Method and system for determining clearances in the environment of a vehicle and for controlling the vehicle based on determined clearances

Abstract

The invention relates to a method for determining clearances in the environment of a vehicle (F) and for controlling the vehicle (F) based on the determined clearances, comprising the following steps: - Detection (S1) of the vehicle's environment (F) using at least one environment detection sensor (2) and generation of sensor data; - Detecting (S2) objects using at least one environmental sensing sensor (2); - Determining (S3) a probability of explicit and implicit free spaces based on the sensor data of at least one environment detection sensor (2); - Entering (S4) the detected objects and the determined probability of the free spaces (FW) into a grid map (G) which has a large number of grid cells (Gz); - Adjusting (S5) vehicle parameters based on the probabilities of the free spaces (FW) in the vehicle environment; - Controlling (S6) the vehicle based on grid cells (Gz) which have a definable minimum probability value (FW) for free spaces.

Description

[0001] The invention relates to a method and a system for determining free spaces in the environment of a vehicle and for controlling the vehicle based on determined free spaces.

[0002] It is known from the state of the art that in the field of assisted driving and autonomous driving, it is common for vehicles to react to objects detected by appropriate sensors, such as lidar, cameras, radar, or ultrasound. Furthermore, it is also very common to use free space representations, such as a grid, as an input for driving functions. This enables vehicles to react to objects that are not, for example, other road users. The driving function can then calculate trajectories that avoid obstacles on the road, for example, by avoiding them or by stopping if the obstacle blocks the entire lane.

[0003] It is an object of the present invention to provide a method and a system that enable reliable and precise clearance determination and safe control of a vehicle. This object is achieved by the subject matter of independent claims 1 and 7. Further advantageous embodiments and embodiments are the subject matter of the dependent claims.

[0004] Initial considerations were that problems could arise if only the end of a free space was considered and provided to a driving function.

[0005] The further away the objects are, the worse the object detection is. Therefore, it could happen that a vehicle enters a situation too quickly and has to brake suddenly, because the quality of detection improves as the distance between the vehicle and the object decreases and an object is closer to the vehicle than originally detected. Furthermore, "clear space" does not necessarily mean that it is passable. Current systems use maps or are activated by drivers. Both inputs can be used to determine whether the space in front of the vehicle is actually passable. However, this no longer works with autonomous vehicles or when no map is available or when the map is out of date, e.g. due to construction work. In such situations, the vehicle must find a new route to its destination.

[0006] According to the invention, a method for determining free spaces in the environment of a vehicle and for controlling the vehicle based on the determined free spaces is therefore proposed, comprising the following steps: - detecting the surroundings of the vehicle by means of at least one surroundings detection sensor and generating sensor data; - Detecting objects by means of the at least one environment detection sensor; - Determining a probability of explicit and implicit free spaces based on the sensor data of the at least one environment detection sensor; - Entering the detected objects and the determined probability of the free spaces into a grid map, which has a large number of grid cells; - Adjusting vehicle parameters based on the probabilities of the free spaces in the vehicle environment; - Control the vehicle based on grid cells that have a definable minimum probability value.

[0007] The vehicle is preferably an autonomous vehicle or at least a semi-autonomous vehicle, which can operate autonomously and without driver intervention in certain scenarios, such as highway scenarios.

[0008] The at least one environment detection sensor is, for example, a camera, in particular a mono or stereo camera, an ultrasonic sensor, a lidar, or a radar sensor. Furthermore, it would be conceivable to use several identical or different sensors of the aforementioned type. The use of multiple sensors is advantageous because, for example, with different sensors, different objects and features of the environment can be detected with varying degrees of success by the respective sensors, and the fusion of the various sensor data enables a more precise depiction of the environment. When cameras are used, they can also be part of a surround-view camera system.

[0009] Object detection can be performed by analyzing sensor data. For example, camera images can be processed using segmentation, such as semantic segmentation or instance segmentation, to identify, among other things, objects in the environment. Depending on the sensor used, the processing methods for the sensor data can be selected accordingly.

[0010] A minimum value for the probability of a clear space could, for example, be set between 80% and 100%. At these values, the vehicle would consider the area to be "clear." For example, the explicit clear space could be detected and used to determine whether this area is passable. The values ​​for the probability of a clear space can decrease with distance from the vehicle, as detection accuracy decreases with distance from the vehicle.

