Method for laying out a vehicle
The method optimizes autonomous vehicle design by calculating permissible stay areas and adjusting vehicle parameters to navigate simulated scenarios safely, addressing sensor and performance discrepancies.
Patent Information
- Application Number
- DE102018006265
- Authority / Receiving Office
- DE · DE
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2018-08-08
- Publication Date
- 2024-11-28
- Estimated Expiration
- 2038-08-08
AI Technical Summary
Existing methods for optimizing autonomous vehicle design using computer simulations face challenges due to discrepancies between the sensor perspectives and technical performance of the vehicle being simulated and the vehicle used to record the scenarios, leading to inaccurate decision-making and trajectory differences.
A method involving selecting a traffic scenario from a database, calculating permissible stay areas, and adjusting vehicle parameters using a kinematic model to ensure the vehicle can navigate the scenario without incidents, while adapting steering commands and considering parameters like engine power, braking deceleration, and chassis properties.
Enables effective optimization of vehicle design by ensuring the vehicle can safely navigate simulated scenarios without accidents, addressing sensor perspective and performance discrepancies.
Smart Images

Figure 00000006_0000 
Figure 00000006_0001 
Figure 00000007_0000
Abstract
Description
[0001] The present invention relates to a method for designing a vehicle.
[0002] For an autonomous vehicle, in particular, to be approved for road use, it requires considerable effort to demonstrate that the vehicle can handle practical traffic scenarios without accidents. Due to time and cost constraints, such a comprehensive demonstration cannot be empirically performed; instead, it must be based primarily on computer simulations.
[0003] Such a verification method based on computer simulation is known, for example, from DE 10 2006 044 086 A1, in which traffic scenarios are simulated by means of a virtually modelled vehicle with an environmental sensor system and with at least one other virtually modelled road user.
[0004] For a computer controlling the vehicle, it makes no difference in principle whether the vehicle is being simulated or controlled in real life, as long as the environmental information it receives in the simulation is realistic enough. For this purpose, databases exist that compile data from real traffic scenarios, which can be used as environmental information and vehicle data in such simulations.
[0005] DE 10 2016 220 913 A1 discloses a database-supported method for automatically generating traffic scenarios for testing autonomous vehicles. This method uses data from a database collected from real vehicles on public roads. Data analysis identifies representative driving situations relevant for testing corresponding vehicles.
[0006] These scenarios and data are specific to the vehicle with which they were recorded in several respects.
[0007] For example, DE 10 2017 200 180 A1 describes a method for validating autonomous vehicle functions in which data can be recorded during test drives so that critical traffic situations can be checked afterwards.
[0008] Firstly, the data recorded in the scenario depends on the sensor equipment of the recording vehicle. This means that if the sensor equipment includes a camera that can monitor the surroundings of the recording vehicle, the images provided by such a camera and the conclusions that the vehicle draws from the images, e.g. regarding the distance and speed of other road users, depend, among other things, on the installation location of the camera on the vehicle. If, therefore, an autonomous vehicle to be optimized is supplied with image data of such a scenario as part of a simulation and analyses this data based on its own camera installation position, this can lead to incorrect conclusions about the vehicle's surroundings and, consequently, to inappropriate decisions by the vehicle's autonomous control system that are not due to errors in the programs underlying the control system.
[0009] On the other hand, the technical performance data of the vehicle to be optimized usually differ from those of the vehicle with which the scenario was recorded, with the result that the trajectory chosen by the vehicle to be optimized to master the traffic situation of the scenario will differ from the trajectory actually driven in the recorded scenario, but these differences do not indicate an error in the control programs of the vehicle to be optimized.
[0010] An object of the present invention is to enable the optimization of a vehicle design using scenarios from a database despite the problems described above.
[0011] According to an embodiment of the invention, the object is achieved by a method for optimising a vehicle to be designed, comprising the steps: a) selecting a traffic scenario from a database containing data from at least one trajectory and data on an environment recorded specifically for a real vehicle (1); b) defining a sequence of permissible locations for the vehicle to be designed by calculating, for each location from a sequence of locations of the actual vehicle along the trajectory, a permissible location in which the vehicle to be designed can stay without incident; c) searching, on the basis of a value of at least one parameter of the vehicle to be designed, a trajectory of the vehicle to be designed which runs along its entire length through the sequence of permitted locations, wherein the trajectory for the vehicle to be designed is calculated using a kinematic model by adapting steering commands and wherein the value of the at least one parameter is taken into account by the kinematic model; d) Assessing the vehicle to be designed as suitable for the selected traffic scenario if the search in step c) is successful.
