Arrangement for guiding an agricultural vehicle across a field using sensor fusion and corresponding procedure

Abstract

An arrangement for guiding an agricultural vehicle (10) across a field (58) and a corresponding procedure comprise: a visual sensor system (62) directed at the field (58), which is connected to an image processing system (36) via signal transmission, which image processing system (36) is configured to identify features of the field (58) in the image signal of the visual sensor system (62) and to calculate a first path to be traveled for the vehicle (10) based on the identified features, a navigation unit (48) for determining the position of a reference point on the vehicle (10), a memory (72) in which stored waypoints for a second route to be traveled by the vehicle (10) are stored, a path calculator (70) configured to convert the first and / or second path into a common coordinate system, and a fusion unit (74) configured to merge the first path and the second path and, based on the merged path obtained in this way, to supply a signal to a steering control unit (34) to steer the vehicle (10) along the merged path and / or to store waypoints calculated on the basis of the merged path for the positions traveled by the vehicle (10) in a memory (72). It is proposed that the first path and / or the second path includes or comprises more than two waypoints and / or that the path calculator (70) or the image processing system (36) is configured to select field features (58) used to determine the first path based on the second path.

Description

[0001] The invention relates to an arrangement for guiding an agricultural vehicle across a field and a corresponding method. State of the art

[0002] In agriculture, there is a need to steer vehicles autonomously across a field to facilitate the operator's work and allow them to concentrate on monitoring and controlling the work being carried out while driving across the field, such as tilling the soil, applying fertilizers, or harvesting plants, or to provide complete automation. In addition to steering systems that generate a steering signal using a navigation system and a map of the paths to be traveled, optical steering systems are also known. These systems use the image signal from a camera or an active, scanning sensor (lidar or similar) to detect features in the field, which are then used to evaluate the path to be traveled.These features can be cultivation boundaries that involve a change in the soil profile, wheel tracks, crop rows, swaths and / or the edge of a still standing crop to be harvested.

[0003] The navigation systems use satellite-based positioning systems (GNSS), which can be supplemented or replaced by local sensors for recording the movement of the vehicle (inertial navigation systems and / or odometric systems), e.g. when driving through fields with suboptimal satellite signal reception under trees. The accuracy of the steering signal depends on the reception conditions, but also requires a sufficiently accurate map of the route to be traveled, which is not always available or may be inaccurate, e.g. if the positioning signal was inaccurate when the map was generated or there is a discrepancy between the position of certain features of the field (e.g. seeds sown there or plants planted in the ground) and the recorded position, for example if a towed seed drill deviates from the path of a tractor towing it, whose position is recorded, due to the influence of a slope.

[0004] In addition, sensor systems based on optical (visual) principles have been described. These systems are not subject to the disadvantages described above with navigation systems, are also more cost-effective, and can be used, for example, when navigating row crops. However, these visual sensor systems sometimes encounter difficulties when unfavorable visibility conditions, crop gaps, or sudden jumps in the crop rows (caused during sowing or due to a change in the image processing system between adjacent rows) occur.

[0005] For these reasons, various guidance systems have been proposed that provide for a fusion (merging, combining) of signals provided by a navigation system and a visual sensor system. In this case, the lateral deviation of the vehicle to be steered from a target path and the angle between the longitudinal axis of the vehicle and the longitudinal axis of the target path are usually provided and merged (fused) separately for each of the systems, usually with a weighting based on information regarding the accuracy of the respective signals (see, for example, US 6 445 983 B1, EP 1 266 553 A2, WO 2008 / 150418 A1, WO 2018 / 067473 A1, Wan, X., Liu, G. (2008). Automatic Navigation System with Multiple Sensors. In: Li, D. (eds) Computer and Computing Technologies in Agriculture, Volume II. CCTA 2007. The International Federation for Information Processing, vol 259. Springer, Boston, Ma.).

[0006] In the case of a row crop, the target path for the visual sensor system is determined by the rows of plants identified in the image, which also applies to the navigation system. The steering signal is typically selected so that the vehicle's wheels do not drive over the plants. The target path can be straight or curved, depending on the field in question.

