vehicle combination. Technical benefits can include that a machine learning model can provide a torque distribution that avoids unsafe and / or unstable operating modes of the vehicle combination, thus ensuring safe operation.

[0010] According to a second aspect of the present disclosure, there is provided a computer system for determining a safe torque distribution for a vehicle combination including a tractor unit and at least one trailer unit, the computer system including a processing circuit configured to receive a plurality of torque distributions that satisfy a torque request for the vehicle combination; receive an operating point of the vehicle combination; and use a machine learning model trained according to the first aspect to output one or more safe torque distributions for the vehicle combination based on the received torque distributions and operating point.

[0011] The second aspect of the present disclosure may seek to output a safe torque distribution based on the state of the vehicle combination in a fast and reliable manner. This can be implemented online such that the vehicle combination can be controlled in a safe manner according to the parameters of the vehicle combination and / or its environment to satisfy the torque request.

[0012] Optionally, in some examples, including in at least one preferred example, the received operating point is the current operating point of the vehicle combination. Technical benefits may include that a safe torque distribution can be output based on current operating parameters and driving conditions, thereby ensuring that the vehicle combination is controlled in an adaptive manner.

[0013] Optionally, in some examples, including in at least one preferred example, the processing circuit is configured to generate an alert or modify the torque request if the one or more output safe torque distributions do not meet one or more second criteria. Technical benefits may include that the operator of the vehicle can be informed that there is no available optimal torque distribution, or the vehicle can be controlled in a different way to meet one or more second criteria.

[0014] Optionally, in some examples, including in at least one preferred example, the machine learning model is configured to output a plurality of safe torque distributions for the vehicle combination, and the processing circuit is configured to determine a safe torque distribution from the plurality of safe torque distributions based on one or more first criteria. Technical benefits may include that a range of possible safe torque distributions can be provided, from which a safe torque distribution can be selected based on given parameters, such as selecting a safe torque distribution that provides good energy efficiency.

[0015] Optionally, in some examples, including in at least one preferred example, the processing circuit is configured to apply a safety margin to the one or more output safe torque distributions. Technical benefits may include that the operation of the vehicle combination can be comfortably maintained within a safe operating range and not approach the limits of unsafe or unstable operation.

[0016] Optionally, in some examples, including in at least one preferred example, the processing circuit is configured to cause the output safe torque distribution to be implemented as the torque distribution for the vehicle combination. Technical benefits may include the ability to safely control the vehicle combination in a fast and reliable manner.

[0017] According to a third aspect of the present disclosure, a vehicle is provided, which includes the computer system of the first aspect and / or the second aspect. The third aspect of the present disclosure may seek to provide a vehicle capable of determining a safe torque distribution based on a given operating point.

[0018] According to a fourth aspect of the present disclosure, there is provided a computer-implemented method for training a machine learning model to determine a safe torque distribution for a vehicle combination including a tractor unit and at least one trailer unit, the method comprising: performing, by a processing circuit of a computer system, a plurality of simulations of the movement of the vehicle combination based on a plurality of torque distributions and a plurality of operating points of the vehicle combination; classifying, by the processing circuit, each simulation as safe or unsafe; and training, by the processing circuit, a machine learning model using the plurality of simulations to determine a safe torque distribution for the vehicle combination based on the operating points of the vehicle combination.

[0019] The fourth aspect of the present disclosure may seek to provide a machine learning model that can quickly and reliably output a safe torque distribution based on the state of a vehicle combination. This can be implemented in the vehicle combination such that it can be controlled in a safe manner according to the parameters of the vehicle combination and / or its environment.

[0020] According to a fifth aspect of the present disclosure, there is provided a computer-implemented method for determining a safe torque distribution for a vehicle combination including a tractor unit and at least one trailer unit, the method comprising: receiving, by a processing circuit of a computer system, a plurality of torque distributions that satisfy a torque request for the vehicle combination; receiving, by the processing circuit, an operating point of the vehicle combination; and outputting, by the processing circuit and using a machine learning model trained according to the fourth aspect, one or more safe torque distributions for the vehicle combination based on the received torque distributions and operating points.

[0021] The fifth aspect of the present disclosure may seek to output a safe torque distribution based on the state of a vehicle combination in a quick and reliable manner. This can be implemented online such that the vehicle combination can be controlled in a safe manner according to the parameters of the vehicle combination and / or its environment.

[0022] According to a sixth aspect of the present disclosure, there is provided a computer program product including program code for performing the computer-implemented methods of the fourth aspect and / or the fifth aspect when executed by a processing circuit. The sixth aspect of the present disclosure may seek to enable convenient configuration of new vehicles and / or traditional vehicles through software installation / updating to determine a safe torque distribution based on a given operating point in a quick and reliable manner.

[0023] According to a seventh aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium including instructions that, when executed by a processing circuit, cause the processing circuit to perform the computer-implemented methods of the fourth aspect and / or the fifth aspect. The seventh aspect of the present disclosure may seek to enable convenient configuration of new vehicles and / or traditional vehicles through software installation / updating to determine a safe torque distribution based on a given operating point in a quick and reliable manner.

[0024] Those of ordinary skill in the art will appreciate that the disclosed aspects, examples (including any preferred examples), and / or the appended claims may be appropriately combined with each other. Additional features and advantages are disclosed in the following description, claims, and drawings, and will be partly apparent to those of skill in the art or will be recognized by practicing the present disclosure as described herein.

[0025] Also disclosed herein are computer systems, control units, code modules, computer-implemented methods, computer-readable media, and computer program products associated with the technical benefits discussed above. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Examples will be described in more detail below with reference to the drawings.

[0027] Figure 1 A top view of a vehicle combination according to an example is schematically shown.

[0028] Figure 2 is a flowchart of a computer-implemented method according to an example.

[0029] Figure 3 is a flowchart of a computer-implemented method according to an example.

[0030] Figure 4 A torque distribution diagram of a machine learning model output is shown according to an example.

[0031] Figure 5 is a schematic diagram of a computer environment according to an example.

[0032] Figure 6 is a schematic diagram of a computer system for implementing the examples disclosed herein.

[0033] Throughout the specification, the same reference numerals refer to the same elements. DETAILED DESCRIPTION

[0034] The detailed description set forth below provides information and examples of the disclosed technology in sufficient detail for those of skill in the art to practice the present disclosure.

[0035] A vehicle combination including a tractor and one or more trailers is often over-actuated, which means that there are more actuators than controlled motions. This means that torque (i.e., braking or accelerating torque) can be distributed between the tractor and the trailer in different ways. This can be used to optimize torque distribution in some way, for example, by maximizing energy efficiency. However, in order to do this, it must first be known which distributions are safe. There is currently no fast and efficient method for performing this online evaluation for different vehicle operating points.

