DE102017211556A1 - Method for route planning for a motor vehicle with an automated vehicle system and motor vehicle with an automated vehicle system - Google Patents

Abstract

In order to provide a method for route planning for a motor vehicle with an automated vehicle system, which takes into account the route-specific localizability and the performance of the automated vehicle system and thus enables situation-adapted route planning, a method for route planning for a motor vehicle with an automated vehicle system is proposed, wherein a driveability analysis at least a route is created, wherein the trafficability analysis, the steps selection of a route to be examined from a starting point to a destination point, - determination of vehicle-independent environmental conditions on the selected route, - determining an expected detection rate for locating suitable landmarks on the selected route using the vehicle-independent ambient conditions, - to determine an expected localization accuracy on the selected route using the expected Detection rate, - Determining whether the expected localization accuracy for a supported by the automated vehicle system driving on the selected route is sufficient includes.

Description

The present invention relates to a route planning method for a motor vehicle having an automated vehicle system. Furthermore, the present invention relates to a motor vehicle with an automated vehicle system, wherein the vehicle system is designed for carrying out a method for route planning.

State of the art

Advanced driver assistance systems (ADAS) and highly automated vehicle systems for UAD (Urban Automated Driving) increasingly require detailed knowledge of the vehicle environment and situational awareness. Sensor measurement data is used as the basis for the perception of the vehicle environment. From these, objects can be extracted with the aid of so-called detection algorithms, by means of which the vehicle environment can be described and analyzed. Modern environment sensors, such as stereo video cameras or laser scanners, in conjunction with detection algorithms, allow the collection of much information from the vehicle environment, for example in the form of landmarks. Such landmarks include traffic signs, traffic lights, lane markings, etc. Detected objects or detected landmarks, in turn, may be used to perform vehicle localization. The performance of the entire automated vehicle system depends significantly on the performance of environmental sensors.

Modern lane departure warning systems rely on reliable detection and vehicle-relative location of the lane markings. Out Borrmann, JM et al., "STELLaR - A Case-Study on Systematically Embedding Traffic Light Recognition", Intelligent Transportation Systems (ITSC), 2014 IEEE, 17th International Conference on, pp. 1258, 1265, 8.-11. October 2014 , a traffic light detection is known, taking into account the very high hardware requirements for object detection algorithms.

An approach to the selection of landmarks for vehicle localization is in Thrun, S., "Finding Landmarks for Mobile Robot Navigation," Robotics & Automation, 1998. Proceedings. 1998 IEEEE International Conference on, Volume 2, pp. 958, 963, 16.-20. May 1998 , described.

The accuracy of the vehicle localization required for the automated vehicle systems does not only depend on the performance of the environmental sensor system. In addition, environmental conditions and the choice of the detection algorithm can have a significant impact. The performance of the entire vehicle system is directly linked to the route to be traveled and the given environmental conditions.

Disclosure of the invention

The present invention has for its object to provide a method for route planning for a motor vehicle with an automated vehicle system, which takes into account the route-specific localizability and the performance of the automated vehicle system and thus allows a situation-adapted route planning.

It is another object of the invention to provide a motor vehicle with an automated vehicle system, which provides a situation-adapted route planning.

• - Wahl einer zu untersuchenden Route von einem Startpunkt zu einem Zielpunkt,
• - Ermittlung von fahrzeugunabhängigen Umfeldbedingungen auf der gewählten Route,
• - Bestimmung einer zu erwartenden Detektionsrate für zur Fahrzeuglokalisierung geeignete Landmarken auf der gewählten Route unter Verwendung der fahrzeugunabhängigen Umfeldbedingungen,
• - Bestimmung einer zu erwartenden Lokalisierungsgenauigkeit auf der gewählten Route unter Verwendung der zu erwartenden Detektionsrate,
• - Ermittlung, ob die zu erwartende Lokalisierungsgenauigkeit für ein von dem automatisierten Fahrzeugsystem unterstütztes Befahren der gewählten Route ausreichend ist,

umfasst.To solve the problem, a method for route planning for a motor vehicle having an automated vehicle system is proposed, wherein a trafficability analysis of at least one route is created, wherein the trafficability analysis comprises the steps:

• Choosing a route to be examined from a starting point to a destination point,
• Determination of vehicle-independent ambient conditions on the selected route,
• Determination of an expected detection rate for landmarks suitable for vehicle localization on the selected route using the vehicle-independent environmental conditions,
• Determination of an expected localization accuracy on the selected route using the expected detection rate,
• Determining whether the localization accuracy to be expected is sufficient for driving on the selected route, supported by the automated vehicle system,

includes.

