WO2021058223A1

WO2021058223A1 - Method for applying automated driving functions efficiently and in a simulated manner

Info

Publication number: WO2021058223A1
Application number: PCT/EP2020/073845
Authority: WO
Inventors: Nicolas Fraikin
Original assignee: Bayerische Motoren Werke Aktiengesellschaft
Priority date: 2019-09-27
Filing date: 2020-08-26
Publication date: 2021-04-01
Also published as: DE102019126195A1

Abstract

The invention relates to a method for applying automated driving functions efficiently and in a simulated manner. According to the invention, it is possible to reduce simulation processes and to achieve simulation results in real time with justifiable computational complexity, said simulation results being not only efficiently generated but also being of reliable quality. The present invention also relates to a correspondingly designed system arrangement. The invention also relates to a computer programme product having control commands which implement the method or operate the proposed system arrangement.

Description

Process for the efficient, simulative application of automated driving functions

The invention relates to a method for the efficient, simulative application of automated driving functions. According to the invention, it is possible to save simulation processes and to achieve simulation results in real time with a justifiable computing effort, which are not only efficiently generated, but are also reliable with regard to their quality. The present invention also relates to a correspondingly configured system arrangement. In addition, a computer program product with control commands is proposed which implement the method or operate the proposed system arrangement.

DE 102012220655 A1 shows a system and a method for managing vehicle configurations, as well as a database that contains a large number of configuration templates. The configuration template contains parameters of configurable functions and settings for a corresponding vehicle. WO 2019/060938 A1 shows a method for generating a dynamic speed profile of a vehicle, which is suitable for simulating an, in particular real, ferry operation on a route or is suitable for specifying target speeds for driver assistance systems, in particular for predictive driving functions.

WO 2007060 134 A1 shows a method for determining a value of a model parameter of a reference vehicle model. The reference value of the state variable is usually determined using a vehicle model. The vehicle models that can be used generally contain several parameters that have to be adapted to a specific vehicle type so that the model correctly reproduces the reference behavior of a specific vehicle.

It is known to carry out simulations with regard to the driving functions, for example in the context of autonomous or automated driving. A sufficient model accuracy must be selected to calculate reliable results. However, the prior art has the disadvantage that increases in efficiency typically come at the expense of system accuracy.

There is a trade-off between model accuracy and complexity / computing time. Depending on the application, either simple models are used, which do not exactly map the dynamics of the vehicle in all situations, but are not complex and enable quick simulation. Alternatively, a highly accurate analytical vehicle model is created, which consists of many individual components, has to be parameterized and validated in a complex manner and is very computationally intensive.

The motivation to develop a method for the efficient, simulative application of automated driving functions arose from the need to enable the parameterization of increasingly complex driving functions with a reasonable effort. Increased demands on the comfort and security of the systems and the growing variety of derivatives ensure increased application effort, so that applications in the vehicle can no longer be carried out with the same quality and level of detail as in the past. Instead Improved simulation tools are to be used to enable a systematic, scenario-based virtual application with the help of optimization processes. Since the computing capacity in the simulation is limited and virtual applications should be available as quickly as possible, the number of system evaluations should be kept small without neglecting important solutions.

It is an object of the present invention to propose an improved, in particular more reliable, method for the efficient, simulative application of automated driving functions, preferably of an automobile or motorcycle. The computational effort or the computational complexity should be reduced without having to accept significant losses in quality. Furthermore, it is an object of the present invention to provide a correspondingly configured system arrangement and to propose a computer program product with control commands which implement the method or at least partially operate the proposed system arrangement.

The object is achieved by a method with the features according to patent claim 1. Further advantageous refinements are given in the subclaims.

Accordingly, a method for efficient, simulative application of automated driving functions is proposed, comprising reading out a plurality of driving scenarios and reading out a plurality of parameter sets from a data memory, the parameter sets each providing parameters as inputs for a simulation of a driving scenario; a division of the parameters of the parameter sets into influential and non-influential parameters per parameter set in relation to a driving scenario; and identifying a reference parameter set for an input parameter set, the reference parameter set and the input parameter set generating essentially the same output in a driving scenario and the input parameter set and the reference parameter set differing in their parameters only within parameters that are not influential. A simulative application of automated driving functions relates to the identification and application of driving functions, i.e. at least partially automated controls of a motor vehicle, in certain driving scenarios. In addition, situations are simulated in advance and it is thus learned how a vehicle should behave in a certain situation. The results of such a simulation or the results of the proposed method are then applied adaptively and situation-specifically in real time in automated driving.