[0036] To solve this problem, systems, methods, and other means for determining safe torque distribution for a vehicle combination are provided. Specifically, a large number of simulations of the motion of the vehicle combination are performed. Simulations are performed for different combinations of torque distribution and vehicle states, and each simulation is classified as safe or unsafe. The results of the simulations can then be used to train a machine learning model. The trained model can be used online and, based on the current state of the vehicle, output which torque distributions are safe and which are unsafe.

[0037] Training and implementing a machine learning model in this way can provide a vehicle combination with a fast and reliable way to output safe torque distributions based on the state of the vehicle combination. This can be implemented online such that the vehicle combination is controlled in a safe manner according to current operating parameters and driving conditions. The implementation also allows for a range of possible safe torque distributions to be provided from which a safe torque distribution can be selected based on given parameters, such as selecting a safe torque distribution that provides good energy efficiency.

[0038] Figure 1 A top view of an exemplary vehicle combination 100 of the type considered in the present disclosure is schematically shown. The vehicle combination 100 includes a plurality of units 110 (including a tractor unit 110-1 and a trailer unit 110-2). Although a single trailer unit 110-2 is shown, it should be understood that the vehicle combination 100 can include additional trailer units. This results in different types and names of vehicle combinations. Figure 1 The illustrated vehicle combination 100 is an example of a truck, however, the systems and methods disclosed herein can be used with any suitable form of vehicle combination 100, such as trucks, buses, etc.

[0039] The tractor unit (such as the tractor unit 110-1) is typically the foremost unit in the vehicle combination 100 and can include a cab for the driver (including steering controls, dashboard displays, etc.). Generally, the tractor unit 110-1 is used to provide propulsion power for the vehicle combination 100. In Figure 1 the example of, the tractor unit 110-1 can also be used to store the cargo being transported by the vehicle combination 100.

[0040] The trailer unit (such as the trailer unit 110-2) is typically used to store the cargo being transported by the vehicle combination 100. The trailer unit can be a truck, a trailer, a dolly, etc. The trailer unit can also provide propulsion force for the vehicle combination 100. In a vehicle combination such as Figure 1 that shown, vehicle motion management can be performed at the unit level to receive requests from a manual or virtual driver to coordinate propulsion, braking, and steering.

[0041] Each unit 110 includes a plurality of axles 120, each axle having a plurality of wheels 130. It should be understood that any suitable number of axles 120 can be provided on the respective unit 110. A towing unit 110-2 without a front axle is called a semi-trailer. It should also be understood that any number of tractor axles and / or trailer axles can be drive axles, including zero (i.e., one of the units 110 can include at least one drive axle while the other does not).

[0042] Figure 1 The relevant parameters of the vehicle combination 100 are shown. Examples of the parameters of the vehicle combination 100 can include, for example, the speed v (including the longitudinal speed v x and the lateral speed v y ), the longitudinal acceleration a x and the lateral acceleration a y . These parameters can also be defined according to the unit. For example, each unit 110 has its own speed and acceleration. For example, each unit 110 has a corresponding speed v i , which can be regarded as two components: the longitudinal speed v xi and the lateral speed v yi , where i is the unit index. Similarly, the acceleration of each unit can be regarded as two components, namely the longitudinal acceleration a xi and the lateral acceleration a yi . Each unit 110 also has a mass m, and the mass can affect the load F z on the axles 120 of the unit 110. Each unit 110 has a corresponding yaw angle ψ, which is the angle between the environmental reference frame and the vehicle reference frame.

[0043] Figure 1 The relevant parameters of the wheels 130 of the vehicle combination 100 are also shown. For example, each wheel 130 can have a running wheel angle δ, which is the angle between the longitudinal axis of the unit 110 of the wheel 130 and the direction the wheel 130 faces. The running wheel angle δ can also be called the steering angle δ. When the vehicle combination 100 is in motion, the wheels 130 experience slip. The slip can be expressed as the slip on the entire vehicle 100, the slip on a given unit 110, the slip on a given axle 120, or the slip on a given wheel 130. These parameters are known in the art and are not discussed in detail here. However, it is worth noting that for a given unit 110, the slip angle α, that is, the angle between the longitudinal dimension of the unit 110 (the direction the unit 110 points) and the traveling direction of the unit 110, can be defined as: The slip angle can also be called the sideslip angle.

[0044] Figure 1Also shown are relevant parameters of the route traveled by the vehicle combination 100. At Figure 1 In the example of, it is shown that the vehicle combination 100 travels on a curved path with a radius of curvature r. The path may have a slope in the longitudinal direction and / or the lateral direction. The surface on which the vehicle combination 100 travels has a surface friction coefficient μ. Although Figure 1 a curved path is shown in, it should be understood that the methods disclosed herein are applicable to any suitable operation of the vehicle combination 100, including straight driving.

[0045] Figure 1 The various parameters shown in can be used to define the operating point 140 of the vehicle combination 100. For example, at a certain point in the journey, the vehicle combination 100 may have a certain speed v (including the longitudinal speed v x and the lateral speed v y ) and a lateral acceleration a y . At this operating point 140, each unit can have a certain yaw rate and axle load F zik on a given axle, where k is the axle index. At the operating point 140, each unit 110 can have a slip angle α and a slip angle change rate , and the wheels 130 of each unit 110 can each have a road wheel angle δ. The operating point 140 can also include parameters of the route and / or environment on which the vehicle combination 100 travels, such as the surface friction coefficient μ, the radius of curvature r, and / or the path slope.

[0046] At any given operating point 140, a torque request for the vehicle combination 100 can be received, for example, from a manual or virtual driver, and the torque request is related to a certain desired movement of the vehicle combination 100. The torque request can be at the global (vehicle) level. The global torque request can be divided into corresponding torque requests for each unit 110. The sum of the respective corresponding torque requests satisfies the total global torque request for the vehicle combination 100. For example, the torque request for the vehicle combination 100 can be regarded as the total torque request T (covering braking and / or propulsion), including the torque request T 请求、牵引车 for the tractor unit 110-1 and the torque request T 请求、挂车 for the trailer unit 110-2. The torque request is converted into a torque distribution for the vehicle combination 100, including actual signals provided to the motion support devices of the vehicle combination 100, and the motion support devices are, for example, one or more motors of the propulsion or braking system of the vehicle combination 100 configured to drive one or more axles 120 or individual wheels 130 of the vehicle combination 100 (e.g., providing torque and / or steering to them). The torque distribution can also be at the vehicle or unit level. This torque distribution can be performed by the Vehicle Motion Manager (VMM) of the vehicle combination 100.

