DE102022208646A1 - Method for detecting obstacles on uneven surfaces - Google Patents

Abstract

The invention relates to methods for obstacle detection, in which a sensor is used to detect a route in front of a vehicle and the sensor data thus obtained are processed by a computing unit, with an expected terrain level being determined in a first step based on the vehicle geometry; in a second step, an upper offset and a lower offset to the expected terrain level are specified, an area being formed by the upper and lower offsets; sensor data obtained in a third step are filtered out if they are within the formed area; upon receipt of a signature outside the formed area, in a fourth step the at least the upper offset is changed in such a way that the formed area is adjusted around the obtained signature.

Description

The present invention relates to a method for obstacle detection in a vehicle, whereby an obstacle is to be understood as an object which cannot be driven over by the vehicle in question. A computer program product is also included.

Various methods for obstacle detection are known from the prior art. For example, concerns DE 10 2013 200 385 A1 a method for supporting a driver of a vehicle when driving on uneven terrain, a corresponding device and a corresponding computer program product. Using optical, radar or lidar sensors, an unevenness in the terrain being traveled is detected and classified as to whether the unevenness represents an obstacle that prevents the vehicle from continuing its journey without causing damage. In addition to an incline or decline, the obstacle or unevenness can be a pothole, a ramp, a transverse or longitudinal bump.

Further revealed DE 10 2012 203 091 A1 Method for detecting objects in the surroundings of a motor vehicle with a distance sensor that works according to the pulse-echo method. The method makes it possible to lower a threshold specified by a characteristic curve in the case of a weakly reflecting background and thereby increase the sensitivity. This means that even weakly reflecting objects can be detected. Furthermore, the method makes it possible, particularly when large interference signals occur due to a highly reflective background, to mask these out by increasing the threshold specified by the characteristic curve. This allows the signals to be clearly assigned to objects and no false measurements due to measurements of ground signals. Furthermore, the method according to the invention not only suppresses ground signals, but it also makes it possible to suppress interference signals that arise, for example, from measurement errors.

The object of the present invention is to provide an improved method for obstacle detection, particularly when operating the vehicle on uneven ground.

This task is solved with the obstacle detection method according to the invention. A sensor is used to record a route in front of a vehicle. The sensor data obtained in this way is processed by a computing unit, with an expected terrain level being determined in a first step.

The vehicle can be any type of vehicle, but the focus is on off-road vehicles. Sometimes it can be a vehicle operated conventionally by an operator, a semi-autonomous or a fully autonomous vehicle. In particular, the vehicle is work machines, for example agricultural, forestry or construction machines. These are typically operated in a work environment on an uneven surface or road. In particular, the route is a road, a (field) path or a construction site or an agricultural area - in other words, an unpaved surface. The route often has uneven topography. This can result from an unpaved or leveled surface, due to bumps, potholes, ruts, curbs, hills or rubble. The sensor is one that provides depth information, for example (3D) lidar, (3D) radar, ToF camera, stereo camera. The sensor types mentioned can also be combined with one another to improve object detection. Depending on the sensor type, the sensor is a transmitter/receiver unit, since, particularly with a radar or lidar sensor, a signal is sent out and a reflected signal is received. The computing unit can be designed as a separate computing unit or can sometimes be integrated into another device. In particular, the computing unit can be integrated into the sensor or the transmitting/receiving unit. Alternatively, the computing unit can also be embedded in another control unit of the vehicle.

Determining the expected terrain level describes the determination of a plane or area in which an expected route lies and is carried out, for example, on the basis of an extrapolation of the route currently or last traveled, on the basis of the vehicle geometry, a vehicle orientation, on the basis of position data and/or map data. In particular, acceleration and/or inclination sensors can be used to determine the expected terrain level. In addition, short-term influences caused by driving on uneven surfaces can be compensated for or eliminated using the acceleration and inclination sensors when determining the expected terrain level. To the extent that knowledge of inclines and slopes is provided via position and/or map data, this information can be taken into account to determine the expected terrain level. Sometimes a data interface can be provided here, by means of which data is exchanged with a server or a remote control in order to achieve the corresponding Obtain and provide topography or map data.

In a second step of the method, an upper offset and a lower offset to the expected terrain level are specified, with an area being formed by the upper and lower offsets. The upper offset and the lower offset can be the same size, i.e. have the same value. In other words, the upper and lower offsets form a boundary which defines or limits the formed area. For example, the length of the formed region can match the distal end of the detection range of the sensor, i.e. the distal length of the expected terrain level and the upper or lower offset. For spatial assignment, a positive sign could be assigned to the upper offset and a negative sign to the lower offset. The positive and negative signs refer to the orientation in relation to the expected terrain level. When viewed from the side, the formed area represents a surface, but in fact a volume is described in the space in front of the vehicle. This results from the fact that the viewing area of the sensor also extends to the side in front of the vehicle.