According to the invention, the computational complexity decreases without the quality of the results decreasing. Simply calculated simulation results are used again. Computing complexity typically refers to the fact that it is described on the basis of necessary computation steps or the required system resources are described. Computing complexity can demand a certain processor power or it can be particularly memory-intensive. A general measure of the computational complexity is the number of computation steps required.

The optimization problem that is addressed belongs to the class of "Simulation Optimization Problems". These are characterized in that a system evaluation (i.e. the evaluation of the target function depending on the input variables) is characterized by a simulation run. In order to solve these problems efficiently, a substitute model of the simulation or of reality is often generated, which has a lower model quality, but enables a fast evaluation of parameter sets. An alternative approach is the application of influence analyzes in advance of the optimization and the a priori reduction of the parameter space to influential areas.

In the area of the application of automated driving functions, it is difficult to create a substitute model, since the performance must always take into account the entire chain of effects (environment recognition, environment processing, movement planning, control, vehicle dynamics) and small model errors falsify the performance measured using comfort and safety KPIs could. Legal framework conditions also require a high model quality in order to be able to use virtual tests for application or validation. With the aid of influence analyzes, the parameter space can be reduced according to the invention before the optimization. It should be noted, however, that the evaluation of an application parameter set requires the simulation of a scenario catalog and subsequent aggregation of the individual KPIs. It is therefore possible that an influence analysis declares certain areas of the parameter space based on the aggregated KPIs to be irrelevant, even though the scenarios have an important influence on specific KPIs. The invention provides for counteracting this problem by integrating the results of a scenario-specific sensitivity analysis into an optimization algorithm and thus saving individual scenario simulations. The cumulative number of scenarios and the total computing time for the virtual application can thus be reduced according to the invention.

The term key performance indicator (KPI) or key performance indicator generally refers to key figures which can be used to measure and / or determine the progress or the degree of fulfillment with regard to objectives or critical success factors. These factors, which are known in a different context, are used, for example, to carry out the division of the parameters of the parameter sets into influential and non-influential parameters per parameter set in relation to a driving scenario. In this way it can be determined empirically which parameters of the parameter sets are classified into influential and non-influential parameters.

According to one aspect of the present invention, in a planned simulation based on a driving scenario using the input parameter set, the result of an earlier simulation using the identified reference parameter set is used. This has the advantage that results that have already been calculated can be used again for other parameter sets. Parameter sets are used as input for a simulation and if other parameters have no significant effects, a result of a parameter set that has already been calculated can be used.

According to a further aspect of the present invention, if the identification fails, a new simulation is carried out for at least one driving scenario carried out and the input parameter set is saved as a reference parameter set. This has the advantage that the library of existing calculations, i.e. results of a simulation, can be expanded. Although no savings can be seen for the time being, since the simulation has to be carried out in the absence of comparable reference parameter sets, this result is then saved and can be used in the next iteration.

According to a further aspect of the present invention, the driving scenarios are provided as simulation functions. This has the advantage that driving scenarios can be treated virtually empirically and models of individual driving scenarios are created. In this way, a real-world driving scenario can be modeled and, consequently, the parameter sets can be used to determine how a motor vehicle should behave.

According to a further aspect of the present invention, the driving scenarios describe a longitudinal maneuver, a braking, an acceleration, a transverse maneuver, a steering movement and / or a parking maneuver. This has the advantage that common driving scenarios can be mapped and modeled. In general, all driving scenarios can be mapped and a simulation can therefore be carried out using the parameters.

According to a further aspect of the present invention, the parameters of the parameter sets are divided into influential and non-influential parameters on the basis of empirical test runs, which decide on the basis of a provided threshold value for which deviation of a simulation result parameters are influential or not. This has the advantage that those parameters that have little influence on the simulation result can be hidden and do not have to be taken into account when comparing the parameter sets. This significantly increases the probability that reference parameter sets will be found, since the parameter sets only have to be similar in terms of relevant parameters. According to a further aspect of the present invention, the essentially identical output describes an output of a simulation process with regard to a driving scenario, which only deviates within the scope of a predetermined tolerance range. This has the advantage that it is possible to specify the criteria according to which outputs are the same. Essentially, this describes an optional feature which in the present case expresses that equality should also exist in the event of a deviation within a tolerance range. The feature “essentially” is thus clearly defined, although this feature is negligible or can be deleted in the present case. As a result, the output may be the same or vary within a tolerance range.