[0007] The visual sensor system uses the image signals from the optical sensor (e.g. lidar and / or camera in stereo or mono version) to record the positions of several plants in front of the vehicle. From this, a line is generated which represents the centers of gravity of the plants in the respective rows. This line can be straight or curved. In DE 10 2019 203 651 A1, it is proposed to define the search area in which the optical sensor suspects plants based on a curvature of the stored path in order to improve the accuracy of the optical guidance. F. Rovira-Más et al., Kalman filter for sensor fusion of GPS, The 2006 ASABE Annual International Meeting, St. Joseph, Michigan, USA, July 9-12, 2006 describe guidance based on a fusion of visual sensor signals and GPS signals, in which the GPS data is only used to extrapolate the previous path based on the visual data in the event of a gap in the visual data.The GPS-based target path is therefore based on the vehicle’s previous path. Task

[0008] In the previous procedure, i.e. the separate calculation of the lateral deviation and the angle of the vehicle for the visual sensor system and the navigation system (based on a previously stored path) and their subsequent fusion, essential information is lost - even if the guidelines for fusion are transformed into a common coordinate system by fuzzy logic (Wan et al., op. cit., regarded as generic) - even though they were originally available, in particular the data regarding a possible curvature of the future course of the path to be followed.

[0009] The object underlying the invention is to provide a steering system for guiding an agricultural vehicle across a field and a corresponding method which does not have the said disadvantages or has them to a reduced extent. Solution

[0010] This object is achieved according to the invention by the teaching of patent claims 1, 2 and 10, wherein the further patent claims list features which further develop the solution in an advantageous manner.

[0011] An arrangement for driving an agricultural vehicle across a field and a corresponding procedure include: a visual sensor system directed at the field, which is connected to an image processing system for signal transmission, which image processing system is configured to identify features of the field in the image signal of the visual sensor system and to calculate a first path to be covered by the vehicle based on the identified features, a navigation unit for determining the position of a reference point on the vehicle, a memory in which stored waypoints for a second route to be covered by the vehicle are stored, a path calculator configured to convert the first and / or second path into a common coordinate system, and a fusion unit configured to merge the first path and the second path and, based on the merged path thus obtained, to supply a signal to a steering controller in order to steer the vehicle along the merged path and / or to generate waypoints calculated from the merged path for the positions travelled by the vehicle in to store in a memory.

[0012] It is proposed that the first path and / or the second path include or include more than two waypoints.

[0013] In other words, a first path is recorded by the visual sensor system and the image processing system based on features optically detected in the field, and a second path is read from a memory. Both paths are converted into a common coordinate system and merged in order to generate a steering signal for a steering controller based on the merged path obtained in this way. The steering control can display the steering signal to an operator for manual steering or feed it to a steering actuator. Alternatively or additionally, waypoints for positions traveled by the vehicle, calculated based on the merged path, can be stored in a memory in order to be able to retraverse the stored path later or to plan a different path based on it. One or both of the paths comprises more than two waypoints and can therefore also represent a curvature.

[0014] In this way, the accuracy of the steering signal or the stored path is improved because it is based on an enlarged database that also takes into account any possible curvature of the path.

[0015] To avoid the problem of the image processing system jumping between series of features, it is proposed that the path calculator or the image processing system be configured to select features of the field used to determine the first path based on the second path. Example

[0016] The drawings illustrate an embodiment of the invention, described in more detail below. They show: Fig. 1 a side view of an agricultural vehicle with an arrangement for automatically guiding the vehicle across a field, Fig. 2 a top view of the vehicle working in the field, Fig. 3 a schematic diagram of the arrangement for guiding the vehicle, Fig. 4 a flow chart showing the arrangement of the Fig. 3 works, Fig. 5 a plan view of the vehicle working the field with a curved path, and Fig. 6 a flowchart for selecting the visually tracked features.

[0017] The Fig. Figure 1 shows a side view of a self-propelled agricultural vehicle 10 in the form of an agricultural tractor and an implement 12 in the form of a field sprayer mounted on a three-point linkage 14 of the vehicle 10. The vehicle 10 is built on a supporting frame 16, which is supported on steerable front wheels 18 and drivable rear wheels 20 and carries a cabin 22 in which an operator's workplace 24 is located. The operator's workplace 24 includes a steering wheel 26, a seat 28, pedals (not shown), and an input and output unit 30. In the following, directional references such as front and rear or sideways refer to the forward direction of the vehicle 10, which in the Fig. 1 runs to the left.

