DE102019127176A1 - Controlling an autonomous vehicle - Google Patents

Abstract

A method is provided for controlling an autonomously driving vehicle at low speed by means of a model for interaction with at least one unprotected road user, with trajectories for future movements of the unprotected road user being calculated, control commands being planned on this basis and the planned control commands of the autonomous vehicle in an information loop into the calculations of the trajectories for future movements of the unprotected road user before revised control commands are sent to the actuators of the autonomous vehicle. An autonomous vehicle is also provided for carrying out the method.

Description

The invention relates to a method for controlling an autonomous vehicle by means of a model for the interaction between the vehicle and unprotected road users.

In the area of parking spaces, unprotected road users, i.e. those who are not inside a protective body, e.g. pedestrians and / or cyclists, are exposed to a particular risk. This is especially true in the case of autonomous vehicles. For an autonomous vehicle to function safely, it is of central importance to recognize unprotected road users and to anticipate their likely movement. The interaction between an autonomous vehicle and an unprotected road user is essential, since an action by the vehicle can influence the unprotected road user and vice versa. Autonomous vehicles are here motor vehicles with a drive by an internal combustion engine, an electric motor and / or a hybrid drive, which can drive without the influence of a human driver.

In road traffic, the interaction between the autonomous vehicle and unprotected road users or the privileges of one of the road users is regulated by traffic signs. The situation in parking areas is different, however, as the priority of road users is often unclear here. This makes the interaction difficult, e.g. when an autonomous vehicle drives into a parking space. In the event that an unprotected road user wants to give the autonomous vehicle the right of way and stops, this cannot be recognized by a simple control strategy, so that the vehicle stops. The unprotected road user could then assume that the autonomous vehicle has stopped to give it the right of way.

In conventional control strategies, a future movement of the unprotected road user is calculated as a trajectory that is independent of the actions of the corresponding autonomous vehicle. As a rule, the most unfavorable behavior of the unprotected road user is assumed for the trajectory. Accordingly, the control strategy of the autonomous vehicle is very conservative in the sense that a collision with unprotected road users should be avoided.

A previously calculated future movement of a pedestrian is, for example, in the publication CN 108172025 A described, this information being used in a driver assistance system to warn the driver.

In the pamphlet DE 10 2017 210 394 B3 pedestrians are detected in a parking area and their future movement is predicted. Input information comes from a camera, vehicle-to-vehicle communication or a remote control. A potential trajectory is calculated with a certain probability, the calculation being carried out using methods of artificial intelligence.

In the pamphlet GB 2562049 A Artificial intelligence methods are also used to determine potential trajectories of other vehicles and pedestrians, which flow into the planning of the further movement of an autonomous vehicle or a warning strategy for a driver. Further methods with reference to autonomous vehicles and the calculation of future movements can be found in the publications, for example JP 6103265 B2 , US 9,604,639 B2 , US 9,766,626 B1 and US 10,055,652 B2 .

None of the conventional strategies consider an interaction between the autonomous vehicle and the unprotected road user. The task is to improve conventional strategies for parking space behavior of autonomous vehicles.

This object is achieved by a method with the features of claim 1 and by a motor vehicle with the features of claim 10. Further advantageous embodiments and refinements of the invention emerge from the subclaims, the figures and the exemplary embodiments. The said embodiments and subjects of the claims can advantageously be combined with one another.