[0047] The ability to differentially allocate torque (i.e., braking or accelerating torque) between the tractor unit 110-1 and the trailer unit 110-2 can be used to optimize torque allocation in some manner, such as by maximizing energy efficiency. However, in order to do this, it must first be known which allocations are safe. There is currently no quick and effective way to perform this online evaluation for the different operating points 140 of the vehicle combination 100. To address this problem, the present disclosure provides systems, methods, and other means for training and implementing a machine learning model to determine safe torque allocations for a vehicle combination 100.

[0048] Figure 2 is a flow chart of an exemplary computer-implemented method 200 for training a machine learning model to determine safe torque allocations for a vehicle combination 100. Method 200 may be executed by a computer system associated with a vehicle combination (such as Figure 1 the vehicle combination 100 shown in the example of). The computer system may be a control system that is implemented on the vehicle combination, implemented remote from the vehicle combination, or a combination of both, as will be discussed in connection with Figure 5 discussed. Method 200 may be executed by the processing circuitry of the computer system.

[0049] At 210, a plurality of simulations of the motion of the vehicle combination 100 are performed. Each simulation is based on the torque allocation and the operating point 140 of the vehicle combination 100. That is, each simulation models the motion of the vehicle combination 100 in response to receiving a particular torque allocation at the operating point 140 defined by one or more parameters of the vehicle combination 100 and / or its environment. The simulations may be performed using a suitable vehicle model. The suitable vehicle model is preferably a vehicle model that captures vehicle dynamics (such as longitudinal, lateral, and vertical dynamics (depending on the instability being considered)). The model preferably includes a sufficient number of parameters and factors to provide an accurate representation of the behavior of the vehicle, particularly under conditions associated with instability. Examples of such models include low-fidelity models (such as single-track or double-track models) and high-fidelity models (e.g., Volvo Transportation models). The model preferably includes a sufficiently accurate tire model, such as a Pajecka tire model, a brush tire model, etc.

[0050] As discussed above, each torque distribution can be a global torque distribution for the vehicle combination 100 or can include corresponding torque distributions for each unit 110 of the vehicle combination 100, where the sum of the corresponding torque distributions meets the total torque distribution for the vehicle combination 100. The torque distributions used in the simulations can be related to braking and / or acceleration of the vehicle combination 100. Also as discussed above, each operating point 140 is defined based on one or more of the following: the speed v of the vehicle combination 100, the lateral acceleration a y of the vehicle combination 100, the axle loads of one or more of the one or more units 110 of the vehicle combination 100, the yaw rate of one or more of the one or more units 110 of the vehicle combination 100 the slip angle α of one or more of the one or more units 110 of the vehicle combination 100, the rate of change of the slip angle of one or more of the one or more units 110 of the vehicle combination 100 the steering wheel angle δ of one or more of the one or more units 110 of the vehicle combination 100, the surface friction coefficient μ, the curve radius r, and the path gradient. In some examples, the slip angle α of one or more axles 120 or one or more wheels 130 can be used additionally or alternatively.

[0051] At 220, each simulation is classified as safe or unsafe. For example, if the simulated motion of the vehicle combination 100 includes instability of the vehicle combination 100, the simulation can be classified as unsafe. In some examples, if any state of the vehicle combination 100 during the simulation is considered unsafe, each time instance of the simulation is classified as unsafe. Multiple different types of instability can be identified. In some examples, the instability is a jackknife, where the tractor unit 110-1 starts to slide laterally and pushes the tractor unit 110-2, causing the tractor unit 110-1 to rotate about the vertical axis until it hits the trailer unit 110-2. In some examples, the instability is a trailer swing, where the wheels of the trailer unit 110-2 slip while the wheels of the tractor unit 110-1 do not slip, resulting in the trailer unit 110-2 swinging about the vertical axis. In some examples, the instability is a combined jackknife and trailer swing, where both a jackknife and a trailer swing occur simultaneously. In some examples, the instability is a rollover, where one or more of the units 110 of the vehicle combination 100 tip over. In some examples, the instability is a deviation, where one or more of the units 110 of the vehicle combination 100 travel outside of their intended path. In some examples, it can be based on the rate of change of the slip angle and / or the yaw rate to identify instability. This captures whether the unit experiences a rapid change in yaw angle ψ over a short period of time, which indicates unstable behavior. In some examples, inability can include backward amplification, which is a measure of signal oscillation during a sudden turn (such as the yaw rate of one or more units 110 ).

[0052] If the simulation does not result in instability of the vehicle combination 100, it is classified as safe. In these examples, the simulated motion of the vehicle combination 100 can be as expected or anticipated by the corresponding motion request.

[0053] In this way, each pair of torque distribution and operating point 140 is classified based on the resulting simulated motion. Thus, a number of data points are defined, which include pairs of torque distribution and operating point and the associated safety or stability classification. Then these data points can be used as training data for a machine learning model. It should be understood that a large number of data points may be required to train a machine learning model to an adequately accurate and reliable level.

[0054] At 230, a machine learning model is trained using the data points to determine a safe torque distribution for the vehicle combination 100. Specifically, at least some of the data points are used to train the machine learning model such that the model adapts its parameters (weights and biases) in order to minimize the error between the predicted value and the actual value. In some embodiments, when training the machine learning model, at least some of the data points are used to monitor the performance of the model in terms of hyperparameters (e.g., learning rate, regularization strength). By providing an independent data set for tuning, this helps prevent overfitting. Additionally, validation data can be used to compare different models and select the best model.

[0055] Any suitable machine learning model can be used. Some suitable models include decision tree models, random forest models, and neural networks. In some examples, a neural network with one, two, or three hidden layers can be used. In one example, a feedforward neural network with three hidden layers is used. Neural networks can capture complex patterns and have the potential to improve understanding through hyperparameter tuning.

[0056] Due to the classification of various simulations, the trained machine learning model is capable of providing a safe torque distribution that avoids unsafe and / or unstable operating modes of the vehicle combination. Once the machine learning model has been trained, it can be used to determine a safe torque distribution based on the input operating point 140 of the vehicle combination 100. This will be explained in conjunction with Figure 3 for interpretation.

[0057] Figure 3 is a flow chart of an exemplary computer-implemented method 300 for determining a safe torque distribution for a vehicle combination 100 according to an example. Method 300 can be performed by a computer associated with a vehicle combination (e.gFigure 1 executed by a computer system associated with the vehicle combination 100 as shown in the example of. The computer system can be a control system that is implemented on the vehicle combination, implemented remotely from the vehicle combination, or a combination of both, as will be described in connection with Figure 5 discussed. Method 300 can be executed by the processing circuitry of the computer system.