In a third step, sensor data obtained is filtered out if it is within the formed area. On the one hand, this means that uncritical sensor data can be ignored, which means that the data to be processed can be significantly reduced. As a result, less computing power is used to process data that is not needed. Alternatively, the detection range of the sensor can also be limited so that detection only takes place outside the formed area.

When a signature is received outside the formed area, in a fourth step the at least the upper offset is changed in such a way that the formed area is adjusted around the obtained signature. A signature describes a signal characteristic that is caused by an object, in particular an obstacle. Consequently, such a signature can be caused by reflected signals (radar, lidar) or by image recognition using imaging sensors (camera). The change in the offset can either take place in the form of an adjustment of the value of the offset for the entire formed area or in a distally limited manner in the immediate distal proximity to the object. The offset can also be adjusted in the form of an adjustment curve. A reduction in the offset is particularly advantageous. The adjustment described is primarily related to the upper offset, but the upper and lower offset can also be adjusted. Both offsets can be adjusted symmetrically to the expected base plane or individually for the upper and lower offsets. The offset can also be deactivated, sometimes temporarily.

According to a further development of the method, in a fifth step the computing unit outputs a signal in response to the signature. The output signal can be an indication on a human-machine interface, for example a display. Alternatively or additionally, driving dynamics parameters of the vehicle can be adjusted, for example a driving speed. Operating parameters of the engine (torque, speed, power), the transmission (gear changing) or differential locks (locking/unlocking) can also be adjusted. Furthermore, a vehicle stop or an emergency stop can be initiated by the signal from the computing unit.

To the extent that a signature has been received, further sensors can primarily detect the adapted trained area in which the signature was received. In particular, the signature received could be qualified or confirmed based on the various sensor data.

In a further development, a value which corresponds to a maximum permissible obstacle height is specified as the maximum value, at least for the upper offset. This means that the vehicle can drive over obstacles below the maximum permissible obstacle height, and consequently an obstacle above a maximum permissible obstacle height would result in a collision and thus damage to the vehicle. A gradient that is too steep could also be considered an obstacle above a maximum permissible obstacle height. The upper offset, like the lower offset, can be specified by an operator or via the data interface. The values for the offset can also be stored on a memory element of the vehicle.

Alternatively or additionally, the upper and/or lower offsets are variable over a distance. This means that a different offset can be provided close to the vehicle than at a greater distance. This allows early detection of distant objects, while precise detection takes place in a close area of the vehicle.

Furthermore, the upper and/or lower offset can be variable depending on a current vehicle speed or a speed specification. This means that it is sometimes difficult to drive over smaller objects at higher vehicle speeds, which can sometimes be passed without any problems at lower vehicle speeds. It may also be necessary to recognize objects at an early stage at higher vehicle speeds and sometimes to carry out precise processing of the sensor data or signatures obtained while approaching. Just like a current vehicle speed, the procedure described above can also be applied to a speed specification. A current vehicle speed differs from a driving target in that sometimes a current vehicle speed deviates from a speed target due to external influences (insufficient engine power, traction, starting process).

In a further development, topography data from other data sources can be fed to the computing unit and merged with the sensor data received. This means that, for example, topography data can be used to check the plausibility of the sensor data obtained.

According to a further aspect of the present invention, this relates to a computer program product comprising instructions which, when the program is executed by a computer, cause it to carry out the method according to the invention. The computer can be, for example, the computing unit. Alternatively, the computer or the computing unit can also be integrated into the sensor or the transmitter/receiver unit. Furthermore, the computer program product can be executed remotely on a smartphone, a tablet PC or via a computer. The computer program product can also be executed on any other control device of the vehicle, for example a transmission or engine control device or on a vehicle control computer.

A machine learning method can also be used to optimize object detection and/or the adjustment of the offset.

The invention is explained below using the only 1 described. The illustration shows a vehicle 1 designed as an agricultural machine. The vehicle 1 has a sensor 2 on the front of the vehicle. The sensor 2 has a forward-facing detection area 3, which is shown in simplified form using dash-dotted lines. The vehicle 1 travels along a route 4, which is shown by a continuous straight line.

Starting from the current route 4, an expected terrain level 5 is determined, which in the present case continues the current route 4 as a dashed line. In an expected section 6, the route 4 has an uneven topography, which is characterized by elevations and depressions. The expectation section lies in front of the vehicle 1 and its extent corresponds to the distal length of the detection area 3.

