Methods for quantifying vehicle path following performance

Abstract

A method for quantifying vehicle path following performance, the method comprising: obtaining samples of path following performance (I); selecting a subset of the path following performance samples so that the selected samples follow a predetermined statistical extreme value distribution; parameterizing the predetermined statistical extreme value distribution based on the selected path following performance samples; and quantifying the vehicle path following performance based on the parameterized statistical extreme value distribution.

Description

Technical Field

[0001] The present disclosure relates to a method, a control unit and a vehicle for quantifying the path following performance of a vehicle. The method disclosed herein is applied, for example, in autonomous driving, advanced driver assistance systems and other safety-critical vehicle functions.

[0002] The invention is applicable to heavy vehicles such as trucks, semi-trailers, construction equipment. Although the invention will be described primarily for cargo transport vehicles, the invention is not limited to this particular vehicle but may also be used in other types of vehicles, such as construction equipment, buses and cars. Background Art

[0003] Automated driving systems and other advanced driver assistance systems (ADAS) are constantly making decisions about how to control the host vehicle to achieve a given goal without exposing the vehicle to risk. These decisions are based on a variety of different factors related to the state of the host vehicle and the surrounding environment, as well as the vehicle's ability to perform different types of maneuvers.

[0004] When controlling a vehicle, there are always uncertainties that need to be taken into account to avoid exposing the vehicle to risk. These uncertainties include uncertain vehicle state estimates and uncertainties in vehicle actuation.

[0005] When performing vehicle trajectory planning, it is often desirable to understand how well a particular vehicle assembly can follow different types of paths. Therefore, extensive testing is often performed using different types of vehicles to perform different types of maneuvers under different conditions. Path following performance is then evaluated and modeled for a given vehicle type and maneuver, for example based on observations of how well the vehicle can perform different maneuvers.

[0006] The problem with observation-based modeling of vehicle path following performance is that a large amount of data is required to accurately model rare events such as vehicle skidding, loss of control, jack-knifing of the tractor relative to the trailer, etc. Collecting such a large amount of data is expensive, time-consuming, and in some cases even impractical.

[0007] Improved methods of quantifying path-following performance are needed that allow rare events to be effectively modeled in a reliable manner. Summary of the invention

[0008] One object of the present disclosure is to provide a method for quantifying and modeling vehicle path following performance. This object is achieved by a method for quantifying vehicle path following performance. The method comprises: obtaining a path following performance selecting a subset of the path following performance samples so that the selected samples follow a predetermined statistical extreme value distribution; parameterizing the predetermined statistical extreme value distribution based on the selected path following performance samples; and quantifying the vehicle path following performance based on the parameterized statistical extreme value distribution.

[0009] An important feature of the disclosed method is that it reduces the amount of data required to quantify the path following performance of a vehicle. By allowing the path following performance to be modeled based on a reduced data set, development and testing time can be reduced, which is an advantage. Also, certain types of analysis that were previously impossible to perform because too much data was required can now be achieved with the disclosed method.

[0010] It is an advantage that the disclosed method is applicable to offline processing and / or online processing. For example, online processing can be used to supplement more extensive offline processing to improve model validation and thus improve vehicle operation.

[0011] The predetermined statistical extreme value distribution may be, for example, a generalized Pareto distribution (GDP) or a generalized extreme value distribution (GEV). These types of extreme value distributions have been extensively studied, which means that the disclosed method can utilize well-known and robust analysis techniques, which is an advantage.

[0012] According to various aspects, the method includes measuring a time between exceedances metric, and monitoring an operating design domain (ODD) based on the time between exceedances metric, the time between exceedances metric indicating the time elapsed between path following performance samples that exceed a threshold value ζ. This is a relatively simple metric for determining when a vehicle is operating outside of its ODD, but is a powerful indicator.