[0058] At 310, a torque request for the vehicle combination 100 is received. As discussed above, for example, a torque request for the vehicle combination 100 can be received from a manual or virtual driver. The torque request can be a global torque request that can be divided into corresponding torque requests for each unit 110. The sum of the respective corresponding torque requests satisfies the total global torque request for the vehicle combination 100.

[0059] At 320, a plurality of different torque distributions that satisfy the torque request are determined. This can be performed by the VMM of the vehicle combination 100, as is well known in the art. The torque distribution can be a global torque distribution for the vehicle combination 100, or can include corresponding torque distributions for each unit 110 of the vehicle combination 100. In some examples, it can be ensured that the torque distribution has sufficient resolution. This can be related to the number of torque distributions determined for the torque request, and / or their distribution between a maximum and a minimum value.

[0060] At 330, the machine learning model receives the torque distribution determined at 320 based on the torque request received at 310. The machine learning model can be trained according to method 200. The machine learning model can be implemented as a software module in the computer system, such as the control system of the vehicle combination 100, as will be described in connection with Figure 5 discussed.

[0061] At 340, the machine learning model receives an input in the form of an operating point 140 of the vehicle combination 100. As discussed above, the operating point 140 can be defined based on a plurality of different parameters of the vehicle combination 100 and / or its environment.

[0062] At 350, the machine learning model outputs one or more safe torque distributions for the vehicle combination 100. Specifically, due to the training performed in method 200, the machine learning model is able to output one or more safe torque distributions for the input operating point 140 from the number of torque distributions received at 330. Depending on how the machine learning model is trained and how the input torque distributions are determined, the output safe torque distribution can be a global torque distribution for the vehicle combination 100 or can include corresponding torque distributions for each unit 110 of the vehicle combination 100. Thus, the machine learning model is able to provide safe torque distributions for vehicle combinations with different unit configurations and can implement the selected distribution at the global or unit level.

[0063] The input operating point 140 can be the current operating point 140 of the vehicle combination 100. In this way, multiple safe torque distributions can be provided online (i.e., when the vehicle combination 100 is in motion and a torque request is received). This ensures that the vehicle combination is controlled in an adaptive manner based on the current operating parameters and driving conditions.

[0064] Figure 4 An example graph 400 showing a torque distribution 402 for a vehicle combination 100 having a tractor unit 110-1 and a trailer unit 110-2 is shown. Specifically, the plotted torque distribution 402 is the braking distribution input to the machine learning model at 330. In this example, each torque distribution 402 represents an allocation T for the tractor unit 110-1 牵引车 and a T for the trailer unit 110-2 挂车 . In some examples, the machine learning model is configured to output multiple safe torque distributions for the vehicle combination 100. In Figure 4 the example, the safe torque distribution 404 represented by a cross is located between two limits 406a, 406b, outside of which are located multiple unsafe torque distributions 408a, 408b represented by circles. In this example, the unsafe torque distribution 408a corresponds to a fold, while the unsafe torque distribution 408b corresponds to a trailer swing.

[0065] When the machine learning model outputs multiple safe torque distributions, at 360, a single one 404a, 404b of the multiple safe torque distributions 404 is determined based on one or more first criteria. For example, a criterion related to the maximum braking level 410 of the tractor unit 110-1 can be set, and any one of the multiple safe torque distributions 404 involving braking above this level can be ignored. In Figure 4In the example of, this results in the selection of torque distribution 404a. Another example of the first criterion includes the minimum energy efficiency of the vehicle combination 100. The selection can be made using one or more optimization functions. In one example, it may be desirable to select a safe torque distribution 404 that maximizes the braking of the tractor unit 110-1.

[0066] In some examples, at 370, a safety margin can be applied to one or more output safe torque distributions. Specifically, the safety margin can be applied at the limit between the safe torque distribution and the unsafe torque distribution. In some examples, the safety margin can be an absolute value. In some examples, the safety margin can be a function of a parameter. In Figure 4 the example of, a safety margin 412 of 10% of the total braking distribution T is applied, resulting in a maximum braking torque distribution 404b. Thus, the braking torque is reduced, making the resulting torque distribution and the resulting vehicle behavior safer. In other examples, such as in the propulsion example, the output safe torque distribution can be made safer by reducing one or more of the speed v and / or acceleration a of the vehicle combination by a certain amount. In this way, the operation of the vehicle combination can be comfortably maintained within the safe operating range and not approach the limits 406a, 406b of unsafe or unstable operation. In some examples, the safety margin can be a function of parameters such as the surface friction coefficient μ, the lateral acceleration a y or the steer wheel angle δ. For example, for a lower surface friction coefficient μ, the safety margin can be larger. In some examples, the safety margin can be an absolute value. In some examples, the safety margin can be a weighted function based on the accuracy and / or uncertainty of the input parameters. When the input variables are signals from measurements or estimates, the uncertainty in the signals can be approximately estimated. Then weights can be assigned to each input parameter based on the accuracy and / or uncertainty of each input parameter. For example, if the measurement or estimate of a certain parameter has a high accuracy and / or a low uncertainty, a lower weight can be given to it in the weighted function, resulting in a smaller safety margin. This means that the safety margin is dynamically adjusted based on the reliability or confidence associated with the input parameters. This provides an adaptive and responsive safety margin and is useful for variables that are difficult to estimate, such as the surface friction coefficient.

[0067] In some examples, at 380, one or more output safe torque distributions 404 can be checked based on one or more second criteria, and resulting actions can be taken. In one example, a machine learning model can be required to output a minimum number of safe torque distributions 404. In one example, the torque distribution can be required to have sufficient resolution (e.g., a sufficiently uniform distribution between a maximum value and a minimum value). In one example, all measurements can be required to comply with a maximum uncertainty and / or a minimum accuracy. If one or more second criteria are not met, the resulting actions can include generating an alert for an operator of the vehicle combination 100 or modifying the torque request for the vehicle combination 100. In this way, the operator of the vehicle can be informed that there is no optimal torque distribution available, or the vehicle combination 100 can be controlled differently to meet one or more second criteria. If one or more second criteria are met, method 300 can continue.

