DE102021213429A1 - Method of providing a fully autonomous evasive maneuver control system for a ego vehicle - Google Patents

Abstract

A method for providing a fully autonomous evasive maneuver control system for a ego vehicle traveling on a road is proposed. In a first part, an environment of the ego vehicle is recognized and detected in order to create a digital environment. Furthermore, additional information is extracted from the digital environment to predict the behavior of other road users identified in the digital environment. Position and map information is also recorded in order to provide information about the availability of one or more potential alternative areas in the vicinity of the ego vehicle, the potential alternative areas not being limited to the edges of the road. In a second part, a risk of collision with one or more other road users is estimated and evaluated based on the predicted behavior of the road users detected in the digital environment. If a collision of the ego vehicle with one or more other road users is predicted, position and map information from the first part is used to identify a target avoidance field and at least one target trajectory is calculated in order to reach the target avoidance field. An autonomous maneuver of the ego vehicle is determined based on the trajectory, and subsystems of the ego vehicle are prepared for performing the autonomous maneuver. In a third part, the target trajectory is received from the second part and the motion actuators of the ego vehicle are controlled to control the longitudinal, lateral and vertical vehicle dynamics so that the avoidance field is reached autonomously.

Description

This application relates to a method for providing a fully autonomous evasive maneuver control system for a host vehicle.

Modern driver assistance systems (Advanced Driver Assistance Systems, ADAS) and autonomous driving (Autonomous Driving, AD) are not only leading automotive technologies, but can also be life-saving systems. A state-of-the-art assistance system is the well-known Evasive Maneuver Assist (EMA), which avoids accidents with objects and obstacles by the vehicle detecting obstacles locally and spontaneously, and via intelligent algorithms that help the driver after starting the maneuver of steering initiated.

As the name EMA suggests, the system (i.e. partially autonomous) actively assists the driver by applying a specific torque to the steering wheel/column to compensate for the driver's input in a way that avoids an accident.

There is currently no evasive maneuver control system available that can act fully autonomously, i. H. without any user intervention. Thus, it is an object of this invention to provide a fully autonomous evasive maneuver control system for a ego vehicle. This object is achieved by the characteristics of the independent patent claims. Advantageous embodiments are the subject matter of the dependent claims.

A method for providing a fully autonomous evasive maneuver control system for a ego vehicle traveling on a road is proposed. In a first part, an environment of the ego vehicle is recognized and detected in order to create a digital environment. Furthermore, additional information is extracted from the digital environment to predict the behavior of other road users identified in the digital environment.

Position and map information is also recorded in order to provide information about the availability of one or more potential alternative areas in the vicinity of the ego vehicle, the potential alternative areas not being limited to the edges of the road. In a second part, a risk of collision with one or more other road users is estimated and evaluated based on the predicted behavior of the road users detected in the digital environment. If a collision of the ego vehicle with one or more other road users is predicted, position and map information from the first part is used to identify a target avoidance field and at least one target trajectory is calculated in order to reach the target avoidance field. An autonomous maneuver of the ego vehicle is determined based on the trajectory, and subsystems of the ego vehicle are prepared for performing the autonomous maneuver. In a third part, the target trajectory is received from the second part and the motion actuators of the ego vehicle are controlled to control the longitudinal, lateral and vertical vehicle dynamics such that the target avoidance field is reached autonomously.

With the proposed fully autonomous evasive maneuver control system, an improved driver assistance system is provided, which has an enlarged evasive range compared to known systems.

According to one aspect, in the first part, additional information from the V2X communication is collected and used to create the digital environment. Using V2X communication (V2X=Vehicle to Everything, which also includes other vehicles, i.e. V2V communication) enables the ego vehicle to significantly increase the amount of information collected. Thus, a more reliable digital environment can be created.

According to one aspect, available avoidance fields are empty lanes and/or fallow fields and/or areas with a certain probability of a collision with another object, e.g. B. vehicles, pedestrians, walls and other objects that could harm the ego vehicle, especially its passengers.