[0013] According to some examples, vehicle path following performance samples include any of the following: lateral vehicle position, longitudinal vehicle position, absolute or squared vehicle position, lateral vehicle velocity, longitudinal vehicle velocity, absolute or squared vehicle velocity, lateral vehicle acceleration, longitudinal vehicle acceleration, vehicle acceleration, absolute or squared vehicle acceleration, curvature, yaw rate, heading, and articulation angle. It is worth noting that these quantities are examples, and several additional examples may be given.

[0014] According to various aspects, vehicle path following performance samples are determined separately for one or more portions of the vehicle. Different portions of the vehicle may be associated with different extreme value distributions, i.e., a trailer may not behave exactly the same as a tractor, so it is best to treat the trailer separately from the tractor. Also, different axles or different groups of wheels on a vehicle may behave differently. By using different vehicle path following performance samples for different portions of the vehicle, a more refined analysis can be performed.

[0015] For example, the vehicle path following performance samples may be obtained based on any of the following: type of vehicle, type of vehicle combination, vehicle physical dimension parameters, weather conditions, road friction, road geometry, and path trajectory geometry.

[0016] According to some aspects, the method further comprises: based on the predetermined statistical extreme value distribution, defining the performance as being within the range The path following performance is quantified as a bounded model with probability greater than 1-γ.

[0017] The bounds on the path following performance can be used to conduct risk assessment for various driving scenarios both offline and in real time. For example, the bounded model obtained from the disclosed method can also be used to decide when to trigger an emergency maneuver of the vehicle.

[0018] According to some other aspects, the method further comprises determining a confidence value β associated with the predetermined statistical extreme value distribution.

[0019] This confidence value is discussed in more detail below. It indicates the confidence obtained in the parameterized GEV or GDP, that is, whether the GEV or GDP can accurately fit the available data or whether more data is needed to get a good fit.

[0020] According to some further aspects, the method includes evaluating whether the collected path following performance data is sufficient for vehicle modeling based on a confidence value β associated with the predetermined statistical extreme value distribution.

[0021] The methods disclosed herein can also provide constructive feedback about when data collection for a given purpose is complete, e.g., when a sufficient amount of data has been collected to quantify path following performance in certain scenarios. For example, according to various aspects, the method includes evaluating whether the collected model data is sufficient for model correctness verification based on a confidence value β associated with a parameterized GEV or GDP.

[0022] Optionally, the disclosed method may further include monitoring an operating design domain (ODD) associated with the vehicle by comparing a parameterized statistical extreme value distribution with a set of benchmark distribution parameters, wherein operation outside the ODD is indicated by differences between the parameterized statistical extreme value distribution parameters and the benchmark distribution parameters.

[0023] In other words, another feature of the disclosed method is that the intermediate results can be used to construct a monitor of at least a portion of the ODD. These particular aspects can also be used as the basis for a separate independent method that does not rely on the above-described method. Therefore, a method for monitoring the ODD associated with a vehicle is disclosed herein. The method includes obtaining a parameterized GEV or GDP from vehicle path following performance samples during vehicle operation, as described herein or described in some other manner. The method also includes: monitoring the ODD associated with the vehicle by comparing the parameterized GEV or GDP with a set of baseline GEV or GDP parameters, wherein operation outside the ODD is indicated by the difference between the parameterized GEV or GDP parameters and the baseline parameters.

[0024] Furthermore, a control unit, a computer program and a vehicle are disclosed herein which are associated with the same advantages discussed above in connection with the different methods.