The automated vehicle system may be a driver assistance system for assisted driving and / or highly automated driving and / or autonomous driving.

Advantageously, in the method for route planning, a sufficiently precise route-specific localizability and the performance of the overall system, in particular of the automated vehicle system, are taken into account. Routes along which no requirement-oriented localization is possible under the given environmental conditions and with the surroundings sensor system available to the vehicle system can be classified as impracticable. Such routes or route sections can then be bypassed. The method uses vehicle-independent ambient conditions along the route to be examined as input variables. The landmarks suitable for vehicle localization can be taken from a localization map.

The method allows the planning of routes on which the requirement-oriented performance of the vehicle system is ensured with high probability, thereby contributing to an advantageous influence on the robustness of the automated vehicle system. In contrast to known methods for route planning for automated vehicle systems, vehicle-independent environmental conditions, that is to say environmental conditions which do not relate to the vehicle and / or the automated vehicle system, are additionally used. However, vehicle-dependent environmental conditions can also be included in the calculation of the expected localization accuracy. Vehicle-dependent environmental conditions can be, for example, the sensor efficiency or the robustness of the landmark detection algorithms.

The environmental sensors may be stereo video cameras or laser scanners or other suitable sensor devices.

The step of determining an expected detection rate for landmarks suitable for vehicle localization does not necessarily have to be carried out in a separate method step in the method, but can also be incorporated implicitly into the determination of an expected localization accuracy. It is essential in the method for route planning that the determination of an expected localization accuracy and a subsequent determination as to whether the expected localization accuracy is sufficient for a driving of the selected route supported by the automated vehicle system is carried out using vehicle-independent ambient conditions on the selected route , For example, it is also possible that the determination of the localization accuracy to be expected or the determination of whether the localization accuracy to be expected is sufficient for driving the selected route supported by the automated vehicle system by means of a parameterized model and / or a neural network or further methods of the machine learning and / or a statistical model. In the case of a parameterized model, the detection rate in the form of a parameter can be explicitly or implicitly included in the method. If the localization accuracy is determined by means of a neural network or the like, then the expected detection rate is implicitly taken into account in the weights learned in the neural network. In the case of the statistical model too, the detection rate via the statistical weights is included in the determination of the localization accuracy. For the understanding of the invention it is important that the localization accuracy implicitly or explicitly depends on the product of the detection rate, that is, on the percentage of landmarks detectable on a selected route, and the number of landmarks present on the selected route.

The detection rate implicit or explicit in the method may also depend on the type of landmarks present on the selected route. For example, the detection rate for traffic signs, lane boundaries, light signals, trees or buildings may be different. In addition, the detection rate may also depend on the efficiency of the sensor device of the environment sensor. Also, the detection algorithms used may have a different efficiency for the detection rate of certain types of landmarks. All of the aforementioned factors can be used individually or jointly via an explicit or implicit consideration of the detection rate input into the determination of an expected localization accuracy and the consequent determination as to whether the localization accuracy to be expected is sufficient for driving the selected route supported by the automated driving system.

It is preferably provided that the vehicle-independent ambient conditions are the traffic density and / or the weather conditions and / or the road conditions for the selected route.

The weather conditions, the traffic volume and the traffic density as well as road conditions can have an influence on the detection performance of the sensors and thus on the detection rate for different types of landmarks. In particular, a high traffic density can lead to a certain proportion of landmarks on the selected route being at least temporarily obscured by vehicles in front of the environmental sensor system, which reduces the detection rate and the localization accuracy. Advantageously, the dependency of the localization accuracy on the current weather conditions and the traffic volume is taken into account. The determination of the vehicle-independent environmental conditions, in particular the weather conditions and the traffic volume, can, for example, by polling Weather databases or traffic databases are made.

It is preferably provided that a concealment rate for landmarks on the selected route is determined, wherein the concealment rate is preferably determined using the environmental conditions and / or the type of suitable landmarks, wherein the localization accuracy and / or the detection rate is determined using the concealment rate.

The occlusion rate advantageously reflects the influence of the route-dependent and vehicle-independent environmental conditions on the localization accuracy. The occlusion rate may also depend on the type of landmarks. Thus, landmarks located at a low height above the road surface are more likely to be obscured by increased traffic and vehicle volume for the environmental sensor system, which consequently reduces the detection rate for such landmarks. In contrast, the occlusion rate is lower for higher altitude landmarks, such as light signals.

