DE102019216816A1 - Method for operating an automated vehicle - Google Patents

Abstract

A method for operating an automated vehicle (100), comprising the steps:
- Sensory detection of an environment of the automated vehicle;
- Predicting a driving behavior of the automated vehicle and another sensory-sensed road user for a defined future time segment in a simulation tree, with an interaction of the automated vehicle with the at least one other sensory-sensed road user and / or an interaction of the automated vehicle with the sensor-sensed environment without road users detected by sensors are simulated; in which
branching of the predicted driving behavior in the simulation tree is only carried out after the presence of at least one defined temporal and / or spatial circumstance of the automated vehicle or of the surroundings; and
- The automated vehicle is guided according to a defined result of the predicted driving behavior.

Description

The invention relates to a method for operating an automated vehicle. The invention also relates to a device for operating an automated vehicle. The invention also relates to a computer program. The invention also relates to a machine-readable storage medium.

State of the art

For automated driving, it makes sense to carry out a so-called behavior planning or pre-calculation of the behavior of the host vehicle and other vehicles detected by sensors (“prediction”) for a defined future time horizon. This behavior planning has the task of calculating in advance the effects of various possible behaviors (e.g. staying in lane, changing lane to the left / to the right, etc.). For this purpose, a forward simulation is usually carried out, which simulates the behavior of the host vehicle and the interaction with all other relevant road users. The effects of a behavior can then be assessed (for example, time saved by overtaking the vehicle in front) and the behavior that is best assessed can be carried out. This also results in a better foresight for the ego vehicle: it can react early to any critical situations (which result from the forward simulation) and avoid these branches.

A decision or simulation tree is usually used to implement the behavior planning mentioned. The "branches" of the decision or simulation tree are the different possible behavior patterns, whereby branches are possible at different times, for example: Branch 1 = keep lane, branch 2 = change lane to the left, branch 3 = change lane to the right.

In this case, however, the problem arises that such branches can theoretically take place at any point in time (for example, the beginning of lane change a little earlier or later). This threatens to "explode" the size of the decision tree and thus the computing time for the simulation of the branches, i.e. the complexity increases sharply.

Disclosure of the invention

It is an object of the invention to provide an improved method for operating an automated vehicle.

• - Sensorisches Erfassen eines Umfelds des automatisierten Fahrzeugs;
• - Prädizieren eines Fahrverhaltens des automatisierten Fahrzeugs und eines anderen sensorisch erfassten Verkehrsteilnehmers für einen definierten zukünftigen Zeitabschnitt in einem Simulationsbaum, wobei eine Interaktion des automatisierten Fahrzeugs mit dem wenigstens einen anderen sensorisch erfassten Verkehrsteilnehmer und/oder eine Interaktion des automatisierten Fahrzeugs mit dem sensorisch erfassten Umfeld ohne sensorisch erfassten Verkehrsteilnehmer simuliert wird; wobei
• - ein Verzweigen des prädizierten Fahrverhaltens im Simulationsbaum jeweils erst nach Vorliegen von wenigstens einem definierten zeitlichen und/oder räumlichen Umstand des automatisierten Fahrzeugs oder des Umfelds durchgeführt wird; und
• - wobei das automatisierte Fahrzeug gemäß einem definierten Ergebnis des prädizierten Fahrverhaltens geführt wird.

The object is achieved according to a first aspect with a method for operating an automated vehicle, comprising the steps:

• - Sensory detection of an environment of the automated vehicle;
• - Predicting a driving behavior of the automated vehicle and another sensory-sensed road user for a defined future time segment in a simulation tree, with an interaction of the automated vehicle with the at least one other sensory-sensed road user and / or an interaction of the automated vehicle with the sensor-sensed environment without road users detected by sensors are simulated; in which
• branching of the predicted driving behavior in the simulation tree is only carried out after the presence of at least one defined temporal and / or spatial circumstance of the automated vehicle or of the surroundings; and
• - The automated vehicle is guided according to a defined result of the predicted driving behavior.