[0025] Generally, all terms used in the claims should be interpreted according to their ordinary meanings in the technical field, unless otherwise clearly defined herein. Unless otherwise clearly stated, all references to "one / an / the element, device, part, device, step, etc." should be openly interpreted as referring to at least one example of the element, device, part, device, step, etc. Unless clearly stated, the steps of any method disclosed herein do not have to be performed in the exact order disclosed. When studying the attached claims and the following description, other features and advantages of the present invention will become apparent. Those skilled in the art will recognize that, without departing from the scope of the present invention, the different features of the present invention can be combined to produce embodiments other than the embodiments described below. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] With reference to the accompanying drawings, below is a more detailed description of embodiments of the present invention cited as examples. In these drawings:

[0027] Figure 1 A vehicle is schematically shown;

[0028] Figure 2 An example situational evasive maneuver of a vehicle is shown;

[0029] Figure 3 An example path with path following error is shown;

[0030] Figure 4 An example threshold operation is shown;

[0031] Figure 5 Another example threshold operation is shown;

[0032] Figure 6 The bounded set of vehicle states is shown;

[0033] Figure 7 Schematically illustrating vehicle path following;

[0034] Figure 8 is a flow chart illustrating a method;

[0035] Fig. 9 A control unit is schematically shown; and

[0036] Fig.10 An example computer program product is shown. DETAILED DESCRIPTION

[0037] Various aspects of the present disclosure will now be described more fully with reference to the accompanying drawings. However, the various devices and methods disclosed herein may be implemented in many different forms and should not be construed as being limited to the aspects set forth herein. The same numbers in the accompanying drawings refer to the same elements throughout.

[0038] The terms used herein are only used to describe aspects of the present disclosure and are not intended to limit the present invention.As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

[0039] Figure 1 The vehicle 100 including the control unit 110 is schematically shown. The control unit 110 may implement various types of support systems and systems for vehicle functional safety, such as an advanced driver assistance system (ADAS) and a function for autonomous driving (AD).

[0040] The vehicle may be connected 120 to a processing device 130 arranged to perform off-line calculations to assist the vehicle control system comprising the control unit 110. The connection 120 is preferably wireless, but it may also be a wired connection or a connection via some storage module (e.g. a hard drive etc.).

[0041] One example function that the control unit 110 (and possibly the processing device 130) may support is the planning and execution of a situation avoidance maneuver (SAM). A safe stop maneuver is an example of a SAM. However, a SAM may also include maintaining a constant speed in a given lane or performing an obstacle avoidance maneuver. In general, the category of situation avoidance maneuvers includes all maneuvers that may be performed to avoid an undesirable situation, such as a detected risky situation. However, a SAM may also relate to, for example, a difficult parking situation, etc.

[0042] Figure 2An example SAM 200 is shown. The vehicle 100 detects that there is a problem at point A and the vehicle needs to perform a safety stop maneuver. For example, the vehicle 100 may have experienced some type of sensor failure at point A, such as a radar system failure or a camera failure condition.

[0043] Figure 2 The example SAM shown in includes moving the vehicle 100 to the side of the road 210 and stopping the vehicle in a controlled manner at point B. Thus, the control unit 110 determines a trajectory with an associated speed profile that the vehicle 100 will follow during the SAM. The vehicle is then controlled by the control unit 110 to follow the determined trajectory including the speed profile, for example. However, the vehicle may not be able to accurately follow the trajectory or maintain an accurate desired speed. Instead, there will be some uncertainty as to how well the vehicle will be able to follow a given planned maneuver trajectory. This uncertainty is Figure 2 The uncertainty region 220 is shown in FIG. 2 , which shows the region in which the vehicle may eventually arrive if the vehicle is controlled to follow the planned path with the desired speed profile. The area and shape of the uncertainty region 220 are generally related to the longitudinal velocity v of the vehicle. x and lateral acceleration a y Associated.

[0044] Here, "safe" is to be given a broad interpretation. A safe manoeuvre or a safe vehicle state may be a manoeuvre or state in which the vehicle and / or vehicle occupants and / or other road users are not at risk of injury or damage.

[0045] Some examples of safe and unsafe conditions can be given;

[0046] According to some aspects, a situation where a risk of collision is unlikely to occur can be considered a safe state.

[0047] According to some other aspects, situations where a collision risk is not impossible may still be considered a safe state, depending on the specific object. That is, a collision with a small bush or tree may be considered safe, while a collision with another vehicle or a larger object (such as a brick wall) may be considered unsafe.