[0011] In a particularly preferred embodiment, adapting vehicle parameters includes setting an optimal vehicle speed. This is particularly advantageous because, in the event of a possible misdetection of an object or the free space, a safe stopping within the free space can be guaranteed with a certain probability. For example, the optimal vehicle speed can be selected such that the vehicle can safely stop in an area with a free space probability of, for example, at least 80%. The speed can be adjusted independently by an autonomous vehicle. If, for example, the vehicle were to move towards an area that is only 40% free, i.e., has a corresponding 60% probability of being occupied, the vehicle would have to reduce its speed accordingly, since the area cannot be classified as clear with certainty, but also cannot be classified as occupied with certainty.As the vehicle approaches the corresponding area, detection accuracy also improves, and the area could be classified as occupied at a certain distance from the vehicle. To ensure that the vehicle can safely stop before reaching the area or the subsequently detected object, the speed must be reduced accordingly when an area with, for example, a 40% clear-space probability is detected.

[0012] In a further preferred embodiment, controlling the vehicle comprises reducing the speed, increasing the speed, and / or steering intervention. Depending on the determined clearance, the vehicle can be controlled, for example, along grid cells with a clearance probability of at least 80%, and these grid cells can be used accordingly for trajectory planning.

[0013] Furthermore, the probability for explicit free spaces is particularly preferably determined by detecting a ground plane or by classifying the sensor data.

[0014] The classification of the sensor data depends on the environment detection sensor(s) used. If, for example, a camera is used, the camera image can be evaluated using semantic segmentation, among other things, to determine whether the area in front of the vehicle is passable.

[0015] In a further preferred embodiment, at least one distance between the vehicle and the detected objects in the surroundings of the vehicle is determined.

[0016] Particularly preferred is the probability of implicit clear spaces being determined based on the determined distance between the vehicle and the respective object. This assumes that the area between the vehicle and a detected object can be considered clear.

[0017] According to the invention, a system for controlling a vehicle based on determined free spaces in an environment of the vehicle is further provided, comprising at least one environment detection sensor for detecting the environment of the vehicle, a control device which is designed to control the vehicle and a computing device which is designed at least to determine free spaces.

[0018] Further advantageous embodiments are shown in the drawings, which show: Fig. 1: a schematic flow diagram of a method according to an embodiment of the invention; Fig. 2: a schematic representation of a system according to an embodiment of the invention; Fig. 3: a schematic representation of a grid map according to an embodiment of the invention; Fig. 4: a schematic representation of a speed curve when approaching an obstacle with different occupancy probabilities; Fig. 5: another schematic representation of a speed curve when approaching an obstacle with different occupancy probabilities; Fig. 6: another schematic representation of a speed curve when approaching an obstacle with different occupancy probabilities.

[0019] In Fig. 1 shows a schematic flow diagram of a method according to an embodiment of the invention. According to the method for controlling a vehicle based on determined clear spaces in the vehicle's surroundings, in step S1 the vehicle's surroundings are detected by at least one surroundings detection sensor, and sensor data is generated. In a further step S2, objects are detected by the at least one surroundings detection sensor. In step S3, a probability of explicit and implicit clear spaces is determined based on the sensor data of the at least one surroundings detection sensor. In a step S4, the detected objects and the determined probabilities of the clear spaces are entered into a grid map, wherein the grid map has a plurality of grid cells. In step S5, vehicle parameters are adjusted based on the probabilities of the clear spaces in the vehicle's surroundings.In a step S6, the vehicle is controlled based on the grid cells that have a definable minimum value of the probability for free spaces.

[0020] Fig. Figure 2 shows a schematic representation of a system according to one embodiment of the invention. The system 1 comprises at least one environment detection sensor 2, a computing device 3, and a control device 4. The computing device 3 is designed to carry out a method for determining open spaces. For this purpose, the computing device 3 can, for example,

[0021] Execute algorithms for sensor data processing and be configured to create a grid map. The created grid map can be transmitted from the computing device to the control device so that the control device 4 can control the vehicle F according to the free spaces FW. The at least one environment detection sensor 2 is connected to the computing unit 3 via a data connection D, and the computing unit 3 is connected to the control device 4 via a data connection D. The data connection D can be wired or wireless.

[0022] The Fig. 3 shows a schematic representation of a grid map G according to one embodiment of the invention. The grid map G shown here consists of a plurality of grid cells Gz. Furthermore, a vehicle F with at least one environment detection sensor 2 is shown, which moves in the direction of travel R. The at least one environment detection sensor 2 is oriented in the direction of travel R. However, it would also be conceivable, particularly when using multiple sensors, that the sensors are oriented not only in the direction of travel R but also laterally and / or against the direction of travel R. The vehicle F furthermore has at least one computing device 3 and a control device for controlling the vehicle F. Detected objects are entered in the grid map G as occupied grid cells Gb. The objects can be, for example, other road users, road markings, obstacles or the like.As shown, for example, a distance d between vehicle F and an object or one or more occupied grid cells Gb can be determined. In the area between vehicle F and occupied grid cells Gb, an implicit free space can be assumed, and a value for a probability of a free space FW, or also free space probability FW, can be assigned to the grid cells Gz therein. Based on the information about the free space probabilities FW and the occupied grid cells Gb, a trajectory T can be planned in the vehicle, for example, and the vehicle F can be controlled accordingly using the control device 4. For this purpose, trajectory planning can be carried out in the computing device 3 based on the created grid map.