[0018] The vehicle 10 comprises an arrangement for automatically guiding the vehicle across a field, which arrangement comprises a visual sensor system 62 with a camera 32 designed as a mono or stereo camera (and / or a scanning, active sensor, e.g. a lidar system), an image processing system 36 connected to the image sensor of the camera 32 in a signal-transmitting manner, a navigation unit 48 mounted on the roof of the cabin 22, a steering signal calculation unit 46 and a steering control 34.

[0019] The navigation unit 48 is configured to receive signals from a satellite-based navigation system (GNSS, such as GPS, Galileo, etc.). The navigation unit 48 can be equipped with an inertial navigation system to improve the accuracy of the position signals, particularly under unfavorable reception conditions. Furthermore, it can be connected to a receiver for receiving correction signals from a corresponding system (DGPS, RTK) to improve the precision of the navigation system signals, and / or to odometric sensors that detect the steering direction and movement of the front wheels 18, i.e., a steering angle sensor 44 and a speed sensor for detecting the rotation of the front wheels. The odometrically detected signals are typically merged with those of the inertial navigation system and the satellite-based, possibly corrected signals, e.g., using a Kalman filter.The navigation unit 48 outputs a signal at regular intervals regarding its two- or three-dimensional position and optionally the speed and / or orientation.

[0020] The image processing system 36, the navigation unit 48, the steering signal calculation unit 46 and the steering control 34 are interconnected via a bus system 50, which can operate, for example, according to the protocol according to ISO 11783 or a successor thereto.

[0021] The steering signal calculation unit 46 is spatially integrated into an input and output unit 30, which is provided with a user interface in the form of a display unit 54 and an input device 56. The input and output unit 30 can fulfill the function of a virtual display according to ISO 11783.

[0022] The steering control 34 is connected to a valve 38 for controlling a steering actuator 40 and to the steering angle sensor 44 for detecting the current steering angle of the front wheels 18.

[0023] In the Fig. Figure 2 shows a plan view of the vehicle 10 working in a field 58 containing rows of crops 60. The steering control 34 is controlled by the steering signal calculation unit 46 based on the (fused) signals from the visual sensor system 62 (camera 32) and the navigation unit 48 such that the front wheels 18 and the rear wheels 20 travel between the rows of crops 60 without driving over the crops 60. The visual sensor system 62 can detect the plants 60 (cf., for example, US 6 445 983 B1, EP 1 266 553 A2, WO 2008 / 150418 A1, WO 2018 / 067473 A1, Wan et al., loc. cit.), and / or detect lanes or tracks between the plants 60, cf. DE 10 2016 209 437 A1. The implement 12 could also be any other implement 12, e.g., a hoe used to remove unwanted weeds between the rows of plants 60.In another embodiment, the implement could be a soil cultivation implement, such as a plough or harrow, and the visual sensor system 62 could detect the boundary between the uncultivated and cultivated field 58. In yet another embodiment, the vehicle 10 could be a harvester with a harvesting header, or it could move a harvesting implement, e.g., a mower, across the field, and the visual sensor system 62 could detect a crop edge of the unharvested crop or a crop swath to be picked up by a pickup.

[0024] The Fig. Figure 3 shows a schematic of the arrangement for automatically guiding the vehicle 10. The visual sensor system 62 with the camera 32 feeds the image signal from the camera 32 to the image processing system 36. Another, optional camera 64, which is arranged on the underside of the vehicle 10 and can be designed as a stereo or mono camera or any other camera, e.g., a three-dimensional PMD camera, feeds its image signal to the image processing system 36. Either only one camera 32 or 64 is active at the same time, or both are active, and their signals are combined and merged in a path calculator 70.

[0025] The image processing system 36 calculates, in a manner known per se, based on the image signals of the camera 32 and / or 64 and the features of the field 58 contained therein and identified by the image processing system 36, a (first) path constructed from several (more than two) waypoints, which the vehicle 10 is to travel in the future. For this purpose, as above, with regard to the Fig. 2. This path, constructed from waypoints, does not necessarily have to be straight, but can have any shape, ie, include one or more bends.

[0026] The first path is calculated using a coordinate system related to the vehicle 10, the origin of which may, for example, be within the camera 32, 64. Typically, the path is calculated in two-dimensional coordinates, although it would also be conceivable to use three-dimensional coordinates (including elevation information). In this case, the images from cameras 32 and / or 64 are processed sequentially and compared with each other while traveling the path, so that individual features are tracked across multiple images ("frames").