[0048] According to some further aspects, for example, a collision with another vehicle that has been predetermined through vehicle-to-vehicle (V2V) communication (tolerating a low risk of collision) may be considered safe.

[0049] According to various aspects, situations in which the vehicle risks leaving the drivable area are considered unsafe.

[0050] According to other aspects, leaving the drivable area may be deemed safe, depending on the characteristics of the terrain outside the drivable area, as described above.

[0051] The vehicle state is a collection of variables that together describe what state the vehicle is currently in. Here, the vehicle state includes variables related to the vehicle's position (coordinates) and orientation (e.g., heading, steering angle, and articulation angle). The vehicle state also includes information related to the dynamic state of the vehicle, i.e., vehicle speed, acceleration, turn rate, etc. The vehicle state is typically represented as a vector of state variables x. As will be discussed in more detail below, the allowed vehicle state space can typically include upper and lower limits on state variables, such as lateral position during maneuvers.

[0052] The vehicle transitions between states based on issued control commands, such as braking commands or steering commands. The vehicle state can usually be defined with high probability as being within a certain set.

[0053] refer to Figure 6 , the vehicle 100 may currently be associated with a state x contained in a certain state set S, and wishes to transition to some other state set X. Therefore, the control unit 110 issues a control command u. For a given target set X, a one-step robust controllable set (or pre-image set) S is defined as the set of states that stably maps to X. Pre-image sets are discussed, for example, by F. Borrelli, A. Bemporad, and M. Morari in "Predicative Control for linear and hybrid systems" (Cambridge University Press, 2015), and are therefore not discussed in detail here. For all possible disturbances, the set S is the set of vehicle states (such as Figure 3 As shown in FIG. 1 ), the control signal changes the vehicle state to a state included in the target state set X. The interference set can be assumed to be bounded, or it can be assumed to be a certain set representing a fixed probability of occurrence. In other words, various known methods can be used to determine that as a result of issuing a control signal u starting from the vehicle state set S, the vehicle will change to a state x included in the state set X. In order to be able to define the interference set, the vehicle path following performance must be modeled.

[0054] Figure 3 A vehicle path 300 is schematically shown. The vehicle path may be associated with a planned SAM, for example. The path starts at position point A at time t=0. The vehicle initial state S(0) includes an initial position vector x s (t), initial velocity vector v s (t) and the initial acceleration vector a s(t). The trajectory to be followed is defined by a sequence of target vehicle states T(t) as a function of time t. Thus, at time t, the vehicle should be in a target state T(t) that includes the target position vector x T (t), target velocity vector v T (t) and the target acceleration vector a T (t). At any given time, for example, at time t1, there may be a difference between the target vehicle state T(t1) at time t1 and the actual vehicle state S(t1) at time t1. This difference is represented herein as a trajectory error e(t1), which can be used as a measure of path following performance. This measure can be sampled and used to characterize the vehicle path following performance. If the path following performance is bounded (at least with high probability), then it can be determined that Figure 2 The uncertainty region in region 220.

[0055] Assume that densely sampled path following performance data has been obtained. Although this data can be used directly to determine the incidence of extreme cases of path following errors or other differences, it may not necessarily capture the behavior of the vehicle well. Instead, it is proposed here to optionally filter the densely sampled data to extract a series of low-dimensional aggregate samples that describe the specific path following performance we are interested in modeling. For example, if the goal is to quantify the frequency of large lateral excursions when attempting to follow a given path, these instances can be quantified by first detecting instances of lateral excursions based on some criteria (such as a threshold), and then using an aggregate measure (such as a two-dimensional sample consisting of a maximum excursion or threshold and the duration of time that the excursion is above a certain threshold). Cumulative density functions can also be used to describe aggregate behavior based on the relative percentage of time that the excursion is above certain values. For example, densely sampled data can also be replaced by a block maximum, block mean, or block median determined for consecutive blocks of the densely sampled data, thereby providing less densely sampled data.