The occlusion rate can also be taken into account implicitly or explicitly in the procedure. In parameterized models or neural networks, an implicit consideration can take place via the parameters or the learned weights.

More preferably, provision can be made for the occlusion rate and / or the detection rate and / or the localization accuracy to be determined using a parameterized model, wherein the parameterized model is preferably a machine-learned model, wherein the machine-learned model is further preferred using previous trafficability analyzes , in particular, using previous environmental conditions and / or previously determined detection rates and / or previously determined occlusion rates and / or prior determined localization accuracies and / or the performance of sensor devices and / or detection algorithms of the Automated Vehicle System.

Advantageously, the occlusion rate and / or the detection rate and / or the localization accuracy can be determined using a parameterized model or using a neural network. The parameterized model or the neural network determines the occlusion rate and / or the detection rate and / or the localization accuracy on the basis of previously conducted trafficability analyzes, that is to say using the results of trafficability analyzes which were carried out prior to the current performance of the method.

Accordingly, it can be provided that, before carrying out the trafficability analysis, in particular before the choice of a route to be examined is carried out, the parameterized model or the neural network undergoes a learning phase.

In a preferred embodiment, the occlusion rate is determined using the parameterized model using as input to the parameterized model the current weather information and / or the current traffic on the selected route and / or the types of landmarks present on the selected route.

The system is based on a model parameterized with machine teaching methods, in particular on a neural network, which establishes the relationship between weather and traffic data as well as other information about the environment, the environment perception modules used or environment sensors on the vehicle side and the resulting detection probability of individual landmark types.

It is further preferred for the detection rate to be determined using the occlusion rate and / or a maximum number and / or the number density and / or the type of landmarks detectable on the selected route and / or the performance of sensor devices and / or detection algorithms of the automated vehicle system and / or the environmental conditions and / or the accuracy of localization data, in particular GPS data, is determined.

The occlusion rate is used to determine the type-dependent fraction of landmarks that are likely to be obscured on the selected route by environmental conditions such as increased traffic. The rate of detection then depends on the occlusion rate and, moreover, on the efficiency of the sensors or sensor devices. Furthermore, the detection rate can also depend on the efficiency of the detection algorithms, wherein different detection algorithms, in particular for different types of landmarks, can be used.

Further preferably, it can be provided that the localization accuracy is determined by means of a statistical model, wherein the localization accuracy preferably by using the expected detection rate and / or the expected occlusion rate and / or the number and / or the number density and / or the type of landmarks along the selected route.

The localization accuracy depends explicitly or implicitly on the expected detection rate and, in particular type-dependent, number or number density of the landmarks along the selected route. The detection rate can be influenced by the occlusion rate.

Further advantageously, it can be provided that the determined localization accuracy is compared with a predetermined threshold value, whereby a driver of the motor vehicle enables the possibility of manually driving the selected route and / or a new trafficability analysis is carried out for another route if the localization accuracy is less than the threshold is.

In the method, the driver may be informed when falling below the threshold for the localization accuracy that the selected route from the vehicle using the automated vehicle system is not passable and then a new route can be planned.

Alternatively or additionally, the driver of the motor vehicle can be offered the opportunity to decide whether he wants to manually drive the originally planned or selected route or would like to accept an automated, possibly longer, journey.

Advantageously, global planning and rescheduling of an overall route for an automated vehicle system to a defined destination point becomes possible.

It can preferably be provided that the determination of whether the expected localization accuracy is sufficient for a driving of the selected route supported by the automated vehicle system takes into account the best possible localization and / or optimizes this determination to the shortest possible route with sufficiently accurate localization accuracy becomes.

It can also be provided that the sensor systems and / or detector algorithms suitable for driving the selected route are supported by the automated vehicle system, preferably if the localization accuracy is greater than or equal to the threshold value.

Advantageously, the method determines which sensor devices are suitable for which type of landmarks for driving on the selected route. The system is based, in particular using a parameterized model, on the establishment of the relationship between weather data, traffic data and other information about the environment, the sensor devices or environment perception modules on the vehicle side and the resulting detection probability of individual landmark types. Together with the information about the density of landmarks, which can be taken for example from a localization map and the detectable landmark types, which are given depending on the localization module of the individual vehicle system, can be determined by the method, whether a route can be currently traveled by the vehicle system , Different methods of detection algorithms on the vehicle side are taken into account by the method, that is to say, depending on the existing sensor device or depending on the existing detection algorithms, individual vehicle systems may possibly travel in areas where others can not. For this purpose, it can be advantageously determined which sensor devices and / or which detection algorithms are used for driving on a route, in particular at which route section of the route.

