This application is based upon and claims priority to, under relevant sections of 35 U.S.C. § 119, German Patent Application No. 10 2020 130 280.0, filed Nov. 17, 2020, the entire contents of which are hereby incorporated by reference.
This disclosure relates to a method for the identification of objects in a warehouse by means of an industrial truck, as well as an industrial truck configured to carry out the method.
Industrial trucks generally comprise a plurality of sensors for the identification of their environment, such as for example laser scanners. Objects in the environment of the industrial truck may be identified by means of these sensors and the industrial truck may be steered on the basis of the identified objects, for example the industrial truck may be steered around the obstacle in the case of an object identified as an obstacle. For example, a load carrier to be picked up, such as a pallet, may also be identified in this manner and subsequently picked up by the industrial truck.
An identification method which is relevant thereto is disclosed in EP 3 316 181 A1. In this method, on the basis of data recorded by a 3D camera, a view of an object located in the vicinity of the industrial truck is generated and compared with a template, i.e. a predefined pattern. The template may be stored in an evaluation apparatus of the industrial truck. During this so-called template matching it is possible to monitor whether the image recorded by the 3D camera, for example of a pallet, sufficiently corresponds to the stored image of the pallet as a template. In the case of sufficient correspondence, the object is identified as a pallet. The industrial truck may then pick up the pallet, in particular automatically, by driving the load-carrying fork of the industrial truck into the pallet pockets and subsequently lifting.
In known methods it may arise that the object identification fails. This is the case, for example, when the detected view of the pallet does not sufficiently correspond to the template during the described template matching. As the inventors have recognized, this occurs in particular when a load arranged on the pallet is wrapped around with a partially transparent film wrapping. Films are frequently used in logistics for fixing loads. At the same time, it regularly occurs that the film wrapping also at least partially conceals the pallet pockets. The recorded image of the pallet then potentially deviates too greatly from the template.
During the course of the object identification, for example, instead of a pallet a free space may also be identified inside a storage rack as the object. Such an identification may also fail when, for example, films from a rack level located thereabove hang into the free space located therebelow. A partially transparent strip curtain, by which the different warehouse regions may be separated from one another, may also prevent objects which are located behind the strip curtain from being successfully identified as an obstacle. In particular, this strip curtain may constitute an apparent obstacle even if it might be possible to drive through said strip curtain when used as intended.
Obstacle identification in road traffic by means of LIDAR systems is also disclosed, for example, in WO 2019/164959 A1, WO 2019/226487 A1 or WO 2019/222684 A1. These LIDAR systems emit laser pulses and evaluate the echoes over the course of time or, respectively, reduce the signal-to-noise ratio by repeated measurements. Thus, obstacles which are concealed by fog or smoke are also intended to be identified.
Proceeding from this prior art, the object of the invention is to provide a method or, respectively, an industrial truck by which objects may be identified in a more reliable manner in a warehouse. In particular, the objects are intended to be identified in spite of being concealed by at least partially transparent obstacles.
An embodiment of a method for the identification of objects in a warehouse by means of an industrial truck comprises emitting an identification signal by means of a transmitting apparatus of the industrial truck. The identification signal reflected by objects in the environment is detected as a signal echo by means of a receiving apparatus of the industrial truck. A solid angle and a receiving time is assigned to each of the signal echoes and then monitored to see whether the same solid angle has been assigned to a plurality of signal echoes. If a plurality of signal echoes have been assigned to a solid angle, then the signal echo which arrived last is evaluated in the receiving apparatus according to the receiving time for this solid angle for the identification of the object. The object is then identified on the basis of the signal echoes.
An embodiment of an industrial truck is disclosed that is configured to employ the method of identifying objects in a warehouse. The method according to the invention may thus be carried out by the industrial truck according to the disclosure. Descriptions made relative to the method accordingly apply to the industrial truck, in particular to the evaluation apparatus, and vice versa.
In an embodiment, the industrial truck comprises a transmitting apparatus for emitting an identification signal, a receiving apparatus for detecting the identification signal as a signal echo reflected by objects in the environment, as well as an evaluation apparatus. The evaluation apparatus is configured to assign a solid angle and a receiving time to each of the signal echoes, to monitor whether the same solid angle has been assigned to a plurality of signal echoes. When a plurality of signal echoes have been assigned to a solid angle, the evaluation apparatus is configured to evaluate the signal echo which arrived last in the receiving apparatus according to the receiving time for this solid angle, i.e. to form the basis of the identification of the object, to identify the object on the basis of the signal echoes, and to steer the industrial truck on the basis of the identified object.