[0068] In some examples, at 390, the output safe torque distribution 404 can be caused to be implemented as the torque distribution for the vehicle combination 100. This can be performed by the VMM of the vehicle combination 100, as will be discussed in conjunction with Figure 5 discussed. Accordingly, the vehicle combination 100 can be caused to move according to the torque distribution that has been ensured to be safe. As discussed above, the torque distribution can be determined based on the current operating point 140 of the vehicle combination 100, which means that the safe torque distribution can be provided online (i.e., when the vehicle combination 100 is in motion).

[0069] By training and implementing the machine learning model in this way, a fast and reliable way can be provided for the vehicle combination 100 to output a safe torque distribution based on the state of the vehicle combination. The training of the model avoids outputting unsafe and / or unstable operating modes of the vehicle combination. The model can be implemented online such that the vehicle combination can be controlled in a safe manner according to the parameters of the vehicle combination 100 and / or its environment. This implementation also allows a range of possible safe torque distributions to be provided from which a safe torque distribution can be selected based on given parameters, e.g., selecting a safe torque distribution that provides good energy efficiency. The safe torque distribution can be comfortably maintained within a safe operating range and not approach the limits of unsafe or unstable operation.

[0070] Figure 5 is a schematic diagram of an exemplary computer environment 500 for implementing methods 200 and 300. The computer environment 500 can include a first computer system 502 for training a machine learning model and a second computer system 504 for implementing the trained machine learning model.

[0071] The first computer system 502 includes processing circuitry 506 configured to perform the steps of method 200. To this end, the first computer system 502 may perform multiple simulations for the movement of the vehicle combination 100, classify each simulation as safe or unsafe, and use the resulting data points to train a machine learning model. The trained machine learning model may then be implemented by the first computer system 502 or another computer system, such as the second computer system 504. In some examples, the first computer system 502 may be a remote system implemented at a distance from the vehicle combination 100, such as in a central operating facility associated with the vehicle combination 100.

[0072] The second computer system 504 includes processing circuitry 508 configured to perform the steps of method 300. To this end, the first computer system 502 may receive an input in the form of an operating point 140 of the vehicle combination 100 and output one or more safe torque distributions for the vehicle combination 100. The second computer system 504 may be a vehicle control unit configured to perform various vehicle control functions, such as vehicle motion management. In some examples, the second computer system 504 may be located locally to the vehicle combination 100.

[0073] The first computer system 502 may be communicatively coupled to the second computer system 504 in any suitable manner, such as via circuitry or any other wired, wireless, or network connection known in the art. Additionally, the communicative coupling may be implemented as a direct connection between the first computer system 502 and the second computer system 504, or it may be implemented as a connection via one or more intermediate entities.

[0074] The second computer system 504 may also be communicatively coupled to the VMM 510, which includes processing circuitry 512 configured to control components of the vehicle combination 100, such as the propulsion and / or braking systems of the vehicle combination 100. The VMM 510 is implemented locally to the vehicle combination 100. The VMM 510 may be implemented at the vehicle and / or unit level, such as as a combined controller or a set of distributed controllers across the units 110 of the vehicle combination 100. The VMM 510 may receive a safe torque distribution from the second computer system 504 and implement it in the relevant systems of the vehicle combination 100. For example, the VMM 510 may receive a safe torque distribution for the vehicle combination 100 from the second computer system 504, which converts the safe torque distribution into torque distributions for each unit 110 of the vehicle combination 100. Alternatively, the VMM 510 may receive safe torque distributions for each unit 110 of the vehicle combination 100 from the second computer system 504.

[0075] In some examples, the computer environment 500 can be implemented at a single location. That is, the first computer system 502 and the second computer system 504 can be co-located, or can actually be implemented as a single computer system. Thus, the training of the machine learning model and its implementation can be performed at the same location. In one example, the combined system is implemented away from the vehicle combination 100, such as in a central operation facility associated with the vehicle combination 100. The implementation of the trained machine learning model can be used to generate a database or look-up table for the safety torque distribution for different operating points, which can then be stored in the control system of the vehicle combination 100 for querying when necessary. This can reduce the computing resources required by the vehicle combination 100. In another example, the combined system is implemented locally to the vehicle combination 100, such as in the control system of the vehicle combination 100. This means that all the functions of the methods disclosed herein can be implemented in an independent manner on the vehicle combination 100.

[0076] Figure 6 is a schematic diagram of a computer system 600 for implementing the examples disclosed herein. Specifically, according to some examples, the computer system 600 can be configured to cause the execution of Figure 2 method 200 and / or Figure 3 method 300. According to some examples, the computer system 600 can be included in the Figure 5 first computer system 502 and / or the second computer system 504. The computer system 600 is adapted to execute instructions from a computer-readable medium to perform these and / or any functions or processes described herein. The computer system 600 can be connected (e.g., networked) to other machines in a LAN, intranet, extranet, or the Internet. Although only a single device is shown, the computer system 600 can include any collection of devices that individually or jointly execute a set (or multiple sets) of instructions to perform any one or more of the methods discussed herein. Thus, any reference in this disclosure and / or the claims to a computer system, computing system, computer device, computing device, control system, control unit, electronic control unit (ECU), processor device, processing circuit, etc. includes a reference to one or more such devices to individually or jointly execute an instruction set (or multiple instruction sets) to perform any one or more of the methods discussed herein. For example, a control system can include a single control unit or multiple control units that are connected to each other or otherwise communicatively coupled such that any executed function can be distributed among the control units as needed. Additionally, such devices can communicate with each other or with other devices through various system architectures, such as directly or via a controller area network (CAN) bus, etc.

[0077] The computer system 600 may include at least one computing device or electronic device capable of including firmware, hardware, and / or executing software instructions to implement the functionality described herein. The computer system 600 may include processing circuitry 602 (e.g., processing circuitry including one or more processor devices or control units), a memory 604, and a system bus 606. The computer system 600 may include at least one computing device having the processing circuitry 602. The system bus 606 provides an interface for system components including, but not limited to, the memory 604 and the processing circuitry 602. The processing circuitry 602 may include any number of hardware components for performing data or signal processing or for executing computer code stored in the memory 604. The processing circuitry 602 may, for example, include a general-purpose processor, a dedicated processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a circuit including processing components, a group of distributed processing components, a group of distributed computers configured for processing, or other programmable logic devices, discrete gate or transistor logic, discrete hardware components designed to perform the functions described herein, or any combination thereof. The processing circuitry 602 may also include computer-executable code for controlling the operation of the programmable devices.