In this way, based on available sensor information for a defined period of time, forward planning of the automated vehicle and the other road users is carried out. A branching of the forward planning in the simulation tree is only carried out if defined conditions are present, which advantageously means a reduced computational effort for calculating the simulation tree because it has only a few branches compared to conventional methods. As a result, the simulation tree is branched and set up in an event-triggered manner, and not, as in the prior art, branched in a rigid time grid, whereby the simulation effort can advantageously be kept low. In this way, a decision for a variant to be actually driven can advantageously be made earlier for the automated vehicle, which can significantly increase a safety level in a traffic scenario. As a result, a predictive, generic approach is pursued with the proposed method, which can be implemented with zero to n traffic participants detected by sensors.

Another advantage is that the behavior of the automated vehicle is predictable by specifying the exact decision events, which makes validation of the function easier or possible in the first place.

According to a second aspect, the object is achieved with a device for operating an automated vehicle which is designed to carry out the proposed method.

According to a third aspect, the object is achieved with a computer program which is designed to carry out the proposed method.

According to a fourth aspect, the object is achieved with a machine-readable storage medium on which the computer program is stored.

Advantageous further developments of the method are the subject of the dependent claims.

An advantageous development of the method provides that the driving behavior is branched in the simulation tree if the automated vehicle approaches a vehicle in front during the prediction, the speed of which is lower than the target speed of the automated vehicle and a distance threshold to the vehicle in front has been undershot. This can be used in combination with other conditions for triggering the event “lane change” during an overtaking maneuver.

Another advantageous development of the method provides, in the event that a lane change trigger distance to a vehicle in front is reached during the prediction of the driving behavior and a vehicle approaching the vehicle from behind in an adjacent lane is detected by sensors, the implementation of a lane change for the automated Vehicle in the neighboring lane is not released.

A further advantageous development of the method provides that the lane change is only released for the automated vehicle if the vehicle approaching from behind does not have to brake more than a defined threshold value. In this way, too, a safe and fluid traffic situation is supported.

A further advantageous development of the method provides that the lane change trigger distance is reached and it is detected by sensors that no vehicle in the neighboring lane is approaching from behind, and that the automated vehicle in the neighboring lane can perform a lane change. In this way, too, a safe and fluid traffic situation is supported.

Another advantageous development of the method is characterized in that the driving behavior of the automated vehicle is branched in the simulation tree when a target position of the automated vehicle has been determined in relation to a gap in a vehicle group and the automated vehicle has reached the gap and is essentially the same drives fast like the gap. In this way, the proposed method is advantageously also suitable for threading the vehicle into a moving column of vehicles.

Another advantageous development of the method is characterized in that a defined calculation rule is used to determine the position at which the automated vehicle is to enter the gap. In this way, different driving styles of the ego vehicle and of the vehicles involved, which generate the gap, can advantageously be taken into account in the proposed method.

The invention is described in detail below with further features and advantages on the basis of several figures. All of the features described or shown, individually or in any combination, form the subject of the invention, regardless of how they are summarized in the patent claims or their reference, and regardless of their wording or representation in the description or in the figures. The figures are primarily intended to clarify the principles essential to the invention and are not necessarily drawn to scale.

• 1-6 mehrere Darstellungen von Szenarien zur Durchführung des vorgeschlagenen Verfahrens; und
• 7 eine prinzipielle Darstellung einer Funktionsweise einer Ausführungsform des vorgeschlagenen Verfahrens zum Betreiben eines automatisierten Fahrzeugs.

In the figures shows:

• 1-6 several representations of scenarios for carrying out the proposed method; and
• 7th a basic illustration of a mode of operation of an embodiment of the proposed method for operating an automated vehicle.

Description of embodiments

In the following, the term automated vehicle is used synonymously in the meanings of partially automated vehicle, autonomous vehicle and partially autonomous vehicle.

A core concept of the invention is in particular to provide an improved method and a device for performing improved behavior planning or improved operation for an automated vehicle.

A method is proposed which reduces the possible points in time of branches of a predicted driving behavior carried out in a simulation tree and their number as much as possible without "missing" "good" points in time. A “good point in time” is understood to mean a point in time at which an immediate decision regarding the driving behavior of the automated vehicle is due.