In one aspect, if a collision of the ego vehicle with one or more other road users is predicted, information is provided to at least one passenger of the ego vehicle in the form of a warning signal and/or information related to the action the ego vehicle is taking is executed.

According to one aspect, if a collision of the ego vehicle with one or more other road users is predicted, information is provided to one or more road users in the form of a warning signal and/or information in which direction the ego vehicle will move.

In one aspect, other road users are vehicles and/or pedestrians. Thus the method is not limited to avoiding a collision with vehicles.

In one aspect, neural networks are used to predict the behavior of other road users identified in the digital environment.

Furthermore, a computer program with software is proposed to carry out all the steps of the method when run on a computer.

Furthermore, an autonomous evasive maneuver control system of an ego vehicle is proposed, wherein the system is adapted to carry out the method. According to one aspect, the steps of the method are performed using at least one of the control units of the ego vehicle and/or using at least one separately provided control unit.

Furthermore, an ego vehicle is proposed, comprising: at least one set of sensors adapted to monitor the surroundings of the ego vehicle and actuators adapted to control the lateral and longitudinal movement of the ego vehicle, and at least one control unit adapted to calculate the steps of the method and capable of communicating with the set of sensors to receive sampled data and with the actuators to activate them fully autonomously based on the calculated action to be performed .

• Die 1 stellt ein Schema bereit, das das Verfahren zum Bereitstellen eines autonomen Ausweichmanöver-Steuersystems für ein Ego-Fahrzeug gemäß einer Ausführungsform der Erfindung veranschaulicht.
• Die 2 stellt ein Schema bereit, das das Verfahren zum Bereitstellen eines autonomen Ausweichmanöver-Steuersystems für ein Ego-Fahrzeug gemäß einer anderen Ausführungsform der Erfindung veranschaulicht.
• Die 3 stellt ein Diagramm bereit, das Schritte des Verfahrens gemäß einer Ausführungsform der Erfindung veranschaulicht.

Embodiments of this disclosure are illustrated by way of example and not limitation in the figures of the accompanying drawings in which like reference numbers indicate similar elements and wherein:

• The 1 provides a schematic illustrating the method for providing an autonomous evasive maneuver control system for a host vehicle according to an embodiment of the invention.
• The 2 provides a schematic illustrating the method for providing an autonomous evasive maneuver control system for a ego vehicle according to another embodiment of the invention.
• The 3 provides a diagram illustrating steps of the method according to an embodiment of the invention.

It should be noted that in the figures the same reference numbers denote the same features.

The main achievement of the present invention is to provide a fully autonomous evasive maneuver control system, i. H. Level 5 that can fully control the movement of the ego vehicle 1 i.e. H. in lateral, longitudinal and vertical directions. The evasive maneuver to be performed by ego vehicle 1 is also not limited to the edges of lanes L1, L2 and/or the road. The solution as described below provides an autonomous active safety feature in cases where emergency braking is not possible for any reason.

The fully autonomous evasive maneuver control system is implemented as a software product (computer program) executable by a computer. This software product executes a method using hardware of the ego vehicle 1 and software components and information derived from the outside of the ego vehicle 1 . The goal is to calculate an action to be performed by the ego vehicle 1 in an actual situation, decide what to do, and act fully autonomously. In addition, a warning about an upcoming action to be carried out as information to the passengers of the ego vehicle 1, z. as a simple warning signal, and/or to one or more other road users 2-4 affected by the action. However, the action, i. H. the movement itself, the passengers of the ego vehicle 1 and/or one or more other road users 2-4 affected by the action are provided.

In general, the method comprises three parts, the first part being the SEE part, the second part being the THINK part, and the third part being the DOING part, as in FIG 3 is shown schematically. The three parts are explained in more detail with reference to the 1 and 2 embodiments shown described.

In all embodiments, the fully autonomous evasive maneuver control system, hereinafter referred to as EMC, can communicate with the sensor sets of the ego vehicle 1, which are used to monitor the surroundings of the car. Such sensors can be (the list is not limited): cameras (front, side, rear...), radars (front, side, rear...), lidars (front, side, rear...), ultrasonic sensors ( front, side, back...). Also, the EMC can communicate with actuators of the ego vehicle 1 and use map data information provided by the ego vehicle 1 . It can also be made possible to communicate with external devices such as servers (e.g. for traffic, map information), other vehicles, etc. (OTA, Over the Air).