In an embodiment, an object identification takes place by emitting identification signals and evaluating signal echoes. The identification signal may be, for example, an optical signal, in particular a laser signal. The transmitting apparatus and receiving apparatus may be part of a common transceiving apparatus, in particular a sensor unit. For example, a LIDAR (“light detection and ranging”) system may be provided as the transceiving apparatus, which outputs two-dimensional environmental data or even depth information comprising three-dimensional environmental data. A “time of flight” camera may also be provided. The identification signal, however, may also be an acoustic signal, in particular an ultrasonic signal. As a result, the environment may be detected by means of an ultrasound sensor system.
In an embodiment, a solid angle as well as a receiving time are assigned to each of the signal echoes received by the receiving apparatus. Thus, a view of the environment of the industrial truck, in particular an object located in the environment of the industrial truck, may be generated. For example, it may be a two-dimensional view. A three-dimensional view may also be generated by considering distance information. Distance information may be derived, in particular, by considering the transmission time of the identification signal as well as the running time of the signals.
The inventors have recognized that partially transparent obstacles in the environment of the industrial truck may lead to a plurality of signal echoes appearing at one and the same solid angle. Thus, when the emitted identification signal encounters the partially transparent obstacle, a part of the signal may be reflected and thus arrive in the receiving apparatus as a first signal echo, whilst a part of the identification signal transmitted by the partially transparent obstacle is reflected on an object located therebehind, for example a pallet, and is received in the receiving apparatus as a second signal echo coming from the same solid angle and arriving later. In the case of the known method described in the introduction for the identification of objects, this may lead to the partially transparent object being identified rather than the object located therebehind. If, for example as described in the introduction, a pallet is intended to be identified as the object, the film possibly hanging in front of the pallet pockets for fixing the load arranged on the pallet may prevent the identification of the pallet located therebehind, viewed from the industrial truck.
In the case where the same solid angle has been assigned to a plurality of signal echoes, the solution according to the invention provides that the signal echo, which arrived last in the receiving apparatus for the respective solid angle, is evaluated for the identification of the object. The evaluation of the signal echo which arrived last for the respective solid angle for the identification of the object may mean that the last signal echo is assigned to the object. In particular, the last signal echo may be exclusively assigned to the object. Previously arrived signal echoes for the same solid angle may be ignored during the evaluation, for example. Moreover, previously arrived signal echoes may form the basis of an identification of the partially transparent object, as is described below in more detail. Moreover, the evaluation of the signal echo which arrived last for the respective solid angle for the identification of the object may mean that for this solid angle a signal echo which previously entered the receiving apparatus is replaced or that none of the signal echoes which entered the receiving apparatus at the same solid angle is assigned to the object. Thus, for example, during the course of the evaluation a pixel based on a previous signal echo may be deleted on the basis of a further signal echo at the same solid angle, as is described below in more detail. By the evaluation of the signal echoes which arrived last for the identification of the object, therefore, a pallet may also be reliably identified, for example, if the pallet pockets thereof are partially or entirely covered by a transparent film. The transparent film is “seen through”. The industrial truck may then be steered on the basis of the object thus identified.
Thus, for example, in the case of an identified pallet the industrial truck may drive, in particular fully automatically, with the load-carrying fork into the pallet pockets and at the same time puncture the transparent film with the fork tines. In addition, for example, when identifying a free space in a rack, transparent film protruding into the free space may be ignored. A partially transparent strip curtain may also be ignored and thus driven through by the industrial truck, in particular fully automatically. The industrial truck is able to see through the strip curtain and reliably identify objects located therebehind, for example the substrate to be driven over by the industrial truck or a potential obstacle, such as another industrial truck. The solution according to the invention is particularly simple and efficient due to the assignment of the signal echo from the object which in principle arrived last, in particular compared to methods which carry out an assignment via the comparison with known pulse shapes of the signal echoes. Thus, for the solution according to the invention, no knowledge is required about known pulse shapes. Indeed, the accurate pulse shape of the signal echo may be irrelevant, since instead it depends on the receiving time.
In an embodiment, the receiving apparatus may, in particular, permanently measure and thus detect a plurality of successive signal echoes. The receiving apparatus may be activated for a predetermined time period after the emission of the identification signal by the transmitting apparatus. Thus, according to one embodiment it may be provided that only signal echoes which arrive in the receiving apparatus within a predetermined time period after the emission of the identification signal are used or evaluated for the identification of the object. According to a further embodiment which is relevant thereto, the predetermined time period may be based on a maximum measuring distance to be anticipated. This increases the probability that the signal echoes arriving at the receiving apparatus are also actually signal echoes of the emitted identification signals. This is relevant, in particular, when the signal echoes which arrived last at the receiving apparatus form the basis of the object identification according to the invention. The limit to a time period dependent on a maximum measuring distance to be anticipated may prevent measuring signals arriving later (caused for example by repeated reflections) from forming the basis of the object identification, which could lead to a faulty object identification. In particular, the object actually to be identified could be overlooked.