A key idea here is that when the behavior planning or forward simulation (prediction) is carried out, branches in the simulation tree are only permitted if clear spatial and temporal events (branch conditions) are present. This advantageously greatly reduces the number of possible branches in the simulation tree and thus also the computing time for calculating the simulation tree in a corresponding manner. This advantageously supports the fact that, despite limited computing capacity, many (or even all currently possible or sensible) tree branches are calculated in the simulation and an optimum can be found therefrom. As a result, operating the automatic vehicle with the optimal simulation results is actually put into practice.

• - Anforderungen an funktionale Sicherheit und Validation (z.B. Mindestabstände, Geschwindigkeits- und Beschleunigungsgrenzen, technische Aktuatormöglichkeiten, usw.)
• - Gesetzliche Vorgaben, wie zum Beispiel länderspezifische Vorgaben zu Geschwindigkeiten, Abständen, Reißverschluss-Verfahren, usw.
• - Fahrstil-Vorgaben des Fahrzeugherstellers zum Beispiel, z.B. sportlich, komfortabel, Stadtverkehr, usw.
• - Fahrstil-Auswahl des Fahrzeugnutzers während der Fahrt (z.B. komfortabel, zeitoptimiert, usw.

In a variant of the proposed method, it is provided that the branching conditions are derived from a defined “driving style” (fixed rules and equations). The defined driving style makes it possible, for example, to meet various requirements mentioned below as examples:

• - Requirements for functional safety and validation (e.g. minimum distances, speed and acceleration limits, technical actuator options, etc.)
• - Legal requirements, such as country-specific requirements for speeds, distances, zipper processes, etc.
• - Driving style specifications of the vehicle manufacturer, for example, e.g. sporty, comfortable, city traffic, etc.
• - Driving style selection of the vehicle user while driving (e.g. comfortable, time-optimized, etc.

By specifying the driving style, the proposed method can advantageously make an automated driving function of the vehicle safer and more attractive.

A “branch within the simulation tree” is understood to mean that in this case a new driving behavior action is predicted, in deviation from the previously predicted driving behavior of the automated vehicle. The simulation tree is thus set up or run through quasi "in real time", i.e. very often and very quickly, with the entire simulation tree being run through or simulated in 20 ms, for example. The period of 20 ms is only to be understood as an example.

In the following, three examples are explained in more detail for the proposed method.

Example 1: Overtaking maneuver with a free passing lane

1. a) sich das Fahrzeug einem Vorderfahrzeug nähert, dessen Geschwindigkeit langsamer als die Geschwindigkeit des Fahrzeugs ist, und
2. b) eine Distanzschwelle zum Vorderfahrzeug gemäß folgender Bedingung unterschritten wurde

$${\text{d}}_{\text{SP}}=\text{f}\left({\text{v}}_{\text{F}},{\text{v}}_{\text{V}},{\text{a}}_{\text{V}}\right)$$

dSP

Spurwechsel-Triggerdistanz

vF

Geschwindigkeit des Fahrzeugs

vV

Geschwindigkeit Vorderfahrzeug

aV

Beschleunigung Vorderfahrzeug

Dabei ist f(..) eine Funktion, welche die genannten Eingangssignale (und gegebenenfalls auch noch weitere Eingangssignale), und dazu einige Parameter enthält. Entsprechend kann nach dem Überholvorgang auch ein Einscher-Abstand als Funktion vorgegeben werden.A lane change to the left for the purpose of overtaking may only be carried out, for example, if:

1. a) the vehicle approaches a vehicle in front whose speed is slower than the speed of the vehicle, and
2. b) a distance threshold to the vehicle in front was not reached in accordance with the following condition

$${\text{d}}_{\text{SP}}=\text{f}\left({\text{v}}_{\text{F.}},{\text{v}}_{\text{V}},{\text{a}}_{\text{V}}\right)$$

dSP

Lane change trigger distance

vF

Vehicle speed

vV

Front vehicle speed

aV

Acceleration of the vehicle in front

Here, f (..) is a function which contains the input signals mentioned (and possibly also further input signals) and a number of parameters. Accordingly, a cut-in distance can also be specified as a function after the overtaking maneuver.