In the first part, SEE, the system uses not only the vehicle's sensors Sets of the ego vehicle 1 to detect the environment of the ego vehicle 1 and to create the digital environment, but also uses z. B. deep neural networks to extract further characteristics and to predict the behavior of other (not fully autonomous driving) road users 2 - 4, e.g. B. whether a driver in oncoming traffic intends a risky overtaking maneuver or not.

The intelligent VIEW part also uses position and map information derived from the ego vehicle 1 . The ego vehicle 1 can also set up V2V and V2X communication, e.g. B. with external servers (“cloud”) and vehicles to receive useful information, such as position and map information, which allows the SEE part to significantly increase the amount of information received, which allows creating a more reliable digital environment . The most important purpose of using this information is to provide the second part THINKING with the availability of avoidance fields in the vicinity of the ego vehicle 1 to which it can safely autonomously steer in order to avoid an accident. These alternative fields 100 can be both an area on the street and also off the street, e.g. B. a corn or fallow field on the side of the road.

In the second part THINKING, the system uses deep neural networks (AI, artificial intelligence, artificial intelligence) to extract further characteristics and to predict the behavior of different road users 2 - 4. Only then can all the complex information be processed easily, and an action can be calculated based on the actual situation and the situation in the near future. As for example in the 1 is shown: If it is predicted that a driver of a motorcycle 2 will overtake a truck 3 in oncoming traffic on a road with only two lanes L1, L2. If the rider of motorcycle 2 (who is overtaking) underestimates the distance required, the maneuver is risky and potentially fatal. In this case, the AI algorithms used in the ego vehicle 1 can already estimate and evaluate the risk of a collision and prepare all subsystems of the ego vehicle 1 in advance to perform the autonomous maneuver, if necessary. This involves calculating the alternate target trajectory T, which could include off-road points, e.g. B. on an empty field on the side of the ego vehicle 1, as in the 1 will be shown.

In the third part ACTION, the alternative target trajectory T is received from the THINK part. This part receives all the information needed to calculate an action to be taken based on the available information. The responsibility of the ACTION part then lies in controlling the motion actuators of the ego vehicle 1 to control the longitudinal, lateral and vertical vehicle dynamics (to ensure stability) in order to execute the target trajectory T and to achieve the desired target avoidance field 100 .

The 1 12 shows an embodiment that includes an evasive maneuver as an action, in which the ego vehicle 1 evades into a fallow area 100 as a target evasion area, which is located adjacent to its lane L2.

In order for it to be able to carry out this action, a digital environment is created in the first part SEEING based on the recognition and detection of an environment of the ego vehicle 1 . In this case, oncoming traffic in the form of a truck 3 and a motorcycle 2 is detected in lane L1. Also, additional information is extracted from the digital environment by using neural networks to predict the behavior of other road users 2 - 4 that have been detected in the digital environment (by the sensors of the ego vehicle 1 sensor set or by other sensors, communicating with the ego vehicle 1). In addition, position and map information is recorded in order to provide information about the availability of one or more potential alternative areas in the vicinity of the ego vehicle 1, with the potential alternative areas not being restricted to the edges of the road. The position and map information can be derived from the ego vehicle 1 and/or from external entities.

In the second part THINK, it is estimated and evaluated whether, if the motorcycle 2 overtakes the truck 3 (predicted behavior in the SEE part), a collision between the ego vehicle 1 and the motorcycle 2 would be risked or not. A target alternative field 100 is also selected from the potential alternative fields, as SEE in the first part from the sensor set of the ego vehicle 1 and additional information, e.g. B. derived from maps and/or from other vehicles if available. The avoidance fields, including the target avoidance field 100, are not limited to the edges of the road or the road lanes L1, L2, but can also be a fallow field or any free space that is accessible to the ego vehicle 1 without collision. Furthermore, at least one alternative target trajectory T (indicated by the arrow between P1 and P2) is calculated from the current position P1 of the ego vehicle 1 to reach the target alternative field, ie the position P2 of the ego vehicle 1. In order to react as quickly as possible, is also an autonomous maneuver of the ego vehicle 1 is determined, and subsystems of the ego vehicle 1 are prepared for performing the autonomous maneuver, if necessary.