[0056] Extreme value theory (EVT) is a statistical field that focuses on the properties of rare events. EVT is discussed, for example, in C. Scarrott and A. MacDonald, "A review of extreme value threshold estimation and uncertainty quantification" (REVSTAT Statistical Journal, Vol. 10, No. 1, 2012) and references cited therein.

[0057] In probability theory and statistics, the generalized extreme value (GEV) distribution is a family of continuous probability distributions developed in extreme value theory that combines the Gumbel, Fréchet, and Weibull families, also known as type I, II, and III extreme value distributions. According to the extreme value theorem, the GEV distribution is the only possible limiting distribution of the appropriately normalized maximum of a series of independent and identically distributed random variables. It should be noted that the existence of a limiting distribution is not required, as this would require regularity conditions on the tails of the distribution in question. Nevertheless, the GEV distribution is often used as an approximation to model the maximum of a long (finite) sequence of random variables. In some application areas, the generalized extreme value distribution is called the Fisher-Tippett distribution, named after Ronald Fisher and LHC Tipett. However, the use of this name is sometimes restricted to represent a special case of the Gumbel distribution.

[0058] Peak Exceedances at Thresholds (POT) are a class of methods within EVT that are used to model extreme events that exceed a certain threshold. Under some assumptions, the exceedances converge to the so-called Generalized Pareto (GDP) distribution. For a particular data set, these exceedances can be used to estimate the parameters of the GDP distribution, which in turn can be used to predict the likelihood of rarer events. This allows statistical justification of the limits of the assumptions using much less data than traditional methods. The results of this analysis can also be used to verify that a set of assumptions do seem to apply to the operational design domain (ODD) in which the data was collected. For example, the ODD can be characterized by maximum own speed, maximum road grade, maximum road inclination, etc.

[0059] Furthermore, these assumptions can be used when evaluating whether the ego vehicle can perform a particular maneuver, i.e., whether the maneuver is safe given the vehicle's expected path-following performance.

[0060] Parameters of the GEV and / or GDP distributions may also be used to create monitoring of some aspects of the ODD (i.e., assumptions about the vehicle's path following performance). More specifically, if the current vehicle's path following performance deviates from the GDP distribution (by more than some probability threshold), this indicates that there is a risk that the ego vehicle is no longer within the defined ODD, and therefore the ego vehicle may need to take preventive action.

[0061] The ODD monitor is also why it can be said that the system will remain safe over time in terms of changing the vehicle’s path following performance. This is because the system will be able to detect changes in path following performance before extreme events that could lead to an accident occur.

[0062] It is important to note that one benefit of using the method proposed in this paper is that there is no explicit need to observe path following errors, where the stated assumption is broken, to detect risks where the original ODD (i.e., original data collection) is no longer valid. It is sufficient to detect the distribution of semi-rare events that do not match the expected distribution (these events are still within the ODD).

[0063] Figure 4 A threshold operation 400 is schematically illustrated, wherein samples of path following performance 410, 420 have been collected. A threshold ζ is applied to select a subset of the path following performance samples such that the selected samples follow a predetermined statistical extreme value distribution.

[0064] Figure 5 Another thresholding operation is shown, here in two dimensions x1, x2. Path following performance samples 510, 520 have been collected. A threshold ζ is again applied to select a subset of the path following performance samples such that the selected samples follow a predetermined statistical extreme value distribution.

[0065] Figure 7 An example 700 of quantifying vehicle path following performance is shown. The vehicle 100 is initially in a state associated with some bounded uncertainty 710 at point A. The vehicle state here includes an indication of having disturbances or uncertainties n h The heading h A , with interference or uncertainty v The speed v A , and interference or uncertainty p The position p A The method disclosed herein can be used to quantify the vehicle path following performance, which means that the path following error 720, 730 can be defined when the vehicle attempts to follow the target trajectory 310. In this way, the future vehicle state can be defined within the set 740 with high probability.