[0027] The first path calculated in this way using the signals from camera 32 or 64 is fed to a path calculator 70, which also receives signals from a location calculator 68. The location calculator 68 receives signals regarding the current position from the navigation unit 48, which are converted by the location calculator 68 into the position of the vehicle's reference point (i.e., camera 32 and / or 64, see the previous paragraph). The path calculator 70 outputs the first path, which is now based not on the coordinate system used by the image processing system 36, which is related to a point on the vehicle 10, but on absolute, georeferenced coordinates. If both cameras 32, 64 are active, the path calculator 70 can merge their signals into a single path in a conventional manner.At the output of the path calculator 70, there is therefore a first path consisting of several waypoints in a georeferenced coordinate system (in “world coordinates”).

[0028] The output signal of the path calculator 70 is fed to a fusion processor 74, which also receives data from a memory 72 in which stored waypoints for a second path to be covered are stored. These waypoints for the second path may have been generated during a previous processing of the field 58, e.g., during the sowing of the plants 60, or they may have been recorded during the ongoing processing of the field 58 during a previous pass adjacent to the current pass and are shifted sideways by the working width or a multiple thereof (cf. US 2009 / 0204281 A1). The fusion processor 74 retrieves the stored waypoints from the memory 72 based on the current position of the navigation unit 48. The second path may also contain more than two waypoints, i.e., include one or more bends, although in the simplest case, the second path may be a straight line (a so-called AB line).

[0029] Accordingly, the merger 74 has two paths, i.e., sets of waypoints, that refer to one and the same coordinate system, namely the global coordinate system. The first set of waypoints is based on the visual sensor system 62 (camera 32 and / or 64), and the second set of waypoints is based on the path stored in memory 72, which is retrieved based on the signals from the navigation unit 48. Both paths define the trajectory that a defined reference point of the vehicle 10 is to travel, e.g., the center of the rear axle.

[0030] In another embodiment, it would be conceivable to use a coordinate system for the first and second path that refers to any reference point on the vehicle 10, e.g. the center of the rear axle of the vehicle 10 or the navigation unit 48 or the camera 32, 64 (cf. Wan et al., ibid.).

[0031] According to the present invention, a plurality (more than two) of waypoints from one or both paths are provided and merged, instead of only taking into account the lateral deviation of the vehicle 10 to be steered from the path to be traveled and the angle between the longitudinal axis of the vehicle 10 and the longitudinal axis of the path to be traveled, as in the prior art (Wan et al., a.aO), i.e., approximating the two paths as straight lines. The procedure according to the invention has the particular advantage that a more precise steering signal can be provided in the case of curvatures in the paths, since the curvature of the path is also taken into account during the merging, as in the example of Fig. 5. There, all plants 60 lying within the field of view of the camera 32 would be detected and used to calculate a first path, the points of which correspond to the positions of the respective plants 60. The second path would be in the example according to Fig. 5 are also curved. Such curved paths can be caused by the shape of the field or obstacles located there that must be avoided. In previous steering systems (Wang et al., op. cit.), however, the curvature in both paths would be approximated as a constantly recalculated straight line, resulting in a less precise and sometimes erratically changing steering signal.

[0032] During fusion, the fusion unit 74 can take into account the accuracies of the individual points of the paths. The respective accuracies can be calculated separately for each point. In the visual sensor system 62, the accuracy can be determined in a conventional manner based on the quality of the image signals from the camera 32, 64 (cf. the prior art cited above), but separately for each feature detected in the image.

[0033] Naturally, more distant features, such as plants, are assigned a lower accuracy than closer features. The situation will be similar for the signals from navigation unit 48, since positions further in the future are subject to greater uncertainty than immediate positions. By using a common coordinate system and a higher number of waypoints than the previous two, the accuracies associated with the different waypoints can also be considered separately.

[0034] The path calculator 70 (or the image processing system 36) is in particular capable of selecting features (e.g. rows of plants 60) of the field 58 used to determine the first path based on the second path, as shown in the Fig. 6, which illustrates aspects of step 102. For this purpose, the path calculator 70 (or the image processing system 36, which receives corresponding information from the path calculator 70) can recognize a plurality of features arranged laterally next to one another in the field (e.g., the rows of plants) in the first path (step 120) and, based on a comparison between the first and second paths, in the event of an unexpected, sudden change in the first path (step 122), can recognize a sudden change in the image processing system 36 between features arranged laterally next to one another that serve to determine the first path and, if necessary, correct it based on the second path (step 124).