Finally, in the third part ACTION, the target trajectory T is received from the second part THINK, and the motion actuators of the ego vehicle 1 are controlled in order to control the longitudinal, lateral and vertical vehicle dynamics, so that the target avoidance field 100 is reached fully autonomously. The ACTION part is only active if a collision z. B. is predicted with a predefined probability.

The 2 12 shows another embodiment that includes an evasive maneuver as an action in an intersection scenario, in which the ego vehicle 1 evades into a fallow area 100 as a target avoidance area that is not adjacent to its lane L2. In this case, oncoming traffic in the form of a truck 3 and a motorcycle 2 is detected in lane L1. In addition, pedestrians 4 crossing the road at right angles to lane L1 are detected. The first part SEE is carried out as above in relation to the 1 is described and will therefore not be repeated.

In the second part THINKING, a collision risk with one or more other road users 2-4 is estimated and evaluated on the basis of the predicted behavior of the other road users 2-4 that have been recognized in the digital environment. In this case, the ego vehicle 1 is to drive straight ahead in its current lane L2 (indicated by the arrow 10). Further, it is predicted that the truck 3 will not avoid the ego vehicle 1 but will cross its lane L2 (indicated by the arrow 30). The motorcycle 2 running behind the truck 3 is predicted to go straight, i. H. remains in its lane L1 (indicated by arrow 20). The pedestrians 4 are predicted to cross the road. Since the truck 3 is predicted to cross the lane L2 of the ego vehicle 1, a collision risk is determined. A target alternative field 100 is thus identified and selected from the potential alternative fields, as SEE in the first part by the sensor set of the ego vehicle 1 and based on additional information, e.g. B. derived from maps and/or from other vehicles if available. In this case, the alternative field 100 beyond the lane L1 of the truck 3 is determined to be available. Also in this embodiment, the avoidance fields including the target avoidance field 100 are not limited to edges of the road or road lanes L1, L2, but can also be a fallow field or any free space accessible to the ego vehicle 1 without collision.

Thus, the EMC calculates a trajectory T for reaching the target alternate zone 100, i. H. for moving from its current position P1 to the new position P2 without hitting other road users, in this case the motorcycle 2 and the pedestrians 4. In order to react as quickly as possible, an autonomous maneuver of the ego vehicle 1 is also determined, and Subsystems of the ego vehicle 1 are prepared for performing the autonomous maneuver, if necessary.

Finally, in the third part ACTION, the target trajectory T is received from the second part THINK, and the motion actuators of the ego vehicle 1 are controlled in order to control the longitudinal, lateral and vertical vehicle dynamics, so that the target avoidance field 100 is reached fully autonomously.

If a collision of the ego vehicle 1 with one or more other road users 2 - 4 is predicted, in all embodiments information can be sent to at least one passenger of the ego vehicle 1 and/or one or more road users 2 - 4 in the form of a warning signal, such as such as a visual, acoustic, etc. signal, and information related to the action that the ego vehicle 1 will perform, e.g. B. in which direction the ego vehicle 1 will move can be discontinued.

In all embodiments, the available avoidance fields 100 are empty lanes and/or fallow fields and/or areas with a certain probability of a collision with another object. Preferably, the determined probability of collision with another object is zero. However, if a collision cannot be avoided, a decision is made as to which of the other road users 2-4 the ego vehicle 1 will collide with. This decision is not the subject of the present invention, but it must be made when designing the system, particularly the THINK part, and may be subject to legal and/or moral requirements.

Additional information can also be derived by communicating with external resources that have useful information for calculating the action to be taken, e.g. B. Cloud Services. This opens up the possibility of having access to other useful information, such as traffic conditions, construction sites, maps, etc. Also, vehicle-to-vehicle (V2V) communication can be used as additional information to share information about road conditions and road users, etc .