[0078] The system bus 606 may be any of several types of bus structures that may further interconnect to a memory bus (with or without a memory controller), a peripheral bus, and / or a local bus using any of a variety of bus architectures. The memory 604 may be one or more devices for storing data and / or computer code to complete or facilitate the methods described herein. The memory 604 may include database components, object code components, script components, or other types of information structures for supporting the various activities herein. Any distributed or local memory device may be utilized with the systems and methods of this specification. The memory 604 may be communicatively connected to the processing circuitry 602 (e.g., via circuitry or any other wired, wireless, or network connection) and may include computer code for performing one or more of the processes described herein. The memory 604 may include non-volatile memory 608 (e.g., read-only memory (ROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.) and volatile memory 610 (e.g., random access memory (RAM)) or any other medium that can be used to carry or store the desired program code in the form of machine-executable instructions or data structures and that can be accessed by a computer or other machine having the processing circuitry 602. The basic input / output system (BIOS) 612 may be stored in the non-volatile memory 608 and may include basic routines that help to transfer information between elements within the computer system 600.

[0079] The computer system 600 may also include or be coupled to a non-transitory computer-readable storage medium such as a storage device 614, which may include, for example, an internal or external hard disk drive (HDD) (e.g., enhanced integrated drive electronics (EIDE) or serial advanced technology attachment (SATA)), an HDD for storage (e.g., EIDE or SATA), flash memory, etc. The storage device 614 and other drives associated with the computer-readable medium and the computer-usable medium may provide non-volatile storage of data, data structures, computer-executable instructions, and the like.

[0080] The hard-coded or soft-coded computer code may be provided in the form of one or more modules. The modules may be implemented as software and / or hard-coded in circuitry to implement all or part of the functions described herein. The modules may be stored in the storage device 614 and / or the volatile memory 610, which may include an operating system 616 and / or one or more program modules 618. All or a portion of the examples disclosed herein may be implemented as a computer program 620 stored on a transitory or non-transitory computer-usable or computer-readable storage medium such as the storage device 614 (e.g., a single medium or multiple media), which includes complex programming instructions (e.g., complex computer-readable program code) that cause the processing circuitry 602 to perform the actions described herein. Thus, the computer-readable program code of the computer program 620 may include software instructions for implementing the functions of the examples described herein when executed by the processing circuitry 602. In some examples, the storage device 614 may be a computer program product (e.g., a readable storage medium) on which the computer program 620 is stored, where at least a portion of the computer program 620 may be loadable (e.g., loaded into the processor) for implementing the functionality of the examples described herein when executed by the processing circuitry system 602. The processing circuitry system 602 may act as a controller or control system for the computer system 600 for implementing the functionality described herein.

[0081] The computer system 600 may include an input device interface 622 configured to receive inputs and selections to be communicated to the computer system 600, such as from a keyboard, mouse, touch-sensitive surface, etc., when executing instructions. Such input devices may be connected to the processor circuitry 602 through the input device interface 622 coupled to the system bus 606, but may be connected through other interfaces (such as a parallel port, Institute of Electrical and Electronics Engineers (IEEE) 1394 serial port, Universal Serial Bus (USB) port, IR interface, etc.). The computer system 600 may include an output device interface 624 configured to forward outputs to, such as a display, video display unit (e.g., liquid crystal display (LCD) or cathode ray tube (CRT)). The computer system 600 may include a communication interface 626 adapted to communicate with a network as appropriate or as needed.

[0082] The operational actions described in any of the exemplary aspects herein are described to provide examples and discussion. These actions may be performed by hardware components, may be embodied in machine-executable instructions to cause a processor to perform these actions, or may be performed by a combination of hardware and software. Although a particular order of method actions may be shown or described, the order of the actions may be different. Additionally, two or more actions may be performed simultaneously or partially simultaneously.

[0083] According to certain examples, it is further disclosed that:

[0084] Example 1: A computer system (500, 502, 504) for training a machine learning model to determine a safe torque distribution (404) for a vehicle combination (100) including a tractor unit (110-1) and at least one trailer unit (110-2), the computer system (500, 502, 504) including processing circuitry (506, 508) configured to: perform a plurality of simulations of the movement of the vehicle combination (100) based on a plurality of torque distributions and a plurality of operating points (140) of the vehicle combination (100); classify each simulation as safe or unsafe; and use the plurality of simulations to train a machine learning model to determine a safe torque distribution (404) for the vehicle combination (100) based on the operating points (140) of the vehicle combination (100).

[0085] Example 2: The computer system (500, 502, 504) according to Example 1, wherein: each torque distribution includes a corresponding torque distribution for each unit (110) of the vehicle combination (100); and the sum of the corresponding torque distributions satisfies the total torque distribution for the vehicle combination (100).

[0086] Example 3: The computer system (500, 502, 504) according to Example 1 or 2, wherein each operating point (140) is defined based on one or more of the following: the speed v of the vehicle combination (100), the lateral acceleration a of the vehicle combination (100) y , the axle load F of one or more units (110) of the vehicle combination (100) z , the yaw rate of one or more units (110) of the vehicle combination (100) , the slip angle α of one or more units (110) of the vehicle combination (100), the slip angle change rate of one or more units (110) of the vehicle combination (100) , the wheel angle δ of one or more units (110) of the vehicle combination (100), the surface friction coefficient μ, the curve radius r, and / or the path gradient.

[0087] Example 4: The computer system (500, 502, 504) according to any of the preceding examples, wherein the processing circuit (506, 508) is configured to classify the simulation as unsafe if a certain simulation results in instability of the vehicle combination (100).

[0088] Example 5: A computer system (500, 502, 504) for determining a safe torque distribution (404) for a vehicle combination (100) including a tractor unit (110-1) and at least one trailer unit (110-2), the computer system (500, 502, 504) including a processing circuit (506, 508), the processing circuit being configured to: receive a plurality of torque distributions (402) that satisfy a torque request for the vehicle combination (100); receive an operating point (140) of the vehicle combination (100); and use a machine learning model according to any one of Examples 1 to 4 to output one or more safe torque distributions (404) for the vehicle combination (100) based on the received torque distributions (402) and the operating point (140).

[0089] Example 6: The computer system (500, 502, 504) according to Example 5, wherein the received operating point (140) is the current operating point of the vehicle combination (100).

[0090] Example 7: The computer system (500, 502, 504) according to Example 5 or 6, wherein: the machine learning model is configured to output a plurality of safety torque distributions (404) for the vehicle combination (100); and the processing circuit (506, 508) is configured to determine a safety torque distribution (404a, 404b) from the plurality of safety torque distributions based on one or more first criteria.

[0091] Example 8: The computer system (500, 502, 504) according to any one of Examples 5 to 7, wherein the processing circuit (506, 508) is configured to generate an alert or modify the torque request if the one or more output safety torque distributions (404) do not meet one or more second criteria.