Furthermore, a lane change to the right for the purpose of exiting the freeway, for example, may only be carried out from the moment when the vehicle has reached a freeway exit on which it is to depart, further conditions having to be met.

1 shows an automated vehicle 100 (Ego vehicle) in the case of a simulated and, after the simulation, also practically carried out overtaking of a vehicle in front 101 . The automated vehicle 100 and the vehicle in front 101 both drive on one lane or street 1 . Thus shows 1 a situation in the future that will then be controlled by the automated vehicle 100 is realized in terms of driving technology in the present according to the result of the previous simulation. If the simulation carried out provides a critical result, this result will be in the presence of the automated vehicle 100 thus not implemented in practice. The automated vehicle 100 is equipped with a powerful environment sensor system (not shown), for example in the form of ultrasound, radar, lidar, mono or stereo camera, lane recognition device, etc.

At least one of the sensors mentioned is arranged in the front and rear areas in order to be on the road 1 to be able to reliably detect at least one road user coming from behind using sensors. The driving directions of the vehicles 100 , 101 are indicated with arrows. The Associated behavioral actions are as follows: the automated vehicle 100 is initially in the "Drive in lane, with or without vehicle in front" action until the desired lane change trigger distance dsp to the vehicle in front _{is reached at the decision position EP 1} 101 has reached.

Here a branching of the forward simulation or prediction is carried out in the simulation tree (not shown) and a lane change to the left is also simulated. Once in the fast lane, the vehicle changes 100 back to the "Drive in lane, with or without vehicle in front" action. After the overtaking process, the automated vehicle arrives 100 at the decision-making position EP ₂ the desired cut-in distance d _ES to the vehicle in front 101 and branches here in the simulation tree to a lane change to the right, which is then actually carried out.

Example 2: Overtaking maneuver with a blocked overtaking lane

This scenario is in the 2 to 4th shown. This is where the automated vehicle arrives 100 in the decision position EP ₁ , in which a branch to a lane change is allowed, the lane change trigger distance dsp (overtaking distance) to the vehicle in front 101 . However, a further necessary branching condition results from the forward simulation carried out, that the lane change or overtaking process for the automated vehicle 100 in this case is not possible because another vehicle 102 comes quickly from behind in the fast lane (compare 3 ).

This is for the automated vehicle 100 It is not possible to carry out the lane change because either a further branching condition is not met in the simulation tree, or the _{forward simulation carried out at decision position EP 1} shows that an overtaking maneuver in a collision of the vehicle 100 with the vehicle 102 would end. The forward simulation continued to be carried out at the decision position EP ₁ with straight-ahead travel, on the other hand, shows that the automated vehicle does not collide 100 is possible, which is why the vehicle 100 even more to the vehicle in front 101 approximates.

The automated vehicle 100 As a consequence, it remains in the "Drive in lane, with or without vehicle in front" function, reduces its speed and approaches the vehicle in front 101 up to the predetermined following distance d _F , which is determined by a known ACC distance controller of the automated vehicle 100 is set.

The above-mentioned overtaking maneuver would be possible in the scenario of 3 however, if this causes the vehicle 102 does not have to brake more than a predetermined threshold value (not shown). This will make the vehicle 102 saved from this, due to the automated vehicle approaching an overtaking maneuver 100 Braking excessively hard (or even braking hard) and thereby risking an accident. This is made possible by sensing the speed of the vehicle 102 by a rear-facing sensor system (not shown) of the automated vehicle 100 .

In the other case it is based on the detected speed of the vehicle 102 decided that no branching to a forward simulation of the lane change should be carried out in the simulation tree because of the difference in speed between the vehicles 100 , 102 is too high so the vehicle 102 when changing lanes of the vehicle 100 could no longer brake in time.

4th shows that in the decision-making position EP2 the fast lane is free, which is why in the decision-making position EP2 a forward simulation is carried out in the simulation tree. An overtaking arrow indicates that the vehicle in front is now overtaking 101 What is simulated is what is possible collision-free according to the simulation and then in practice by the automated vehicle 100 is actually carried out.