[0035] In this way, possible jumps in the first, visually detected path can be detected and compensated. Therefore, if the first path determined based on the visual sensor system 62 contains sudden jumps indicating that the image processing system 36 has jumped one or more rows to the left or right, i.e., considers a different row of plants 60 than previously as a reference row to be tracked and brought into lateral alignment with a reference point of the vehicle 10 (which, for example, represents the row that should correspond to the center of the vehicle 10 or another reference point of the vehicle 10), the second path to be followed, provided by the memory 72, can serve to inform the path calculator 70 (and / or the image processing system 36) which of the rows of plants 60 is to be considered as the reference row, as shown in the Fig. 3 represented by the arrow 78.

[0036] The procedure described in the two preceding paragraphs is an invention independent of the subject matter of claim 1, because the automatic selection of the path to be followed by the image processing system 36 using the navigation system 48 is not limited to the fact that the paths coming into the fusion contain more than two points, ie can be curved, but would also be applicable if both paths consist of straight lines, as in the prior art.

[0037] The fusion unit 74 therefore calculates a fused path based on both paths and, based thereon, the lateral deviation of the vehicle 10 from the path and the angle between the longitudinal axis of the vehicle 10 and the longitudinal axis of the path, using the signals from the navigation unit 48 and / or the visual sensor system 62. The target value for the path to be traveled is the fused path, and the actual position and direction of the vehicle 10 are provided based on the fused signals from the navigation unit 48 and the visual sensor system 62. This occurs in a similar manner to the fusion of the two paths, but here serves to determine the deviation of the lateral position and orientation of the vehicle from the fused path. The errors resulting from the comparison (lateral deviation, angle) are fed by the fusion unit 74 to an external control loop 76, which in turn feeds a steering error signal to the steering control 34, i.e.a signal representing any deviation between the current steering angle of the front wheels 18 and the steering angle appropriate for traveling the path. The steering controller 34 controls the steering actuator 40 and receives feedback signals from the steering angle sensor 44. The steering controller 34 can be implemented in a conventional manner as a closed-loop controller (taking into account feedback values ​​for a variable to be set, here the steering angle), e.g., as a PID controller. Analogously, the outer control loop 76 can also be implemented as a closed-loop controller and receive a feedback value regarding the steering angle of the front wheels 18 from the steering controller 34.

[0038] This procedure can be used not only when actually crossing the field, but also when turning on the headland. Useful information can be expected from the visual sensor system 62 at least when there are tracks or tramlines from previous processing operations, so that steering can also be carried out on the headland based on the fused signals from the visual sensor system 62 and the signals from the navigation unit 48. If useful features for the visual sensor system 62 are missing on the headland, the visual sensor system 62 can provide signals again at the latest when the actual field is within its field of vision, thus facilitating merging into the next track (cf. EP 2 954 772 A1 or EP 3 146 825 A1).

[0039] In one possible embodiment, the waypoints calculated by the fusion unit 74 in step 108 for the positions traveled from the reference point of the vehicle 10 are stored in the memory 72 so that they can later be used by the vehicle 10 or any other vehicle (e.g., a harvester) as a second route to retraverse the route or to base route planning on them. This makes it possible, for example, to drive across the field for the first time after a sowing operation after the plants have emerged based on the signal from the camera 32, which can optionally be fused with a position of the plants stored during the sowing operation, and to map the route in the process. The mapped route can be used as a reference route for planning during subsequent passes when the plants have grown larger and are possibly harder to recognize.

[0040] As in the Fig. As shown in Figure 3, the location calculator 68, the path calculator 70, the memory 72, the merger 74, and the outer control loop 76 are integrated into the steering signal calculation unit 46. They can be implemented as separate hardware components or are formed by one or more processors and software running thereon of the steering signal calculation unit 46 or the input and output unit 30.