[0023] Fig. 4 shows a schematic representation of a speed curve when approaching an obstacle with different occupancy probabilities. The first representation shows a high occupancy probability BW of the cells with a potential obstacle H. When approaching the obstacle H, the speed is reduced by means of a braking process vb without prior speed adjustment va. As shown in the first representation, comfortable braking is possible if the braking process vb is initiated early. In the second representation, the occupancy probability BW is lower than in the previous representation. Accordingly, the braking process vb is initiated later because the obstacle is only reliably detected later. This leads to less comfortable braking. With an even lower occupancy probability BW, braking takes place even later and leads to a late, uncomfortable braking process vb.

[0024] If the occupancy probability is lower, intervention occurs later, but with longer delays. The underlying assumption is that even with a low probability, a stop must be made in any case (a very low probability threshold may be assumed below which no intervention occurs). Since a low probability could result in a false positive, this strategy reduces the number of incorrect interventions. However, this may lead to uncomfortable behavior.

[0025] The Fig. Figure 5 shows another schematic representation of a speed curve when approaching an obstacle with different occupancy probabilities. The different representations also show, as in Fig. Four different occupancy probabilities BW exist. In the strategy shown here, a speed reduction va is already performed before the actual braking process vb. Here, the decelerations occur earlier but more gently, so that the speed is reduced earlier and the high, uncomfortable deceleration only occurs at the end.

[0026] Fig. Figure 6 shows a further schematic representation of a speed curve when approaching an obstacle with different occupancy probabilities. In Fig. Figure 6 shows how the behavior would develop if the occupancy probability BW drops back to zero, meaning a false detection occurred. The lower the initial occupancy probability, the less serious the impact of a false intervention. As long as the occupancy probability BW remains at least at a certain value, a speed reduction va is performed as a precaution. If the value of the occupancy probability BW drops to zero, an acceleration vb can be performed, since there is obviously no obstacle H in the direction of travel of vehicle F.

[0027] The three different representations of the different occupancy probabilities also show the advantage of more cautious braking at lower probabilities compared to a strategy that would react the same for all probabilities, i.e. would always brake early and gently: In this case, after the probability has dropped to zero, acceleration would have to resume from a very low speed level.

[0028] With a strategy like in Fig. As shown in Figure 4, a false detection has no effect if it is rejected early enough. However, if it is rejected later, the effect is even more serious.

[0029] The Fig. The strategy shown in Figure 5 is therefore a compromise between both strategies, i.e. the one in Fig. 4 shown and the one that always reacts the same to all probabilities. List of reference symbols 1 system 2 environment detection sensor 3 Computing device 4 Control device BW occupancy probability d distance D Data connection F vehicle FW Probability for Free Space G Grid map Gb occupied grid cell Gz Grid Cell H Obstacle R direction of travel T trajectory S1-S6 process steps va speed reduction vb braking process v2 acceleration

Claims

[1] Method for determining free spaces in the environment of a vehicle (F) and for controlling the vehicle (F) based on the determined free spaces, comprising the following steps: - detecting (S1) the surroundings of the vehicle (F) by means of at least one surroundings detection sensor (2) and generating sensor data; - detecting (S2) objects by means of the at least one environment detection sensor (2); - determining (S3) a probability of explicit and implicit free spaces based on the sensor data of the at least one environment detection sensor (2); - entering (S4) the detected objects and the determined probability of the free spaces (FW) into a grid map (G) which has a plurality of grid cells (Gz); - Adjusting (S5) vehicle parameters based on the probabilities of the free spaces (FW) in the vehicle environment; - Controlling (S6) the vehicle based on grid cells (Gz) which have a definable minimum probability value (FW) for free spaces. [2] Method according to claim 1, characterized by that adjusting vehicle parameters includes setting an optimal vehicle speed. [3] Method according to claim 1, characterized by that controlling the vehicle (F) comprises reducing the speed, increasing the speed and / or steering intervention. [4] Method according to claim 1, characterized by that the probability for explicit free spaces (FW) is determined by detecting a ground plane or by classifying the sensor data. [5] Method according to claim 1, characterized by that at least one distance (d) between the vehicle (F) and the detected objects in the surroundings of the vehicle (F) is determined. [6] Method according to claim 5, characterized bythat the probability for implicit clearances is determined based on the determined distance (d) between the vehicle and the respective object. [7] System (1) for determining free spaces in an environment of a vehicle (F) and for controlling the vehicle (F) based on the determined free spaces, comprising at least one environment detection sensor (2) for detecting the environment of the vehicle, a control device (4) which is designed to control the vehicle (F) and a computing device (3) which is designed at least to determine free spaces.

Images