According to the invention there is also provided an ego vehicle 1 comprising: at least one set of sensors adapted to monitor the surroundings of the ego vehicle and actuators adapted to control the lateral and longitudinal movement of the ego vehicle Vehicle 1 to control, and at least one control unit adapted to calculate the steps of the method and capable of communicating with the set of sensors for receiving sampled data and with the actuators to operate them fully autonomously on the basis of the activate the calculated action to be performed.

The proposed method can be used in all types of vehicles, in particular in passenger cars as well as in trucks and motorcycles.

The prerequisite is that the vehicle is able to act fully autonomously, at least when performing the evasive manoeuvre. An advantage is that the evasive maneuver is not limited to edges of the road. This opens up a much larger avoidance area for avoiding a collision between road users.

Reference List

1

ego vehicle

2 - 4

other road users (trucks 2, motorcycles 3, pedestrians 4)

100

Target Dodge

10

path of the ego vehicle

20

path of the motorcycle

30

path of the truck

P1, P2

Position of the ego vehicle

L1, L2

road tracks

T

target trajectory

Claims (10)

A method of providing a fully autonomous evasive maneuver control system for an ego vehicle (1) traveling on a road, wherein - in a first part (SEE) - recognition and detection of an environment of the ego vehicle (1) is carried out in order to create a digital environment, and - additional information is extracted from the digital environment to predict the behavior of other road users (2-4) detected in the digital environment, and - Position and map information is recorded in order to provide information about the availability of one or more potential alternative fields in the vicinity of the ego vehicle (1), the potential alternative fields not being limited to the edges of the road, - in a second part (THINKING) a risk of a collision with one or more other road users (2 - 4) is estimated and evaluated on the basis of the predicted behavior of the other road users (2 - 4) recognized in the digital environment and, in case a collision of the ego vehicle (1) with one or more other road users (2-4) is predicted, the following actions are taken: - Position and map information from the first part (SEE) is used to identify a target alternate field (100), and - at least one target trajectory (T) for reaching the target avoidance field (100) is calculated, and based on the trajectory (T) an autonomous maneuver of the ego vehicle (1) is determined, and subsystems of the ego vehicle (1) are prepared to perform the autonomous maneuver, - in a third part (ACT) - the target trajectory (T) is received from the second part (THINKING) and the movement actuators of the ego vehicle (1) are controlled in order to control the longitudinal, lateral and vertical vehicle dynamics, so that the target avoidance field (100) autonomously is reached. procedure after claim 1 , where in the first part (SEE) additional information from the V2X communication is collected and used to create the digital environment. Method according to one of the preceding claims, wherein the available avoidance fields are empty lanes and/or fallow fields and/or areas with a certain probability of a collision with another object. Method according to one of the preceding claims, wherein, if a collision of the ego vehicle (1) with one or more other road users (2 - 4) is predicted, information is sent to at least one passenger of the ego vehicle (1) and/or one or several road users (2 - 4) in the form of a warning signal and/or information relating to the action that the ego vehicle (1) is going to carry out. Method according to any one of the preceding claims, wherein other road users (2-4) are vehicles and/or pedestrians. A method according to any one of the preceding claims, wherein neural networks are used to predict the behavior of other road users (2-4) identified in the digital environment. Computer program with software for performing all steps of the method according to one of the preceding claims when run on a computer. Autonomous evasive maneuver control system for a ego vehicle (1), the system being adapted to implement the method according to any one of Claims 1 until 6 to execute. Autonomous evasive maneuver control system according to claim 8 , wherein the steps of the method are carried out by using at least one of the control units of the ego vehicle (1) and/or by using at least one separately provided control unit. Ego vehicle (1) comprising - at least one set of sensors adapted to monitor the surroundings of the ego vehicle (1), and - actuators adapted to control the lateral and longitudinal movement of the ego vehicle (1) to control, and - at least one control unit adapted to carry out the steps of the method according to any one of Claims 1 until 6 and capable of communicating with the set of sensors to receive sampled data and the actuators to activate them fully autonomously based on the calculated action to be performed.

Images