[0092] Example 9: The computer system (500, 502, 504) according to any one of Examples 5 to 8, wherein the processing circuit (506, 508) is configured to apply a safety margin to the one or more output safety torque distributions (404).

[0093] Example 10: The computer system (500, 502, 504) according to any one of Examples 5 to 9, wherein the processing circuit (506, 508) is configured to cause the output safety torque distribution (404) to be implemented as a torque distribution for the vehicle combination (100).

[0094] Example 11: A computer system (500, 502, 504) for determining a safety torque distribution (404) for a vehicle combination (100) including a tractor unit (110-1) and at least one trailer unit (110-2), the computer system (500, 502, 504) including a processing circuit (506, 508), the processing circuit being configured to: perform a plurality of simulations of the movement of the vehicle combination (100) based on a plurality of torque distributions and a plurality of operating points (140) of the vehicle combination (100); classify each simulation as safe or unsafe; train a machine learning model using the plurality of simulations; receive a plurality of torque distributions (402) that satisfy a torque request for the vehicle combination (100); receive an operating point (140) of the vehicle combination (100); and use the trained machine learning model to output one or more safety torque distributions (404) for the vehicle combination (100) based on the received torque distributions (402) and operating points (140).

[0095] Example 12: A vehicle (100) comprising a computer system (500, 502, 504) according to any one of the preceding examples.

[0096] Example 13: A computer-implemented method (200) for training a machine learning model to determine a safe torque distribution (404) for a vehicle combination (100) comprising a tractor unit (110-1) and at least one trailer unit (110-2), the method (200) comprising: performing (210) a plurality of simulations of the movement of the vehicle combination (100) by a processing circuit (506, 508) of a computer system (500, 502, 504) based on a plurality of torque distributions and a plurality of operating points (140) of the vehicle combination (100); classifying (220) each simulation as safe or unsafe by the processing circuit (506, 508); and training (230) a machine learning model by the processing circuit (506, 508) using the plurality of simulations to determine a safe torque distribution (404) for the vehicle combination (100) based on the operating points (140) of the vehicle combination (100).

[0097] Example 14: The computer-implemented method (200) according to Example 13, wherein: each torque distribution comprises a corresponding torque distribution for each unit (110) of the vehicle combination (100); and the sum of the corresponding torque distributions satisfies the total torque distribution for the vehicle combination (100).

[0098] Example 15: The computer-implemented method (200) according to Example 13 or 14, wherein each operating point (140) is defined based on one or more of: the speed v of the vehicle combination (100), the lateral acceleration a y of the vehicle combination (100), one or more axle loads F of one or more units (110) of the vehicle combination (100) z the yaw rate of one or more units (110) of the vehicle combination (100) the slip angle α of one or more units (110) of the vehicle combination (100), the slip angle change rate of one or more units (110) of the vehicle combination (100) the wheel angle δ of one or more units (110) of the vehicle combination (100), the surface friction coefficient μ, the curve radius r and / or the path gradient.

[0099] Example 16: The computer-implemented method (200) according to any one of Examples 13 to 15, comprising classifying the simulation as unsafe by the processing circuit (506, 508) if a certain simulation results in instability of the vehicle combination (100).

[0100] Example 17: A computer-implemented method (300) for determining a safety torque distribution (404) for a vehicle combination (100) comprising a towing vehicle unit (110-1) and at least one towed unit (110-2), the method (300) comprising: receiving (330) by a processing circuit (506, 508) of a computer system (500, 502, 504) a plurality of torque distributions (402) that satisfy a torque request for the vehicle combination (100); receiving (340) by the processing circuit (506, 508) an operating point (140) of the vehicle combination (100); and outputting (350) by the processing circuit (506, 508) and using a machine learning model trained according to Example 12, one or more safety torque distributions (404) for the vehicle combination (100) based on the received torque distributions (402) and operating point (140).

[0101] Example 18: The computer-implemented method (300) according to Example 17, comprising receiving (340) by a processing circuit (506, 508) a current operating point (140) of the vehicle combination (100).

[0102] Example 19: The computer-implemented method (300) according to Example 17 or 18, comprising: outputting (350) by the processing circuit (506, 508) and using the trained machine learning model a plurality of safety torque distributions (404) for the vehicle combination (100); and determining (360) by the processing circuit (506, 508) a safety torque distribution (404a, 404b) from the plurality of safety torque distributions (404) based on one or more first criteria.

[0103] Example 20: The computer-implemented method (300) according to any one of Examples 17 to 19, further comprising: generating an alert or modifying the torque request by the processing circuit (506, 508) if the one or more output safety torque distributions (404) do not meet one or more second criteria.

[0104] Example 21: The computer-implemented method (300) according to any one of Examples 17 to 20, further comprising applying (370) a safety margin to the one or more output safety torque distributions (404) by the processing circuit (506, 508).

[0105] Example 22: The computer-implemented method (300) according to any one of Examples 17 to 21, further comprising causing (380) the output safety torque distribution (404) to be implemented as a torque distribution for the vehicle combination (100) by the processing circuit (506, 508).

[0106] Example 23: A computer-implemented method (200, 300) for determining a safety torque distribution (404) for a vehicle combination (100) comprising a tractor unit (110-1) and at least one trailer unit (110-2), the method comprising: performing (210) a plurality of simulations of the movement of the vehicle combination (100) by a processing circuit (506, 508) of a computer system (500, 502, 504) based on a plurality of torque distributions and a plurality of operating points (140) of the vehicle combination (100); classifying (220) each simulation as safe or unsafe by the processing circuit (506, 508); training (230) a machine learning model by the processing circuit (506, 508) using the plurality of simulations; receiving (330) a plurality of torque distributions (402) that satisfy a torque request of the vehicle combination (100) by the processing circuit (506, 508); receiving (340) an operating point (140) of the vehicle combination (100) by the processing circuit (506, 508); and outputting (350) one or more safety torque distributions (404) for the vehicle combination (100) by the processing circuit (506, 508) and using the trained machine learning model, based on the received torque distributions (402) and operating point (140).

[0107] Example 24: A computer program product comprising program code for performing the computer-implemented method (200, 300) according to Examples 13 to 23 when executed by a processing circuit (506, 508).

[0108] Example 25: A non-transitory computer-readable storage medium comprising instructions that, when executed by a processing circuit (506, 508), cause the processing circuit to perform the computer-implemented method (200, 300) according to any one of Examples 13 to 23.