According to a further aspect of the present invention, the proposed method is carried out iteratively for all the parameter sets that have been read out and all of the driving scenarios that have been read out. This has the advantage that all the parameter sets read out and all the driving scenarios read out are combined with one another and all the results of comparable parameter sets are stored, this being done for each driving scenario. The result is consequently a database that provides a result for a subset of the parameter sets for each driving scenario. According to a further aspect of the present invention, the database contains results for a subset of the driving scenarios for a subset of the parameter sets. The parameter set-scenario combination including the result is only entered in the database if a simulation is required for the scenario under consideration. Due to the similarity of the parameter sets, each of the parameter sets ultimately finds a result. This is the case because there is a simulation result for one or more input parameter sets, which, however, was only calculated for one reference parameter set. The number of simulation runs is thus advantageously less than the number of input parameter sets, since a simulation is not carried out for each of the input parameter sets, but rather a simulation of the reference parameter set can be used in some cases.

The object is also achieved by a system arrangement for the efficient, simulative application of automated driving functions, having a Interface unit set up for reading out a plurality of driving scenarios and for reading out a plurality of parameter sets from a data memory, the parameter sets each providing parameters as inputs for a simulation of a driving scenario; a computing unit set up to split the parameters of the parameter sets into influential and non-influential parameters per parameter set in relation to a driving scenario and a comparison unit set up to identify a reference parameter set for an input parameter set, the reference parameter set and the input parameter set generating essentially the same output in a driving scenario and the parameters of the input parameter set and the reference parameter set only differ within parameters that are not influential.

The object is also achieved by a computer program product with control commands which execute the method and operate the proposed arrangement when they are executed on a computer.

According to the invention, it is particularly advantageous that the method can be used to operate the proposed devices and units or the system arrangement. The proposed devices and devices are also suitable for carrying out the method according to the invention. The device thus in each case implements structural features which are suitable for carrying out the corresponding method. The structural features can, however, also be designed as method steps. The proposed method also provides steps for implementing the function of the structural features.

Further advantages, features and details of the invention emerge from the following description, in which aspects of the invention are described in detail with reference to the drawings. The features mentioned in the claims and in the description can be essential to the invention individually or in any combination. Likewise, the features mentioned above and those detailed here can each be used individually or together several can be used in any combination. Parts or components that are functionally similar or identical are in some cases provided with the same reference symbols. The embodiments shown and described are not to be understood as conclusive, but are exemplary for explaining the invention. The detailed description is provided for the information of those skilled in the art; therefore, in the description, known circuits, structures and methods are not shown or explained in detail in order not to make the present description more difficult to understand. In the figures show:

FIG. 1: a schematic flow diagram of a method for efficient, simulative application of automated driving functions according to one aspect of the present invention; and

FIG. 2: a schematic flow diagram of a method for the efficient, simulative application of automated driving functions according to a further aspect of the present invention.

FIG. 1 shows the proposed method in a schematic flow diagram. The basic principle of the present invention and the flow of information are illustrated in FIG. Based on scenario-specific sensitivity information, each new parameter set to be simulated is checked during the optimization to determine whether a simulation of the entire scenario catalog is necessary or whether individual scenario simulations can be saved. For this purpose, already simulated parameter set-scenario combinations are sought whose parameter values differ from the current parameter set only within non-influential areas.

Parameters whose values in the current record are in influential areas may only deviate within a defined tolerance in the reference data record. With the help of this procedure, computing times for scenario-based optimization problems can be significantly reduced. Especially in the area of the application of automated driving functions, scenarios are to be expected depending on different influences of parameters (for longitudinal maneuvers, lateral maneuvers, parking maneuvers, ...) so that the application of the method promises a significant increase in efficiency. A scenario is thus, for example, a longitudinal maneuver (braking / accelerating), lateral maneuver (steering movement), a parking maneuver or the like.