• - Erfassen des ungeschützten Verkehrsteilnehmers mittels mindestens eines Sensors des Fahrzeugs,
• - Vorausberechnen der künftigen Bewegung des ungeschützten Verkehrsteilnehmers in Form mindestens einer Trajektorie,
• - Planen von Steuerungsbefehlen zur Fahrt des Fahrzeugs unter Berücksichtigung der Trajektorie zum Vermeiden einer Kollision,
• - Erteilen von Steuerungsbefehlen an Aktuatoren des Fahrzeugs,

dadurch gekennzeichnet, dass die Steuerungsbefehle in die Vorausberechnung der künftigen Bewegung des ungeschützten Verkehrsteilnehmers einfließen.A first aspect of the invention relates to a method for controlling an autonomous vehicle at low speed by means of a mathematical model for the interaction between the vehicle and at least one unprotected road user in the vicinity of the autonomous vehicle, comprising the steps:

• - Detection of the unprotected road user by means of at least one sensor of the vehicle,
• - Predicting the future movement of the unprotected road user in the form of at least one trajectory,
• - Planning of control commands for driving the vehicle taking into account the trajectory to avoid a collision,
• - Issuing control commands to actuators of the vehicle,

characterized in that the control commands flow into the forecast of the future movement of the unprotected road user.

The core of the present invention is an interaction between autonomous vehicles and unprotected road users. In the method according to the invention, an interaction between an autonomous vehicle and an unprotected road user is advantageously included in the calculation of future movements of the unprotected road user. The mathematical model is based on an artificial neural network. It is essential to the invention that planned control actions of the autonomous vehicle flow into the prediction of the movement of the unprotected road user. The movement behavior of the autonomous vehicle, especially in a parking space, corresponds more to that of a human driver than a conservative automated system.

The environment of the autonomous vehicle includes the area in which a planned movement of the vehicle could lead to a collision with an unprotected road user.

Low speed means that the autonomous vehicle moves slowly compared to driving on a clear lane, depending on the conditions of a parking space. For example, walking pace can be assumed for a low speed, e.g. 4-6 km / h.

The method according to the invention can be explained using two modules. In the first module, a future movement of an unprotected road user is calculated. In this case, one or more (N≥1) trajectories are determined, with a probability preferably being calculated for each precalculated trajectory of the unprotected road user. In the second module, a decision is made as to which control commands are issued to the corresponding actuators of the vehicle. The control commands are preferably selected from the group comprising actuations of the steering, the power control of the drive and the brake. The planned control commands of the autonomous vehicle then flow into the prediction of the movement of the unprotected road user. In internal combustion engines, the control commands can also relate to the throttle. In an alternative embodiment, the method can also be carried out with a module that has the features of both of the above-mentioned modules.

Before they are issued to the corresponding actuators, the control commands are preferably checked for their convergent action. This advantageously ensures that a collision with the unprotected road user can be avoided. If there is no convergence, the method is preferably repeated from the precalculation of the future movement of the unprotected road user. That means that as long as between module 1 and module 2 is iterated until the control commands no longer change and convergence is thus achieved.

The method is preferably carried out in a parking space. The method is particularly advantageous in parking spaces because in the confusing environment it takes into account the behavior of unprotected road users through the deliberate interaction between the autonomous vehicle and unprotected road user than conventional methods. This increases the safety for the unprotected road users and the performance of the autonomous vehicle in terms of comfort and the time required to complete the maneuver in the parking space.

The unprotected road user is preferably a pedestrian. The method is particularly advantageously suitable, on the basis of the interaction between the autonomous vehicle and unprotected road user, to continue and complete a movement of the autonomous vehicle that has started, while ensuring the safety of the pedestrian.

The sensors for detecting the unprotected road user are preferably selected from the group comprising a camera, radar, lidar and devices for vehicle-to-vehicle, vehicle-to-pedestrian and vehicle-to-cyclist communication. The unprotected road users can periodically send status data with a mobile radio device, for example

In the method, the autonomous vehicle preferably moves at an approximately constant speed. This advantageously provides the unprotected road user with clarity about the planned movement of the vehicle. In contrast to conventional methods in which autonomous vehicles drive at varying speeds and stop when an unprotected road user is detected, the movement of the vehicle is completed relatively quickly and safely here.

A second aspect of the invention relates to an autonomous vehicle with a control device on which a first module for precalculating a trajectory of an unprotected road user using a mathematical model is based the basis of an artificial neural network and a second module for calculating control commands for moving the autonomous vehicle are implemented and designed to control a method according to the invention.