[0109] The terms used herein are for the purpose of describing particular aspects only and are not intended to limit the disclosure. As used herein, unless the context clearly dictates otherwise, the singular forms "a" and "the" are intended to include the plural forms as well. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items. It should also be understood that the terms "comprises" and / or "comprising," when used herein, specify the presence of stated features, integers, acts, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, acts, steps, operations, elements, components, and / or groups thereof.

[0110] It should be understood that although the terms first, second, etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, without departing from the scope of the present disclosure, the first element may be referred to as the second element, and similarly, the second element may be referred to as the first element.

[0111] Relative terms such as "below" or "above" or "upper" or "lower" or "horizontal" or "vertical" may be used herein to describe the relationship of one element to another, as shown in the figures. It should be understood that these terms, as well as those discussed above, are also intended to cover different device orientations in addition to the orientation depicted in the figures. It will be understood that when an element is referred to as being "connected" or "coupled" to another element, the element may be directly connected or directly coupled to the other element, or intervening elements may be present. In contrast, when an element is referred to as being "directly connected" or "directly coupled" to another element, no intervening elements are present.

[0112] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. It should also be understood that unless explicitly defined herein, the terms used herein should be interpreted as having a meaning consistent with their meaning in the context of this specification and the relevant art, and should not be interpreted in an idealized or overly formal sense.

[0113] It should be understood that the present disclosure is not limited to the aspects described above and shown in the drawings; rather, those skilled in the art will recognize that many changes and modifications can be made within the scope of the present disclosure and the appended claims. In the drawings and the specification, the aspects have been disclosed for illustrative purposes only and not for purposes of limitation, and the scope of the disclosure is set forth in the appended claims.

Claims

1. A computer system (500, 502, 504) for training a machine learning model to determine a safe torque distribution (404) for a vehicle combination (100) including a tractor unit (110-1) and at least one trailer unit (110-2), the computer system (500, 502, 504) including processing circuitry (506, 508) configured to: perform a plurality of simulations of the movement of the vehicle combination (100) based on a plurality of torque distributions and a plurality of operating points (140) of the vehicle combination (100); classify each simulation as safe or unsafe; and use the plurality of simulations to train a machine learning model to determine a safe torque distribution (404) for the vehicle combination (100) based on the operating points (140) of the vehicle combination (100).

2. The computer system (500, 502, 504) according to claim 1, wherein: each torque distribution includes a corresponding torque distribution for each unit (110) of the vehicle combination (100); and the sum of the corresponding torque distributions satisfies the total torque distribution for the vehicle combination (100).

3. The computer system (500, 502, 504) according to claim 1 or 2, wherein each operating point (140) is defined based on one or more of the following: the speed v of the vehicle combination (100), the lateral acceleration a of the vehicle combination (100) y , the axle load F of one or more axles of one or more units (110) of the vehicle combination (100) z , the yaw rate of one or more units (110) of the vehicle combination (100) , the slip angle α of one or more units (110) of the vehicle combination (100), the slip angle change rate of one or more units (110) of the vehicle combination (100) , the wheel angle δ of one or more units (110) of the vehicle combination (100), the surface friction coefficient μ, the curve radius r, and the path gradient.

4. The computer system (500, 502, 504) according to any one of the preceding claims, wherein the processing circuitry (506, 508) is configured to classify a simulation as unsafe if the simulation results in instability of the vehicle combination (100).

5. A computer system (500, 502, 504) for determining a safe torque distribution (404) for a vehicle combination (100) including a tractor unit (110-1) and at least one trailer unit (110-2), the computer system (500, 502, 504) including processing circuitry (506, 508) configured to: receive a plurality of torque distributions (402) that satisfy a torque request for the vehicle combination (100); receive the operating points (140) of the vehicle combination (100); and use a machine learning model trained according to any one of claims 1 to 4 to output one or more safe torque distributions (404) for the vehicle combination (100) based on the received torque distributions (402) and operating points (140).

6. The computer system (500, 502, 504) according to claim 5, wherein the received operating points (140) are the current operating points of the vehicle combination (100).

7. The computer system (500, 502, 504) according to claim 5 or 6, wherein: the machine learning model is configured to output a plurality of safe torque distributions (404) for the vehicle combination (100); and the processing circuitry (506, 508) is configured to determine a safe torque distribution (404a, 404b) from the plurality of safe torque distributions (404) based on one or more first criteria.

8. The computer system (500, 502, 504) according to any one of claims 5 to 7, wherein the processing circuit (506, 508) is configured to generate an alert or modify the torque request if the one or more output safe torque distributions (404) do not meet one or more second criteria.

9. The computer system (500, 502, 504) according to any one of claims 5 to 8, wherein the processing circuit (506, 508) is configured to apply a safety margin to the one or more output safe torque distributions (404).

10. The computer system (500, 502, 504) according to any one of claims 5 to 9, wherein the processing circuit (506, 508) is configured to cause the output safe torque distribution (404) to be implemented as the torque distribution for the vehicle combination (100).

11. A vehicle (100) comprising the computer system (500, 502, 504) according to any one of the preceding claims.

12. A computer-implemented method (200) for training a machine learning model to determine a safe torque distribution (404) for a vehicle combination (100) comprising a tractor unit (110-1) and at least one trailer unit (110-2), the method (200) comprising: performing (210) a plurality of simulations of the movement of the vehicle combination (100) by a processing circuit (506, 508) of a computer system (500, 502, 504) based on a plurality of torque distributions and a plurality of operating points (140) of the vehicle combination (100); classifying (220) each simulation as safe or unsafe by the processing circuit (506, 508); and training (230) a machine learning model by the processing circuit (506, 508) using the plurality of simulations to determine a safe torque distribution (404) for the vehicle combination (100) based on the operating points (140) of the vehicle combination (100).

13. A computer-implemented method (300) for determining a safe torque distribution (404) for a vehicle combination (100) comprising a tractor unit (110-1) and at least one trailer unit (110-2), the method (300) comprising: receiving (330) by a processing circuit (506, 508) of a computer system (500, 502, 504) a plurality of torque distributions (402) that satisfy the torque request of the vehicle combination (100); receiving (340) by the processing circuit (506, 508) the operating points (140) of the vehicle combination (100); and outputting (350) by the processing circuit (506, 508) and using the machine learning model trained according to claim 12, one or more safe torque distributions (404) for the vehicle combination (100) based on the received torque distributions (402) and operating points (140).

14. A computer program product comprising program code for performing the computer-implemented method (200, 300) according to claim 12 or 13 when executed by a processing circuit (506, 508).

15. A non-transitory computer-readable storage medium comprising instructions that, when executed by a processing circuit (506, 508), cause the processing circuit to perform the computer-implemented method (200, 300) according to claim 12 or 13.

Images