The method uses weather data, traffic information, accuracy of GPS, and information about the density of detectable landmarks along a route as an input. A model whose parameters have been determined by a machine learning method is used in order to be able to make a statement for the selected route as to whether and, if so, with which detection algorithms the route is passable.

Another solution to the problem underlying the invention is to provide a motor vehicle with an automated vehicle system, wherein the automated vehicle system is designed to carry out the method described above.

Further preferably, it can be provided that the automated vehicle system is designed for assisted driving and / or highly automated driving and / or autonomous driving.

list of figures

• 1 ein Ablaufdiagramm eines Verfahrens zur Routenplanung für ein Kraftfahrzeug mit einem automatisierten Fahrzeugsystem,
• 2 eine Übersicht über die Eingangsgrößen für eine Lernphase eines parametrisierten Modells.

The invention is explained below with reference to the drawings. Show it

• 1 a flowchart of a method for route planning for a motor vehicle with an automated vehicle system,
• 2 an overview of the input variables for a learning phase of a parameterized model.

Detailed description of the figures

1 shows a flowchart of a method for route planning for a motor vehicle with an automated vehicle system.

In a first, optional step S1 is in a learning phase for a parameterized model, in particular for a neural network, based on information about weather data, detection rates of landmarks of various types and other information sources such as traffic data, a relationship between these inputs and expected concealment rates for individual landmark types, preferably at the respective given position, made. The position can be determined eg via GPS, dead reckoning or localization on the vehicle.

In the case of use, the procedure begins with the trafficability analysis in step S2 , In step S2 Depending on a starting point and a destination point, a route to be examined for driving with the automated vehicle system, which may include a navigation system, is selected.

This is done in step S3 the determination of vehicle-independent ambient conditions on the selected route. These environmental conditions may be the current traffic volume or the current weather conditions on the selected route, and are determined by querying a database. Furthermore, the expected landmarks are taken from a localization map.

In the following step S4 the parameterized model is used to determine a current occlusion rate for each desired landmark type on the route to be examined. In this case, the current traffic volume and / or the current weather conditions and possibly also the landmarks stored in a localization map for the selected route are used as input variables.

In step S5 For example, a detection rate for landmarks suitable for vehicle localization is made on the selected route using the vehicle-independent environmental conditions. In this case, the determination of the detection rate takes place in the step S4 determined occlusion rate input. Furthermore, vehicle-dependent conditions, such as the detection efficiency of the different sensor devices, such as stereo cameras or laser systems, as well as the performance of the possible detection algorithms for different types of landmarks, especially in the given environmental conditions (weather, traffic, ...), can be considered.

In step S6 an expected localization accuracy is then performed on the selected route using the expected detection rate. In other words, the combination of the determined occlusion rate, the detection rate and the number, the number density and the type of landmarks present along a route allows an estimation of the localization accuracy by means of a statistical model.

The localization accuracy for the selected route results, in particular, from the product of the detection rate and the number density of the landmarks present on the selected route or on sections of the selected route for a specific landmark type.

In a further step S7 it is determined whether the expected localization accuracy is sufficient for driving on the selected route supported by the automated vehicle system. On the basis of the trafficability analysis, a decision is therefore made on the necessity of re-planning the route. Should such be necessary, this can be carried out according to the procedure, whereby the procedure with a new selected route again with the trafficability analysis in the step S2 starts.

If the localization accuracy is sufficient for a driving of the selected route supported by the automated vehicle system, then in a further step S8 in addition, the sensor devices and detector algorithms to be used by the vehicle system can be indicated by the method for a route which is considered as passable. Furthermore, a plan to switch detector algorithms along the selected route can be made to respond to changing landmark types.

By way of example, a route is planned for an automated vehicle system along a main road. Enough surrounding information is available along this route, so that a sufficiently accurate localization of the vehicle system is possible by the detection of all landmarks of the "traffic sign" type.