[0012] In the method, step c) is repeated using different values of the at least one parameter. If several trajectories through the permissible travel areas are found, all of the underlying parameter values are obviously suitable for mastering the scenario. To select one of these multiple parameter values, a cost function of the at least one parameter is defined. The ideal parameter value with respect to this cost function is implemented in the actual vehicle to be designed.
[0013] An incident in the above sense can be any type of undesirable behavior of the vehicle, but at least a collision with another road user or a stationary object and / or crossing a lane boundary and / or violations of any traffic rule or route prohibition.
[0014] In addition to the trajectory of the vehicle from which the data forming the scenario was recorded, the scenario may contain further data on the environment of this vehicle, such as the trajectories of other road users, information on the course of the road, route restrictions and / or prohibitions, and stationary objects.
[0015] By defining the sequence of stopping areas, the goal of optimization cannot be to achieve the greatest possible match between the trajectory of the selected traffic scenario and the trajectory selected for the vehicle being designed, but rather simply that the vehicle being optimized must master the scenario without being involved in an accident. This goal is achieved once a trajectory is found that the vehicle is capable of driving, taking into account its performance parameters, and on which it does not leave the stopping areas.
[0016] If the search is unsuccessful, step c) can be repeated using a modified value of the at least one parameter to finally find a value for which a trajectory passing through the lounges exists.
[0017] The parameters that can be varied here can include, for example, engine power or braking deceleration of the vehicle or parameters of its chassis such as slip stiffness, self-steering gradient or coefficient of adhesion or parameters of a system algorithm.
[0018] If the examination of a trajectory shows that it does not run entirely within the occupied space, it may be necessary to decide whether the vehicle to be designed reaches the limit of the permitted occupied space in the direction of travel, transversely to the direction of travel, against the direction of travel or transversely to the direction of travel.
[0019] The vehicle is unsuitable for the scenario under consideration, i.e., unable to complete the scenario without incident, if it reaches the limit of the permitted area in the direction of travel or perpendicular to the direction of travel. This scenario is equivalent to striking or approaching an object next to or in front of the vehicle being tested, such as a vehicle driving ahead or alongside it, or a lane marking, and can occur particularly if, in order to complete the scenario at all until the limit is reached, the vehicle must accelerate to a speed that is no longer safely controllable.
[0020] Conversely, it may also be the case that the vehicle being interpreted reaches the edge of the permissible waiting area in the opposite direction of travel. This requires that the rear edge of the permissible waiting area is defined by a following vehicle and that this vehicle is traveling faster than the vehicle being interpreted. It should be noted here that the vehicle with which the scenario was originally recorded will generally have followed a different trajectory than that assumed for the vehicle being interpreted, and that the trajectory of the following vehicle was coordinated with that of the receiving vehicle, but did not take into account the trajectory selected for the vehicle being interpreted. In other words: if the receiving vehicle had followed the trajectory now examined for the vehicle being interpreted at the time the scenario was recorded, the following vehicle would have taken this into account and avoided a collision.The fact that the vehicle to be designed reaches the limit of the permissible area in the opposite direction of travel does not allow the conclusion that something would have to be changed in the control algorithms or the performance parameters of the vehicle to be designed in order for the scenario to be driven without collision, but at most that the scenario in question is not relevant for the optimization of the vehicle to be designed.
[0021] Further features and advantages of the invention will become apparent from the following description of exemplary embodiments with reference to the accompanying drawings. They show: Fig. 1 an initial state and Fig. 2 an end state of a scenario; and Fig. 3 a flowchart of an optimization procedure
[0022] Fig. Figure 1 shows a traffic situation at a motorway entrance ramp as a basic application example of the invention. A vehicle 1 is located in a merging lane 2 and must merge from there into flowing traffic in an adjacent lane 3 containing vehicles 4, 5, and 6.
[0023] To do this, it must reach the speed of the flowing traffic in merging lane 2 and then merge into a gap between vehicles 4, 5, 6.
[0024] A scenario describing the movements of vehicles 1, 4, 5, 6 is present in the form of a sequence of data recorded by one or more environmental sensors of vehicle 1 during acceleration and merging. A camera 7 is particularly suitable as an environmental sensor; the sequence of recorded data can then be present, for example, as a video. The scenario can be obtained by simulation or by capturing environmental data during a driving maneuver of real vehicles and is stored in a database from which it is downloaded for use in optimizing an autonomous vehicle.