According to one embodiment, the object is identified by generating a view of the object on the basis of the signal echoes and comparing the view with a template, wherein in the case of sufficient correspondence between the view and the template the object is regarded as identified. As a result, as mentioned in the introduction, a template matching may be carried out. Thus, for example, a view of a pallet front may be generated from the signal echoes. This image of the rack front may then be compared with a stored template, i.e. a pattern image, of a pallet front by means of the evaluation apparatus of the industrial truck. With sufficient correspondence between the view and template, the industrial truck identifies the pallet and is able to pick up this pallet. The view may be generated in this case by the respective last signal echoes being assigned to the object, previous signal echoes at the same solid angle, however, being ignored as already discussed above. The object, therefore, may exclusively be identified on the basis of the signal echoes which arrived last. Moreover, previously arrived signal echoes may additionally form the basis of the identification of the object. Thus, for example based on the first signal echoes, initially pixels may be generated for the view, in particular an image of the entire rack front. Subsequently, for the solid angles at which a further last signal echo arrives, the respective pixel may be deleted or, respectively, replaced.
According to an embodiment which is relevant thereto, the view of the object is generated by creating a two-dimensional data set by projecting pixels onto a projection plane. The signal echoes detected by the receiving apparatus may in this case be understood as image data which comprise a plurality of pixels. Point clouds are also referred to here. These pixels are projected onto the projection plane. The projection plane may be preferably selected such that it corresponds to an assumed arrangement of the object to be identified. The pixels distributed over the projection plane may be understood as a two-dimensional view of the object. A three-dimensional object identification is also possible. In this case, according to a further embodiment it may be provided that a further data set which comprises a depth profile along a line at the position of the projection plane is created, and that the further data set is compared with a predefined depth profile pattern assigned to the object, wherein the object is finally identified when a sufficient correspondence is established during this comparison. According to this embodiment, therefore, a further template matching takes place. The reliability of the identification thus may be increased. Relative to further details for such a template matching, reference is made to EP 3 316 181 A1.
Should the signal echo which arrived last, i.e. the signal echo generated by an object located behind an assumed transparent object, only differ slightly from a signal noise, an anticipated value may be calculated and used when selecting the respective signal echo. The anticipated value may be based, for example, on the geometry of a load carrier. If this geometry is assisted—even if only slightly—by a plurality of adjacent depth values, the corresponding object is thus regarded as identified. The anticipated value may also be provided by the warehouse floor being illustrated.
According to one embodiment, if the same solid angle has been assigned to a plurality of the signal echoes, one of the signal echoes which previously arrived in the receiving apparatus may be assigned to a second object and this object is identified as partially transparent. As described, with the identification of the first object, the signal echoes which arrived last in the receiving apparatus are evaluated, so that previously arrived signal echoes caused by partially transparent obstacles are ignored. According to this embodiment, however, the partially transparent obstacle as such is also identified, namely by evaluating the previously arriving signal echoes. In particular, the signal echo which arrived first in the receiving apparatus may be assigned to the second object and this object thus identified as partially transparent. The identification of the second object as a partially transparent object may in this case encompass, in particular, that in principle in the case of a plurality of signal echoes, those first arriving are assigned to the partially transparent object. In particular, a view of this second partially transparent object may also be created. Thus, in addition to the aforementioned point cloud for the first object, the evaluation apparatus may create a second point cloud for the second transparent object. Thus, the transparent obstacle may be specifically identified and, if desired, removed in a targeted manner. In principle, therefore, the partially transparent object may be classified more accurately, for example it may be established whether it is a film wrapping of goods or a strip curtain. For example, it may be monitored, when the partially transparent object is a film wrapping of goods, whether the fixing of the goods has given way. Moreover, in principle, further views may be generated, in particular, as point clouds. Thus, a first point cloud could be based on the strongest signal echo and a second point cloud on the second strongest echo. Moreover, a first point cloud could be based on the signal echo furthest to the front when viewed from the industrial truck, whilst a second point cloud could be based on the signal echo located therebehind. In this sense, third, fourth or even further point clouds may also be provided.