However, heavy traffic is expected during peak hours on this route. An automated vehicle system implementing the method determines a detection rate for this traffic situation and the planned route, which is clearly below 100% due to the expected occultation rate, which is based on vehicle-independent environmental conditions. Nevertheless, the system determines as the expected localization accuracy a value which is sufficiently good, and in particular lies above a predetermined threshold value. The system thus enables driving of the selected route supported by the automated vehicle system.

In another case, it can come at the same time of day to a strong precipitation. The method again determines a detection rate. Due to the bad weather conditions, this is not sufficiently high, which is due in particular to an increased concealment rate due to the bad weather conditions. A sufficiently accurate localization accuracy is therefore not possible. In the context of the method, an alternative route can then be selected and evaluated within the framework of a further trafficability analysis. The new route leads through less traveled roads and is therefore associated with a lower rate of obstruction. The driver can optionally be informed via a man-machine interface that a longer route has been planned and now a choice between automatic operation on this route or a manual drive on a shorter route are given.

2 shows an overview of the input variables for a learning phase of the parameterized model, in particular the neural network.

• GPS-Daten 10,
• Verkehrsdaten 11,
• Wetterdaten 12 und
• frühere Detektionsraten 13

Be in the learning phase

• GPS data 10 .
• traffic data 11 .
• Weather data 12 and
• earlier detection rates 13

as input parameter into a neural network 14 given. The training phase results in a model 15 for the type-dependent probability of landmark detection under different environmental conditions.

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited non-patent literature

• Borrmann, J.M. et al., "STELLaR - A Case-Study on Systematically Embedding Traffic Light Recognition," Intelligent Transportation Systems (ITSC), 2014 IEEE, 17th International Conference on, pp. 1258, 1265, 8.-11. October 2014 [0003]
• Thrun, S., "Finding Landmarks for Mobile Robot Navigation," Robotics & Automation, 1998. Proceedings. 1998 IEEEE International Conference on, Volume 2, pp. 958, 963, 16.-20. May 1998 [0004]

Claims (10)

A method for route planning for a motor vehicle with an automated vehicle system, wherein a drivability analysis of at least one route is created, wherein the drivability analysis the steps Choosing a route to be examined from a starting point to a destination point, Determination of vehicle-independent ambient conditions on the selected route, Determination of an expected detection rate for landmarks suitable for vehicle localization on the selected route using the vehicle-independent environmental conditions, Determination of an expected localization accuracy on the selected route using the expected detection rate, Determining whether the localization accuracy to be expected is sufficient for a driving on the selected route supported by the automated vehicle system. Method according to Claim 1 , wherein the vehicle-independent environmental conditions are the traffic density and / or the weather conditions and / or the road conditions for the selected route. Method according to Claim 1 or 2 in which a masking rate for landmarks is determined on the selected route, the masking rate being preferably determined using the environmental conditions and / or the type of suitable landmarks, the localization accuracy and / or the detection rate being determined using the masking rate. Method according to Claim 3 wherein the occlusion rate and / or the detection rate and / or the location accuracy is determined using a parameterized model, wherein the parameterized model is preferably a machine-learned model, wherein more preferably the machine-learned model using previous trafficability analyzes, in particular using former ambient conditions and / or previous determined detection rates and / or previous determined occlusion rates and / or previous determined localization accuracies and / or the performance of sensor devices and / or the detection algorithms of the automated vehicle system. Method according to one of the preceding claims, wherein the detection rate using the occlusion rate and / or one, in particular maximum, number and / or number density and / or the type of detectable on the selected route landmarks and / or the performance of sensor devices and / or Detection algorithms of the automated vehicle system and / or the environmental conditions and / or the accuracy of localization data, in particular GPS data, is determined. Method according to one of the preceding claims, wherein the localization accuracy is determined by means of a statistical model, the localization accuracy preferably using the expected detection rate and / or the expected occlusion rate and / or the number and / or the number density and / or the type Landmarks along the chosen route. Method according to one of the preceding claims, wherein the determined localization accuracy is compared with a predetermined threshold, allowing a driver of the motor vehicle the possibility of manually driving the selected route and / or a new trafficability analysis is carried out for another route, if the localization accuracy is smaller as the threshold is. Method according to one of the preceding claims, wherein for driving the selected route supported by the automated vehicle system suitable sensor devices and / or detector algorithms are determined, preferably when the localization accuracy is greater than or equal to the threshold value. Motor vehicle with an automated vehicle system according to one of the preceding claims. Motor vehicle after Claim 9 The driver assistant system is designed for assisted driving and / or highly automated driving and / or autonomous driving.

Images