[0025] To prepare the optimization, it may be necessary (according to step S1 of the flow chart of Fig. 3) adapt the recorded data of the downloaded scenario to the sensors of the autonomous vehicle. If the installation position of the camera 7 on the autonomous vehicle differs from that of the real vehicle 1, then correct control of the autonomous vehicle by its on-board computer 9 would not be guaranteed if the original video recordings of the stored scenario were used as input data. Since these recordings show the objects in the environment of the vehicle 1 from different perspectives over time and as the vehicle 1 moves, it is possible to first convert the original video recordings into a sequence of three-dimensional models of the environment and, from these, to derive video recordings that show the environment from the perspective of the camera 7 of the autonomous vehicle and can be used as input data for its on-board computer 9.
[0026] Each of these three-dimensional models contains, among other things, the current coordinates of vehicles 4, 5, 6 and their dimensions for a particular point in time in the scenario. Based on this data, a permissible location area for the autonomous vehicle is calculated (S2) for the respective point in time, in which it can fit without reaching the other vehicles 4, 5, 6 and the boundaries of the roadway. Such a permissible location area 8 includes, at a given point in time in the scenario, the current location of vehicle 1 as well as locations in the environment to which vehicle 1 could be moved without touching the other vehicles 4, 5, 6 at their current locations or other objects. The permissible location area includes, for example, as in Fig. 1, Fig. 2, at the beginning of the scenario the merging lane 2 in front of vehicle 1 and towards the end the space between vehicles 5, 6 in lane 3.
[0027] The on-board computer 9 of the autonomous vehicle is programmed using any known technology to generate a sequence of steering commands for actuators that control, for example, the steering, fuel metering, or the stroke of a brake cylinder of the autonomous vehicle, such that they guide the autonomous vehicle from the merging lane 2 into lane 3 without collision. During the simulation, these steering commands are not received by real components of the vehicle, but rather by a kinematic model, also implemented by a computer, which calculates the movement of the autonomous vehicle resulting from the steering commands, taking into account vehicle-specific parameters such as engine power or braking deceleration of the vehicle, slip stiffness, self-steering gradient, or coefficient of adhesion.
[0028] Within the scope of the method, a realistic value is determined for each of the parameters mentioned (S3), and based on these values, the on-board computer 9 determines a sequence of steering commands (S4) that guide the autonomous vehicle into lane 3. If the trajectory of the autonomous vehicle, determined by the kinematic model in step S5 based on these steering commands, ends between vehicles 5, 6, then a check is performed (S6) to determine whether the trajectory runs within the permissible operating areas 8 over its entire length. If so, then the vehicle is capable of handling the scenario with the currently selected set of parameter values. The method could then be aborted at this point, and the autonomous vehicle could be built with the selected parameter values.
[0029] If the trajectory does not lie within the permissible storage areas 8 for its entire length, an attempt is made to find a trajectory that lies within the permissible storage areas 8 for its entire length by adjusting the steering commands (S7). If this attempt (S8) fails (even after possibly several attempts), then an investigation is carried out (S9) to determine where the autonomous vehicle reaches the boundary of the permissible storage area. If it reaches the front boundary of the permissible storage area, this corresponds to a rear-end collision with the vehicle 6 in front. If this cannot be avoided, then the autonomous vehicle is clearly not capable of handling the scenario under investigation with the currently selected set of parameter values.
[0030] In this case, a new set of parameter values is selected (S10), and the method returns to step S4. In the case of approaching the front limit of the permissible operating area, the change in the parameter values may, in particular, involve increasing the braking deceleration, i.e., using more powerful brakes.
[0031] If the vehicle hits the lateral limits of the permissible area, this also constitutes an accident and requires the selection (S10) of a new set of parameter values, whereby the parameter values changed in this case preferably concern the chassis or tires, such as cornering stiffness, self-steering gradient or adhesion coefficient of the tire).
[0032] If there is sufficient certainty that an impact with the front or lateral limits cannot be prevented - especially after changing the parameter values several times - then it can be concluded that the autonomous vehicle is technically not capable of driving through the scenario without an accident (S11).
[0033] If, however, the autonomous vehicle reaches the rear boundary of the permissible area, this is not due to a driving error on the part of the autonomous vehicle, but simply means that the speed of the following vehicle 5 is predetermined in the scenario and cannot account for the movement of the autonomous vehicle. In this case, a meaningful statement about whether the autonomous vehicle can complete the scenario without an accident is not possible, since the question of whether an accident occurs depends on the behavior of the following vehicle 5 and cannot be directly influenced by the autonomous vehicle. In this case, the scenario is irrelevant to the autonomous vehicle (S14), and a different scenario should be used to optimize the autonomous vehicle.