Example 3: threading

A complex driving maneuver for threading the automated vehicle 100 in a gap is shown schematically in the 5 and 6th shown. In 5 is the automated vehicle 100 on a drive-in lane 2 (e.g. motorway entrance), and intended to be in on the street 1 threading flowing traffic. There are three gaps as an example for the threading maneuver S1 , S2 , S3 between vehicles 103-106 possible.

For every gap S1 , S2 , S3 becomes for the automated vehicle 100 A target position is calculated using a corresponding driving style specification or function.

1. (i) Anfahren der Sollposition mit Anpassung der Geschwindigkeit an die Lücke
2. (ii) Beim Erreichen der Sollposition an der Entscheidungsposition EP₃ erfolgt ein Einfahren des automatisierten Fahrzeugs 100 in die Lücke S2

Threading the automated vehicle 100 happens in two steps:

1. (i) Approaching the target position with adaptation of the speed to the gap
2. (ii) When the target position is reached at the decision position EP ₃ , the automated vehicle is retracted 100 into the gap S2

For each of the gaps S1 , S2 , S3 a corresponding forward simulation can be carried out in the simulation tree in order to determine the for the automated vehicle 100 To find the “most suitable” gap for the threading maneuver. If no gap is possible, the automated vehicle reduces 100 in the "Drive in lane, with or without vehicle in front" function, the speed on the threading lane 2 , and if necessary stops at the end.

The driving style specification or calculation specification can be used, for example, to specify the area of the gap S2 the vehicle 100 into the gap S2 should retract (e.g. at 30%, 50%, 70%, etc. of the gap).

In 6th a forward simulation is carried out at decision position EP3 in the simulation tree, from which it follows that the vehicle is now merging 100 into the gap S2 is possible. This is then subsequently carried out by the automated vehicle 100 also implemented in practice.

• - Abstand des automatisierten Fahrzeugs 100 zu anderen Fahrzeugen
• - Differenzgeschwindigkeiten zwischen Fahrzeugen (relativ zum Ego-Fahrzeug und zu anderen Fahrzeugen
• - absolute Geschwindigkeiten der Fahrzeuge (Ego-Fahrzeug und andere Fahrzeuge)
• - absolute oder relative Position auf einer Karte (zum Beispiel 100m vor einer Ausfahrt)
• - Position in der Lücke zwischen Fahrzeugen
• - Geschwindigkeit einer Lücke (Fahrgeschwindigkeit der beiden Fahrzeuge)
• - Öffnungs-/Schließgeschwindigkeit einer Lücke (Relativgeschwindigkeit der Fahrzeuge)
• - Breite einer Spur (Verengung, Split, Vereinigung, usw.)
• - Typ einer Fahrspur (z.B. Ein-/Ausfädelspur, Standstreifen, usw.).
• - Geschwindigkeitsbegrenzung einer Fahrspur
• - Wetterbedingungen (z.B. Sichtweite, Seitenwind, Fahrbahn-Reibwert, usw.).
• - Energievorrat (Füllstand Benzintank, Akku-Ladestand, usw.)

According to a general principle, it is understood that possible spatial and temporal conditions for branching to alternative behavior within a forward simulation are almost infinite. In the following, therefore, possible decision-making conditions are only listed as examples, with combinations of the options mentioned also being possible.

• - Distance of the automated vehicle 100 to other vehicles
• - Differential speeds between vehicles (relative to the ego vehicle and to other vehicles
• - absolute speeds of the vehicles (ego vehicle and other vehicles)
• - absolute or relative position on a map (for example 100m before an exit)
• - Position in the gap between vehicles
• - Speed of a gap (driving speed of the two vehicles)
• - Opening / closing speed of a gap (relative speed of the vehicles)
• - Width of a track (narrowing, split, union, etc.)
• - Type of a lane (e.g. entry / exit lane, hard shoulder, etc.).
• - Limiting the speed of a lane
• - Weather conditions (e.g. visibility, cross wind, road friction coefficient, etc.).
• - Energy reserve (fill level of the petrol tank, battery charge level, etc.)

The decision conditions mentioned can be implemented in the program and evaluated in each forward cycle of the simulation tree. If defined conditions are met, a branch is carried out in the simulation tree in the manner explained above.