An upper offset 7 and a lower offset 8 form an area 9 around the expected terrain level 5, which is shown hatched. The upper and lower offsets 7, 8 thus delimit an area around the expected terrain level 5 in the side view shown here. In a perspective view it would be apparent that the formed area 9 is not a surface, but a volume, which can be described by the expected terrain level 5 and the upper and lower offsets 7, 8. In the present case, an upper offset value I and a lower offset value II are the same size, so that the formed area 9 is symmetrical about the expected terrain level 5. Consequently, a height III of the formed region 9 corresponds to the addition of the upper and lower offset values I, II or can be described by the equation (III = I + II).

Without the method according to the invention, the elevations and depressions of the route 4 in the expected section 6 would be obtained sensor data for signatures, which are to be viewed as undesirable disturbance variables. Since these lie in the formed area 9, they are filtered out or not taken into account. These filtered out sensor data 10 are shown as empty circles in the figure.

However, as soon as an object 11 protrudes from the formed area 9, the sensor data is recorded and a signature of the object 11 is obtained. The sensor data 12 taken into account are shown as filled circles. Consequently, an object 11 was detected, which represents an obstacle for the vehicle 1, which sometimes cannot be driven over. To further qualify the sensor data, the upper and/or lower offset 7, 8 can now be adjusted in an area around the object 11 or around the sensor data 12 taken into account in order to obtain a more precise image of the object 11. In the present case, the upper offset 7 would be reduced so that deeper signatures are also taken into account. These supplemented sensor data 13 are shown as circles with hatching. Example rish becomes in 1 just added another point. However, depending on the method, the offset 7, 8 could be further adjusted until a complete signature of the object 11 was obtained and the object 11 could sometimes be classified.

As a result, based on the sensor data 12, 13 received, a signal would be output that there is an obstacle in front of the vehicle 1 that cannot be driven over. Consequently, a warning could be issued, a stopping process could be initiated, or an alternative trajectory could be determined in order to avoid the object 11 through a steering intervention.

Not shown in the 1 a computing unit. However, this does not mean that it is not covered by the invention. The computing unit can also be integrated into the sensor 2, for example. In addition, no other sensors of the type mentioned are shown. These can also be integrated into the sensor 2 as a unit or provided on or in the vehicle 1 at a suitable location. In particular, the additional sensors are used to observe the adapted formed area 9 or the obtained signatures of the object 11.

Reference symbols

1

vehicle

2

sensor

3

Detection area

4

route

5

expected terrain level

6

Expectations section

7

upper offset

8th

lower offset

9

trained area

10

filtered out sensor data

11

object

12

sensor data taken into account

13

supplemented sensor data

I

Offset value upper offset

II

Offset value lower offset

III

Height range

QUOTES INCLUDED IN THE DESCRIPTION

This list of 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.

Cited patent literature

• DE 102013200385 A1 [0002]
• DE 102012203091 A1 [0003]

Claims (8)

Method for obstacle detection, wherein a route (4) lying in front of a vehicle (1) is detected using a sensor (2) and the sensor data (10, 12, 13) thus obtained are processed by a computing unit, wherein - in a first step an expected terrain level (5) is determined; - in a second step, an upper offset (7) and a lower offset (8) to the expected terrain level (5) are specified, an area (9) being formed by the upper and lower offsets (7, 8); - Sensor data (10) obtained in a third step are filtered out if they are located within the formed area (9); - When a signature is received outside the formed area (9), in a fourth step the at least the upper offset (7) is changed in such a way that the formed area (9) is adjusted around the obtained signature. Procedure according to Claim 1 , characterized in that in a fifth step the computing unit outputs a signal in response to the signature. Method according to one of the preceding claims, characterized in that further sensors primarily detect the adapted formed area (9) in which the signature was obtained. Method according to one of the preceding claims, characterized in that a value (I) which corresponds to a maximum permissible obstacle height is specified at least for the upper offset (7). Method according to one of the preceding claims, characterized in that the upper and/or lower offset (7, 8) is variable over a distance. Method according to one of the preceding claims, characterized in that the upper and/or lower offset (7, 8) is variable depending on a current vehicle speed or a speed specification. Method according to one of the preceding claims, characterized in that topography data from further data sources are supplied to the computing unit and combined with the sensor data (10, 12, 13) obtained. Computer program product, comprising instructions which, when the program is executed by a computer, cause it to carry out the method according to one of the Claims 1 until 7 to carry out.

Images