[0034] Even in the case that in S5 the currently selected parameter values have proven to be suitable for driving the scenario, a new set of parameter values can be selected (S10) and the process can be repeated from step S4, in which case the change in the parameter values can be aimed at minimizing costs.
[0035] In the course of the method, through iterations of steps S4 to S10, one or more sets of parameter values can be found that allow driving a trajectory that runs entirely within the permissible residence areas 8. If a cost function is stored for the parameters that allows quantifying manufacturing costs or another variable to be optimized for a given selection of parameter values, then in step S13, the one among the multiple sets that achieves the best value of the cost function is selected, and the autonomous vehicle is built with the parameter values selected in this way (S14).
[0036] Scenarios generated during the process that differ from the one originally downloaded can be uploaded to the database for further use.
[0037] It should be understood that while the above detailed description and the drawings illustrate certain exemplary embodiments of the invention, they are intended for illustrative purposes only and should not be construed as limiting the scope of the invention. Various modifications to the described embodiments are possible without departing from the scope of the following claims and their range of equivalence. In particular, features of the embodiments not recited in the claims will become apparent from this description and the figures. Such features may also appear in combinations other than those specifically disclosed herein.The fact that several such features are mentioned in the same sentence or in another type of textual context does not justify the conclusion that they can only occur in the specifically disclosed combination; instead, it must generally be assumed that individual features of several such features can also be omitted or modified, provided that this does not call into question the functionality of the invention. List of reference symbols 1 vehicle 2 merging lanes 3 lanes 4 vehicle 5 vehicle 6 vehicles 7 Camera 8 Lounge 9 on-board computers
Claims
[1] Method for optimising a vehicle to be designed with the following steps: a) selecting a traffic scenario from a database containing data from at least one trajectory and data on an environment recorded specifically for a real vehicle (1); b) defining a sequence of permissible locations (8) for the vehicle to be designed by calculating, for each location from a sequence of locations of the real vehicle (1) along the trajectory, a permissible location (8) in which the vehicle to be designed can stay without another vehicle (4, 5, 6) at its current location and without reaching the limits of a roadway; c) searching (S5), on the basis of a value of at least one parameter of the vehicle to be designed, a trajectory of the vehicle to be designed which runs along its entire length through the sequence of permissible residence spaces (8), wherein the trajectory for the vehicle to be designed is calculated using a kinematic model by adapting steering commands and wherein the value of the at least one parameter is taken into account by the kinematic model; d) Evaluating the vehicle to be designed as suitable for the selected traffic scenario if the search of step c) is successful, wherein in the method step c) is repeated on the basis of different values of the at least one parameter in order to find a plurality of trajectories, the values used in step c) are evaluated using a cost function (S13), and wherein the values of the at least one parameter that are optimal with regard to the cost function are implemented in the real vehicle to be designed (S14). [2] Method according to claim 1, comprising the step e) if the search is unsuccessful, repeating step c) using a modified value of the at least one parameter. [3] Method according to claim 2, comprising the step f) if the search is unsuccessful, decide (S9) whether the vehicle to be laid out reaches the limit of the permitted area in the direction of travel, transversely to the direction of travel or against the direction of travel. [4] Method according to claim 3, comprising the step (g) assessing the vehicle to be designed as unsuitable for the scenario (S11) if it reaches the limit of the permitted space (8) in the direction of travel or transversely to the direction of travel. [5] Method according to claim 3 or 4, comprising the step h) Assess (S12) the scenario as not relevant if the vehicle to be designed reaches the limit of the permitted space (8) in the opposite direction of travel. [6] Method according to one of the preceding claims, wherein the at least one parameter is selected from dimensions of the vehicle; position and performance parameters of a sensor monitoring the environment of the vehicle, slip stiffness, coefficient of adhesion of a tire; self-steering gradient.
Citation Information
Patent Citations
Automatic braking and steering system for a vehicle
DE10036276A1
Traffic situation e.g. accident-critical driving situation, simulating system, has reference vehicle, where behavior of vehicle temporally takes place based on information detected by detection unit during simulated traffic situation
DE102006044086A1
Method and device for generating test cases for autonomous vehicles
DE102016220913A1
Method and test unit for predicting the movement of road users during a passively operated vehicle function
DE102017200180A1