If no forward simulation is used, the same conditions can also be used for a decision for or against a behavior in the present (scalability down to systems without any forward simulation).

The best driving strategy is then determined from the fully simulated simulation tree, which is then used by the automated vehicle 100 is also implemented in practice.

7th shows a basic sequence of a method for the optimized operation of an automated vehicle 100 .

In one step 200 a sensor-based detection of an environment of the automated vehicle takes place 100 .

In one step 210 there is a prediction of a driving behavior of the automated vehicle 100 and another road user detected by sensors for a defined future time segment in a simulation tree, with an interaction of the automated vehicle 100 with the at least one other road user detected by sensors and / or an interaction of the automated vehicle 100 is simulated with the sensor-recorded environment without sensor-recorded road users, whereby a branching of the predicted driving behavior in the simulation tree only occurs after at least one defined temporal and / or spatial circumstance of the automated vehicle is present 100 or the environment.

In one step 220 becomes the automated vehicle 100 guided according to a defined result of the predicted driving behavior.

As a result, the proposed method can advantageously increase a safety level in road traffic and support a homogeneous flow of traffic.

The method according to the invention can advantageously be implemented as software that, for example, runs on an electronic device in the form of a control unit of the automated vehicle 100 expires. A simple adaptability of the method is supported in this way.

Although the proposed method was previously described exclusively on the basis of right-hand traffic, it goes without saying that the proposed method can also be used in left-hand traffic, in which case the previously mentioned direction / orientation information must be adapted accordingly.

The person skilled in the art will modify the features of the invention in a suitable manner and / or combine them with one another without departing from the essence of the invention.

Claims (11)

A method for operating an automated vehicle (100), comprising the steps: - Sensory detection of an environment of the automated vehicle (100); - Predicting a driving behavior of the automated vehicle (100) and of another sensory registered road user for a defined future time segment in a simulation tree, wherein an interaction of the automated vehicle (100) with the at least one other sensory registered road user and / or an interaction of the automated vehicle (100) is simulated with the sensory recorded environment without sensory recorded road users; in which - a branching of the predicted driving behavior in the simulation tree is only carried out after at least one defined temporal and / or spatial circumstance of the automated vehicle (100) or of the surroundings is present; and - The automated vehicle (100) being guided according to a defined result of the predicted driving behavior. Procedure according to Claim 1 , wherein the driving behavior is branched in the simulation tree when the automated vehicle (100) approaches a vehicle in front during the prediction, the speed of which is lower than the target speed of the automated vehicle (100) and a distance threshold to the vehicle in front (101) is not reached has been. Procedure according to Claim 1 or 2 In the event that, during the prediction of the driving behavior, a lane change trigger distance (d _SP ) to a vehicle in front is reached and a vehicle (102) approaching the vehicle (100) from behind in an adjacent lane is detected by sensors, performing a lane change is not released for the automated vehicle (100) in the neighboring lane. Proceed yourselves Claim 3 wherein the lane change for the automated vehicle (100) is only released if the vehicle (102) approaching from behind does not have to brake more than a defined threshold value. Procedure according to Claim 3 or 4th In the event that the lane change trigger distance (d _SP ) has been reached and it is detected by sensors that no vehicle in the neighboring lane is approaching from behind, the automated vehicle (100) in the neighboring lane can be enabled to change lanes. Procedure according to Claim 1 , wherein the driving behavior of the automated vehicle (100) is branched in the simulation tree when a target position of the automated vehicle (100) in relation to a gap in a vehicle group (103 ... 106) has been determined and the automated vehicle (100) the gap has reached and drives essentially as fast as the gap. Procedure according to Claim 6 , wherein a defined calculation rule is used to determine the position at which the automated vehicle (100) is to enter the gap. Device which is designed, the method according to one of the Claims 1 to 7th to operate an automated vehicle (100). Use of a method according to one of the Claims 1 to 7th in the operation of an automated vehicle (100). Computer program with program code means for carrying out the method according to one of the Claims 1 to 7th when running on an electronic device (400). Machine-readable storage medium on which the computer program is based Claim 10 is stored.

Images