According to one embodiment, it may be provided that a speed of the industrial truck is adapted if the object is identified as a strip curtain. In this sense, the industrial truck may be steered on the basis of the identified second object. Thus, the speed of the industrial truck may be reduced so that the industrial truck drives carefully through the strip curtain. Thus, damage to the strip curtain and also the load transported by the industrial truck may be avoided. Moreover, such a speed reduction is expedient since when driving through the strip curtain the object identification of the industrial truck is potentially impaired by the strip curtain. If the partially transparent object, for example, were to be identified as a film wrapping of goods stored on a pallet, the industrial truck is able to puncture this film wrapping when driving the load-carrying fork into the pallet pockets, as already mentioned. An adaptation of the speed of the industrial truck, in particular a reduction of the speed, may also be provided here. This is expedient, in particular, when it is a relatively lightweight load, which could be otherwise displaced by means of the load-carrying fork by puncturing the film in front of the pallet pockets. For example, the industrial truck may be informed that it is a lightweight load via a warehouse management system.
According to one embodiment, the individual movement of the industrial truck is considered when identifying the object. For example, a direction of travel, a driving speed, an acceleration and/or a deceleration of the industrial truck may be considered as an individual movement for the identification of the object. In particular, an object identification may take place repeatedly, for example, at different positions which the industrial truck adopts during the course of its individual movement. Thus, according to one embodiment, a first view of the object is generated on the basis of signal echoes detected in a first position of the industrial truck and compared with a first template, wherein a second view of the object is generated on the basis of signal echoes detected in a second position of the industrial truck and compared with a second template. The object is then regarded as identified when a sufficient correspondence is present both between the first view and the first template and between the second view and the second template. According to this embodiment, therefore, in particular an image sequence with two or more views of the object may be generated and compared with corresponding templates. It may be concluded with greater reliability therefrom whether the signal echoes are caused by a partially transparent object or by a non-transparent object. Thus, possibly occurring multipath reflections, which otherwise may lead to a faulty object identification, may also remain unconsidered. Thus, in the case of multipath reflections it may also lead to a plurality of signal echoes at the same solid angle. With a movement of the industrial truck, however, these multipath reflections generally no longer occur or at least no longer occur at the same solid angle. A partially transparent second object with a non-transparent first object located therebehind, however, will also lead to the arrival of at least two signal echoes at the same solid angle. By considering the individual movement of the industrial truck, it may also be identified whether the alleged signal echo is possibly only a noise. For example, a relatively weak signal echo may also be made plausible by this embodiment and thus distinguished from the noise. If the pixel resulting from the signal echo moves in harmony with the individual movement of the industrial truck, it is probably an object worthy of identification and not noise.
The industrial truck may be steered on the basis of the identified object, as already discussed. Thus, according to one embodiment, the industrial truck may be automatically oriented on the basis of the identified object. As already discussed, the industrial truck for example may drive automatically with the load-carrying fork into pallet pockets of the pallet, even if these pallet pockets are concealed by a partially transparent film. Moreover, the industrial truck could automatically drive through a partially transparent strip curtain. Partially automatic steering is also conceivable. Thus, according to one embodiment, an orientation or movement of the industrial truck to be carried out or to be initiated by an operator of the industrial truck on the basis of the identified object may be optically and/or acoustically displayed to the operator. The industrial truck may comprise corresponding display devices. For example, it could be displayed to the operator on a display how the load-carrying fork is to be oriented, in order to permit picking up a pallet. An acoustic feedback could be implemented in the case of a correct orientation. In a partially automated case, it could be optically or acoustically displayed that a pallet has been identified and that the industrial truck is ready for the corresponding orientation. An automatic orientation of the industrial truck could thus be carried out by an interaction of the operator with the industrial truck, for example by holding down a corresponding operating element.
According to one embodiment, environmental elements are considered for the identification of the object. Thus, it could be concluded from the environment in which the object generally is to be anticipated whether the known signal echoes are actually the object. Thus, for example, the wall behind a storage area may be an indication that detected signal echoes actually originate from a pallet located on the storage area. The wall may also provide an indication as to where a corresponding signal echo is highly likely to be found.
The invention is described hereinafter with reference to figures, in which:
Unless indicated otherwise, the same reference numerals denote the same objects hereinafter.
On the basis of the signal echoes received by the sensor unit 18 an evaluation apparatus 20 arranged in the drive part 12 of the industrial truck 10 generates views of objects located in the environment of the industrial truck. The evaluation apparatus 20 compares these views with stored templates, wherein in the case of sufficient correspondence between the respective view and the respective template, the object forming the basis of the template is regarded as identified. Such a template matching is described in detail in EP 3 316 181 A1.
In
This is shown in
As the inventors have recognized, the partial transparency of the film 25 leads to repeated signal echoes at the same solid angle. As shown in
In
However, the pallet 22 may also be already identified successfully here. In this arrangement the second echo E2 results from a reflection of the identification signal on a lower face of the pallet top deck. Accordingly, the view D from
In
Number | Date | Country | Kind |
---|---|---|---|
10 2020 130 280.0 | Nov 2020 | DE | national |