[0041] The Fig. Figure 4 shows a flow chart showing the arrangement of the Fig.1 to 3. After starting in step 100, a first path is determined in step 102 using the visual sensor system 62 and converted into a global coordinate system in step 104. In step 106, a second path is determined using the memory 72 and the navigation unit 48, and both paths are merged in step 108. Based on this, a steering signal is generated in step 110, and the system continues with step 102 as long as field 58 has not yet been processed (step 112). QUOTES CONTAINED IN THE DESCRIPTION

[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This 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

[0000] US 6 445 983 B1 [0005, 0023] EP 1 266 553 A2 [0005, 0023] WO 2008 / 150418 A1 [0005, 0023] WO 2018 / 067473 A1 [0005, 0023] DE 10 2019 203 651 A1

[0007] DE 10 2016 209 437 A1

[0023] US 2009 / 0204281 A1

[0028] EP 2 954 772 A1

[0038] EP 3 146 825 A1

[0038] Zitierte Nicht-Patentliteratur

[0000] Wan, X., Liu, G. (2008). Automatic Navigation System with Multiple Sensors

[0005] Li, D. (eds) Computer and Computing Technologies in Agriculture, Volume II. CCTA 2007. The International Federation for Information Processing, vol 259. Springer, Boston, Ma.

[0005] F. Rovira-Más et al., Kalman filter for sensor fusion of GPS, The 2006 ASABE Annual International Meeting, St. Joseph, Michigan, USA, July 9-12, 2006

[0007] ISO 11783

[0021]

Claims

[1] Arrangement for guiding an agricultural vehicle (10) across a field (58), comprising: a visual sensor system (62) directed at the field (58) and connected to an image processing system (36) for signal transmission, which is configured to identify features of the field (58) in the image signal of the visual sensor system (62) and to calculate a first path to be traveled by the vehicle (10) based on the identified features, a navigation unit (48) for determining the position of a reference point on the vehicle (10), a memory (72) in which stored waypoints for a second route to be covered by the vehicle (10) are stored, a path calculator (70) configured to convert the first and / or second path into a common coordinate system, and a fusion unit (74) configured to merge the first path and the second path and, based on the merged path thus obtained, to supply a signal to a steering controller (34) in order to steer the vehicle (10) along the merged path and / or to store waypoints calculated from the merged path for the positions traveled by the vehicle (10) in a memory (72), characterized by that the first path and / or the second path includes more than two waypoints. [2] Arrangement according to the preamble of claim 1, wherein the path calculator (70) or the image processing system (36) is configured to select features of the field (58) used to determine the first path based on the second path. [3] Arrangement according to claim 2, wherein the path calculator (70) or the image processing system (36) is configured to detect a plurality of laterally adjacent features on the field (58) in the first path and to detect a sudden change between laterally adjacent features in the first path which serve to determine the first path and, if necessary, to correct it using the second path. [4] An arrangement according to claim 3, wherein the features are plants (60) arranged in rows. [5] An arrangement according to any one of the preceding claims, wherein the first and / or second path comprises one or more bends. [6] Arrangement according to one of the preceding claims, wherein the path calculator (70) is configured to convert the first path provided by the image processing system (36) into a coordinate system that is georeferenced or referenced to a reference point on the vehicle (10) based on the position determined by the navigation unit (48). [7] Arrangement according to one of the preceding claims, wherein the fusion unit (74) is configured to calculate, based on the fused path and the signals from the navigation unit (48) and / or the visual sensor system (62), a lateral deviation of the vehicle (10) from the fused path and a deviation of the forward direction of the vehicle (10) from the direction suitable for traveling along the fused path and to supply both deviations to the steering control (34). [8] Arrangement according to one of the preceding claims, wherein the fusion unit (74) is configured to take into account the waypoints of the first and second paths in a weighted manner based on their accuracy during the fusion. [9] Arrangement according to one of the preceding claims, wherein the steering control (34) is configured to control a steering actuator (40). [10] Method for guiding an agricultural vehicle (10) across a field (58), comprising the following steps: a visual sensor system (62) directed at the field transmits signals to an image processing system (36), which identifies features of the field (58) in the image signal of the visual sensor system (62) and calculates a first path to be traveled by the vehicle (10), a navigation unit (48) determines the position of a reference point on the vehicle (10), stored waypoints for a second route to be covered by the vehicle (10) are stored in a memory (72), a path calculator (70) converts the first and / or second path into a common coordinate system, and a fusion unit (74) merges the first path and the second path and, based on the merged path thus obtained, supplies a signal to a steering control unit (34) to steer the vehicle (10) along the merged path and / or stores waypoints calculated from the merged path for the positions travelled by the vehicle (10) in a memory (72), characterized by that the first path and / or the second path comprises more than two waypoints and / or that the path calculator (70) or the image processing system (36) is configured to select features of the field (58) used to determine the first path based on the second path.

Images