The advantages of the autonomous vehicle correspond to the advantages of the method according to the invention.

• 1 eine Verkehrssituation mit einer Ausführungsform eines erfindungsgemäßen Fahrzeugs in einem Parkraum.
• 2 ein Fließdiagramm eines herkömmlichen Verfahrens zum Steuern eines Manövers eines autonomen Fahrzeugs.
• 3 ein Fließdiagramm einer Ausführungsform des erfindungsgemäßen Verfahrens.
• 4 eine modulare Darstellung einer Einrichtung zum Ausführen des erfindungsgemäßen Verfahrens.

The invention is explained in more detail with reference to the figures. Show it

• 1 a traffic situation with an embodiment of a vehicle according to the invention in a parking space.
• 2 Figure 12 is a flow diagram of a conventional method for controlling a maneuver of an autonomous vehicle.
• 3 a flow diagram of an embodiment of the method according to the invention.
• 4th a modular representation of a device for carrying out the method according to the invention.

In 1 is a traffic situation in a parking space 1 shown. The parking space includes 1 a number of parking spaces. A first vehicle 10 is between two vehicles 11 and parked across from other vehicles. The first vehicle 10 is an autonomously driving vehicle. The direction of movement of the first vehicle 10 is illustrated by the dashed arrow emanating from it.

The first vehicle 10 has at least one sensor in the rear area 20th on, which is formed, the surroundings of the first vehicle 10 capture. The sensor 20th can be a camera, a radar or a lidar, without being limited to them. The said embodiments of the sensors 20th also all occur and in different places of the first vehicle 10 be arranged, ie next to the front area, for example in the rear area and on the sides of the first vehicle 10 .

Furthermore is the first vehicle 10 trained in vehicle-to-vehicle communication, vehicle-to-pedestrian and vehicle-to-cyclist communication. The corresponding facilities of the first vehicle 10 are also covered by the term sensor 20th includes. The unprotected road users can periodically send status data with a mobile radio device, for example.

The sensor 20th is with a control device 30th connected. In the control device 30th is a first module 31 implemented to precalculate a trajectory of an unprotected road user using a mathematical model based on an artificial neural network. Furthermore is in the control device 30th a second module 32 implemented for calculating control commands for moving the first vehicle. The modules 31 , 32 are as separate software in the control device 30th to understand which interact with each other, the function of the second module 32 on the information of the first module 31 builds up. The modules 31 , 32 can also be spatially separated, ie, for example, the first module can also be outside the control device 30th , e.g. on a hard drive.

An unprotected road user in the parking area 1 is a pedestrian 40 . The direction of movement of the pedestrian 40 is illustrated by the trajectories T1, T2, T3 emanating from it in the form of dashed arrows. The trajectories T1, T2, T3 mean possible movement paths that the pedestrian could follow from his current position. The first module calculates a probability of the actual impact by the pedestrian for the trajectories. The first trajectory T1 has the highest probability here, the second highest probability the second trajectory T2 and the lowest probability the third trajectory T3.

In 2 Figure 3 is a flow diagram of a conventional method for controlling the movement of an autonomous vehicle in a parking space environment. The vehicle moves out of its parking position at a slow speed. In this case, in a first step S1, an unprotected road user, for example a pedestrian, becomes 40 , detected by a sensor of the vehicle. The sensor data are sent to a control device of the vehicle. In a second step S2, the future movement of the unprotected road user is calculated in the form of at least one trajectory, regardless of the vehicle. Several trajectories can also be calculated, as well as the probability with which a corresponding trajectory will actually be taken by the unprotected road user. In a third step S3, control commands for driving the vehicle are planned taking into account the trajectory to avoid a collision. The trajectory is selected that the unprotected road user would take with the highest probability. In a fourth step S4, the corresponding control commands are issued to actuators of the vehicle which, for example, control the steering, acceleration and braking of the vehicle.