[0066] Figure 8 is a flow chart showing a method summarizing the above discussion. A method for quantifying the path following performance of a vehicle is shown here. The method includes obtaining S1 path following performance For example, these samples may be obtained as a previously stored sample set S11 or may be obtained as a sample set during operation of the own vehicle 100 S12.

[0067] As described above, the path tracking performance samples may include any of the following: lateral vehicle position, longitudinal vehicle position, absolute or squared vehicle position, lateral vehicle velocity, longitudinal vehicle velocity, absolute or squared vehicle velocity, lateral vehicle acceleration, longitudinal vehicle acceleration, absolute or squared vehicle acceleration, curvature, yaw rate, heading, and vehicle articulation angle. It will be appreciated that this list is not exhaustive and further examples may be given.

[0068] According to some aspects, vehicle path following performance samples are determined separately for one or more parts of the vehicle. For example, a vehicle part may be a unit such as a trailer in a vehicle combination, or a vehicle part such as an axle or a pair of wheels.

[0069] According to some other aspects, the vehicle path following performance samples are obtained based on any of: type of vehicle, type of vehicle combination, vehicle physical dimension parameters, weather conditions, road friction, road geometry, and path trajectory geometry.

[0070] The method includes selecting a subset of S2 path following performance samples so that the selected samples follow a predetermined statistical extreme value distribution. For example, the predetermined statistical extreme value distribution can be a generalized Pareto distribution (GDP) or an instance of a generalized extreme value distribution (GEV).

[0071] According to some aspects, the selection includes: determining S21 a threshold value ζ so that the vehicle path following performance samples exceeding the threshold value ζ follows the predetermined statistical extreme value distribution. Figure 4 and Figure 5 Some example thresholds are shown in . These thresholds can be fixed or dynamically adjusted based on some performance criteria.

[0072] The method further comprises: parameterizing S3 the predetermined statistical extreme value distribution based on the selected path following performance samples, and quantifying S4 the vehicle path following performance based on the parameterized statistical extreme value distribution. Methods for parameterizing extreme value distributions are known and therefore will not be discussed in detail here.

[0073] According to some aspects, the method includes measuring S7 a time metric between limit violations (the time metric between limit violations indicates the time elapsed between vehicle path following performance samples exceeding a threshold value ζ), and monitoring the ODD based on the time metric between limit violations. In this way, it can be determined whether the vehicle is operating in its ODD, or whether traffic conditions have changed so much that the vehicle 100 is no longer operating in the expected ODD.

[0074] According to some aspects, the method includes: based on the predetermined statistical extreme value distribution, by defining the performance as being within the range The vehicle path following performance is quantified as S41 into a bounded model with a probability greater than 1-γ.

[0075] According to some aspects, the method comprises determining S5 a confidence value β associated with said predetermined statistical extreme value distribution. can be used to estimate parameters of a GDP or GEV (with corresponding confidence levels). The identified distribution can in turn be used to find bounds on the vehicle path following performance, with confidence levels β satisfying these bounds. The vehicle path following performance model has a verified correctness γ and a confidence value β, which is associated with the parameterized GEV or GDP. Methods for determining the confidence value β are known and will not be discussed in detail herein. The confidence value depends at least in part on the number of observations collected over a threshold ζ and provides constructive feedback on the amount of data required. In this way, it is possible to estimate when a sufficient set of data has been collected to model the path following performance and when more data is needed. Therefore, an advantage of the disclosed methods is that they can be used to assess whether the collected model data is sufficient for model correctness verification based on the confidence value β associated with the parameterized GEV or GDP. Therefore, if the confidence value β after the GEV or GDP has been parameterized is too low, more data may be needed to increase the confidence value. In other words, according to some aspects, the method includes evaluating whether the vehicle path following performance data collected at S51 is sufficient for quantification based on a confidence value β associated with the predetermined statistical extreme value distribution.