The increase in efficiency achieved with this method can be seen in a representative example (simulative application of a level 3 driving function). A transferability of the procedure for the protection is also advantageously possible. It can be used to identify the most safety-critical situations and to significantly reduce the test effort required for approval (200 million km). All areas in which a system evaluation represents the simulation of a scenario catalog can use this method.

The terms “left”, “right”, “above” and “below” used in the description of the exemplary embodiments relate to the drawings in an orientation with normally legible figure designation or normally legible reference symbols. The process steps can be described as follows:

10: For each scenario s (provision of scenarios s)

11: For each parameter set P * (provision of parameter sets P *)

12: Looking for a reference parameter set Pref that deviates from P * only within non-influential areas and maintains a maximum deviation D to P * in influential areas? (maximum deviation D as tolerance range)

13: Found a valid reference pref? If no: control flow downwards, if yes: control flow to the right.

14: Simulate scenario s for parameter set P *

15: Add P * to the reference parameter set database (library will be enriched for future iterations)

16: Take the result of Pref for P * 17: All parameter sets checked? If no: control flow to the left, if yes: control flow upwards.

18: All scenarios checked? If no: control flow to the left, if yes: control flow to the right. FIG. 2 shows in a schematic flow diagram a method for the efficient, simulative application of automated driving functions, having a readout 100 of a plurality of driving scenarios and a readout 101 of a plurality of parameter sets from a data memory, the parameter sets each providing parameters as inputs for a simulation of a driving scenario ; dividing 102 the parameters of the parameter sets into influential and non-influential parameters per parameter set in relation to a driving scenario; and identifying 103 a reference parameter set for an input parameter set, the reference parameter set and the input parameter set generating essentially the same output in a driving scenario and the input parameter set and the reference parameter set differing in their parameters only within parameters that are not influential.

The person skilled in the art recognizes here that the steps can have further substeps and in particular that the method steps can each be carried out iteratively and / or in a different order.

Claims

1. Method for efficient, simulative application of automated driving functions, comprising:

Reading out (100) a plurality of driving scenarios and reading out (101) a plurality of parameter sets from a data memory, the parameter sets each providing parameters as inputs for a simulation of a driving scenario;

- Splitting (102) the parameters of the parameter sets into influential and non-influential parameters per parameter set in relation to a driving scenario; and

- Identifying (103) a reference parameter set for an input parameter set, the reference parameter set and the input parameter set generating essentially the same output in a driving scenario and the input parameter set and the reference parameter set differing in their parameters only within parameters that are not influential.

2. The method according to claim 1, characterized in that in a planned simulation based on a driving scenario by means of the input parameter set, the result of an earlier simulation by the identified (103) reference parameter set is used.

3. The method according to claim 1 or 2, characterized in that if the identification (103) fails, a new simulation is carried out for at least one driving scenario and the input parameter set is stored as a reference parameter set.

4. The method according to any one of the preceding claims, characterized in that the driving scenarios are provided as simulation functions.

5. The method according to any one of the preceding claims, characterized in that the driving scenarios describe a longitudinal maneuver, braking, acceleration, transverse maneuver, a steering movement and / or a parking maneuver.

6. The method according to any one of the preceding claims, characterized in that the division (102) of the parameters of the parameter sets into influential and non-influential parameters takes place on the basis of empirical test runs, which decide on the basis of a threshold value for which deviation of a simulation result parameters are influential or are not influential.

7. The method according to any one of the preceding claims, characterized in that the essentially the same output describes an output of a simulation process with respect to a driving scenario, which deviates only within a predetermined tolerance range.

8. The method according to any one of the preceding claims, characterized in that the proposed method is carried out iteratively for all read parameter sets and all read driving scenarios.

9. System arrangement for efficient, simulative application of automated driving functions, comprising:

an interface unit set up for reading out (100) a plurality of driving scenarios and for reading out (101) a plurality of parameter sets from a data memory, the parameter sets each providing parameters as inputs for a simulation of a driving scenario;

- A computing unit set up for dividing (102) the parameters of the parameter sets into influential and non-influential parameters per parameter set in relation to a driving scenario; and

- A comparison unit set up to identify (103) a reference parameter set for an input parameter set, wherein the reference parameter set and the input parameter set are essentially one Generate the same output in a driving scenario and the input parameter set and the reference parameter set differ in their parameters only within parameters that are not influential.

10. Com puterprogramm product with control commands which carry out the method according to any one of claims 1 to 8 when they are brought to execution on a computer.