If there is no convergence between the effects of the planned control commands, the Difference to the conventional method according to 2 in one embodiment of the method according to the invention as shown in FIG 4th an information loop is passed back from step S3 to step S2. Information about planned control commands flows into the calculation of the trajectories and their probabilities. Based on the planned movements of the first vehicle 10 become the predicted trajectories of the movement of the pedestrian 40 and revised their corresponding probabilities. The control commands are then revised in step S3. The pedestrian 40 and the first vehicle 10 change their actions using game theory, for example, until a state of equilibrium is established. Only then are the control commands sent to the corresponding actuators of the first vehicle in step S4 10 passed on.

The modular functioning of the method according to the invention is shown in 4th shown. During a parking maneuver at low speed, data is received from the sensor 20th the first module 31 provided. The first module 31 processes the data as an artificial neural network. Based on an interaction between the first vehicle 10 and the pedestrian 40 trajectories of the future movement of the unprotected road user are predicted and their probability is calculated. These calculations are sent to the second module 32 delivered. In the second module 32 become control commands for further movement of the first vehicle 10 calculated while avoiding a collision. The decision is made on the assumption that the unprotected road user of the first vehicle 10 lets go by. Before the control commands are issued to the corresponding actuators, the control commands flow into the calculations of the first module 31 on (feedback loop). Then the control commands are issued to the actuators. The first vehicle 10 completes its movement at a low, but as constant as possible, speed.

List of reference symbols

1

Parking space

10

first vehicle

11

another vehicle

20th

sensor

30th

Control device

31

first module

32

second module

40

pedestrian

QUOTES INCLUDED IN THE DESCRIPTION

This list of the documents listed by the applicant was 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.

Patent literature cited

• CN 108172025 A [0005]
• DE 102017210394 B3 [0006]
• GB 2562049 A [0007]
• JP 6103265 B2 [0007]
• US 9604639 B2 [0007]
• US 9766626 B1 [0007]
• US 10055652 B2 [0007]

Claims (10)

A method for controlling an autonomous vehicle (10) at low speed by means of a mathematical model for the interaction between the vehicle (10) and at least one unprotected road user (40) in the vicinity of the autonomous vehicle (10), comprising the steps of: - Detecting the unprotected Road user (40) by means of at least one sensor (20) of the vehicle (10), - precalculation of the future movement of the unprotected road user (40) in the form of at least one trajectory (T1, T2, T3), - planning of control commands for driving the vehicle ( 10) taking into account the trajectory (T1, T2, T3) to avoid a collision, - issuing control commands to actuators of the vehicle (10), characterized in that the control commands flow into the forecast of the future movement of the unprotected road user (40). Procedure according to Claim 1 , a probability being calculated for each precalculated trajectory (T1, T2, T3) of the unprotected road user (40). Method according to one of the preceding claims, wherein the control commands are selected from the group comprising actuations of the steering, the power control of the drive and the brake. Method according to one of the preceding claims, wherein the control commands are checked for their convergent action. Procedure according to Claim 4 In the event of a lack of convergence, the method is repeated from the precalculation of the future movement of the unprotected road user (40). Method according to one of the preceding claims, wherein the method is carried out in a parking space (1). Method according to one of the preceding claims, wherein the unprotected road user (40) is a pedestrian (40). Method according to one of the preceding claims, wherein the sensors (20) for detecting the unprotected road user (40) are selected from the group comprising a camera, radar, lidar and devices for vehicle-to-vehicle, vehicle-to-pedestrian and Vehicle-to-cyclist communication. Method according to one of the preceding claims, wherein the vehicle (10) moves at an approximately constant speed. Autonomous vehicle (10) with a control device (30) on which a first module (31) for precalculating a trajectory (T1, T2, T3) of an unprotected road user (40) using a mathematical model based on an artificial neural network and a second module (32) for calculating control commands for moving the vehicle (10) are implemented and designed, a method according to one of the Claims 1 - 9 to control.

Images