[0076] According to some other aspects, the method includes monitoring S6 an operating design domain ODD associated with the vehicle by comparing the parameterized predetermined statistical extreme value distribution to a set of benchmark distribution parameters, wherein operation outside the ODD is indicated by differences between the parameterized predetermined statistical extreme value distribution parameters and the benchmark distribution parameters.

[0077] Although the described vehicle path following performance modeling can be performed offline (i.e., at the design stage), the identified GDP or GEV distribution can also be used to build an ODD monitor for online use. In particular, if vehicle path following performance samples are collected online, the distribution of the exceedances can be monitored. A significantly different distribution of exceedances compared to the distribution obtained offline indicates that the system exceeds its ODD. In other words, by comparing the parameterized GDP or GEV to a set of baseline GDP or GEV parameters, using the disclosed technology, it becomes possible to monitor the ODD associated with the vehicle. For example, a difference between the parameterized GDP parameters and the baseline GDP parameters indicates operation outside the ODD. In the event that operation outside the ODD is detected, a warning signal or SAM can be triggered.

[0078] This simple monitor counts the return period (ie, the average time between limit violations) over some time window. A significant reduction in the return period would indicate that the vehicle is operating outside its ODD.

[0079] Fig. 9 Components of the control unit 110 according to the embodiments discussed herein are schematically shown in a number of functional units. The processing circuit 910 is provided using any combination of one or more of the following: a suitable central processing unit CPU, a multiprocessor, a microcontroller, a digital signal processor DSP, etc., which is capable of executing software instructions stored in a computer program product, for example, in the form of a storage medium 930. The processing circuit 910 may also be provided as at least one application specific integrated circuit ASIC or a field programmable gate array FPGA.

[0080] In particular, the processing circuit 910 is configured to cause the control unit 110 to perform a set of operations or steps, such as combining Figure 4 For example, the storage medium 930 may store the set of operations, and the processing circuit 910 may be configured to retrieve the set of operations from the storage medium 930 to cause the control unit 110 to perform the set of operations. The set of operations may be provided as a set of executable instructions. Thus, the processing circuit 910 is thereby arranged to perform the method as disclosed herein.

[0081] The storage medium 930 may also include a permanent storage, such as any one or combination of magnetic storage, optical storage, solid-state storage, or even a remotely mounted storage.

[0082] The control unit 110 may also include an interface 920 for communicating with at least one external device (e.g., a remote server 130 and / or an onboard vehicle system). Thus, the interface 920 may include one or more transmitters and receivers, including analog and digital components and a suitable number of ports for wired or wireless communication.

[0083] The processing circuit 910 controls the general operation of the control unit 110, for example, by sending data and control signals to the interface 920 and the storage medium 930, by receiving data and reports from the interface 920, and by retrieving data and instructions from the storage medium 930. In order not to obscure the concepts presented here, control nodes are omitted.

[0084] Fig.10 A computer program product 1000 is schematically shown, which includes a set of operations 1010 that can be performed by the control unit 110. The set of operations 1010 can be loaded into the storage medium 930 in the control unit 110. The set of operations can correspond to the above-mentioned Figure 6 Methods of discussion.

[0085] exist Fig.10 In the example of , the computer program product 1000 is shown as an optical disc, such as a CD (compact disc) or a DVD (digital versatile disc) or a Blu-ray disc. The computer program product may also be implemented as a memory, such as a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), or an electrically erasable programmable read-only memory (EEPROM), and more specifically, as a non-volatile storage medium of a device in an external memory (such as a USB (universal serial bus) memory), or a flash memory (such as a compact flash). Therefore, although the computer program is schematically shown here as a track on the depicted optical disc, the computer program may be stored in any manner suitable for a computer program product.

Claims

1. A method for controlling a vehicle (1) to follow a vehicle path, performed by a control unit (110), the method comprising: determining a vehicle path (300) having an associated speed profile for the vehicle (100) to follow, wherein the vehicle path is defined by a series of target vehicle states as a function of time, Obtain (S1) path following performance The sample, where the path following performance p i is measured as a trajectory error indicating the difference between the target vehicle state and the actual vehicle state when following the vehicle path, Select (S2) a subset of the path following performance samples such that the selected samples follow a predetermined statistical extreme value distribution, parameterizing (S3) the predetermined statistical extreme value distribution based on the selected path following performance samples, quantifying (S4) the vehicle path following performance based on a parameterized statistical extreme value distribution, and The vehicle (1) is controlled by the control unit (110) to follow the vehicle path based on the quantified vehicle path following performance.

2. The method according to claim 1, wherein: The obtaining includes: obtaining (S11) a set of previously stored vehicle path following performance samples.

3. The method according to any one of claims 1 to 2, wherein: The obtaining comprises: obtaining (S12) a set of vehicle path following performance samples during the operation of the own vehicle (100).

4. The method according to any one of claims 1 to 2, wherein: The selection includes: determining (S21) a threshold value ζ so that the vehicle path following performance samples exceeding the threshold value ζ Complying with the predetermined statistical extreme value distribution.

5. The method according to claim 4, comprising: A time between limit violations metric is measured (S7) and the ODD is monitored based on the time between limit violations metric, the time between limit violations metric indicating the time elapsed between vehicle path following performance samples exceeding the threshold ζ.

6. The method according to any one of claims 1 to 2, wherein: The predetermined statistical extreme value distribution is the generalized Pareto distribution GDP.

7. The method according to any one of claims 1 to 2, wherein: The predetermined statistical extreme value distribution is a generalized extreme value distribution GEV.

8. The method according to any one of claims 1 to 2, wherein: Vehicle path following performance samples include any of the following: lateral vehicle position, longitudinal vehicle position, absolute or squared vehicle position, lateral vehicle velocity, longitudinal vehicle velocity, absolute or squared vehicle velocity, lateral vehicle acceleration, longitudinal vehicle acceleration, absolute or squared vehicle acceleration, curvature, yaw rate, heading, articulation angle.

9. The method according to claim 8, wherein: Vehicle path following performance samples are determined individually for one or more portions of the vehicle.

10. The method according to any one of claims 1 to 2, wherein: The vehicle path following performance samples are obtained based on any of the following: the type of vehicle, the type of vehicle combination, vehicle physical dimension parameters, weather conditions, road friction, road geometry, and path trajectory geometry.

11. The method according to any one of claims 1 to 2, comprising: Based on the predetermined statistical extreme value distribution, the performance is defined as being within the range The vehicle path following performance is quantified (S41) as a bounded model with a probability greater than 1-γ.

12. The method according to any one of claims 1 to 2, comprising: A confidence value β associated with said predetermined statistical extreme value distribution is determined (S5).

13. The method according to claim 12, comprising: Whether the collected vehicle path following performance data is sufficient for performance quantification is evaluated ( S51 ) based on the confidence value β associated with the predetermined statistical extreme value distribution.

14. The method according to any one of claims 1 to 2, comprising: An operational design domain (ODD) associated with the vehicle is monitored (S6) by comparing the parameterized predetermined statistical extreme value distribution with a set of benchmark distribution parameters, wherein operation outside the ODD is indicated by differences between the parameterized predetermined statistical extreme value distribution parameters and the benchmark distribution parameters.

15. A computer-readable medium (810) carrying a computer program (820), the computer program comprising program code components for performing the steps of the method of any one of claims 1 to 14 when the computer program is run on a computer or on a processing circuit (710) of a control unit (110).

16. A control unit (110) for quantifying a path following performance of a vehicle, the control unit (100) being configured to perform the steps of the method according to any one of claims 1-14.

17. A vehicle (1) comprising a control unit (110) according to claim 16.

Images