Method and apparatus for determining radio tag transits radio tag detection gates

Abstract

A method is proposed for determining which radio tag (RF) passes through radio tag detection gates (A, B, C, D, E, F, G) within a monitored area (UB). Multiple radio tag detection gates (A, B, C, D, E, F, G) are disposed within the monitored area (UB), and each radio tag detection gate (A, B, C, D, E, F, G) has one or more antennas for receiving radio signals from radio tags (RF) within the monitored area (UB). The method comprises the following steps: a) receiving radio signals from radio tags (RF) through at least one of the multiple radio tag detection gates (A, B, C, D, E, F, G); b) detecting the signal strength curve of each radio signal received by at least one of the multiple radio tag detection gates (A, B, C, D, E, F, G); and c) determining, based on the detected signal strength curve, which of the multiple radio tag detection gates (A, B, C, D, E, F, G) the radio tag (RF) passed through.

Description

Technical Field

[0001] The present invention relates to a method for determining passage through a radio tag detection gate, and an apparatus for determining passage through a radio tag detection gate. Background Technology

[0002] To monitor or detect goods, merchandise, or other items in a warehouse, radio tags, such as RFID transponders or RFID tags (RFID: Radio Frequency Identification), can be used. These radio tags can be read by a corresponding reading device using high-frequency radio waves. In many cases, the radio tag is powered by the radio waves generated by the reading device, thus requiring no power source of its own. Radio tags allow for simple identification of objects with corresponding tags and the transmission of information between the radio tag and the reading device.

[0003] RFID technology enables the identification of radio tags approaching a transmitting / receiving station (e.g., a so-called RFID gate or radio tag detection gate). Such radio tag detection gates typically have multiple antennas that receive radio signals from the radio tags and are capable of predicting the location and / or passage of objects carrying radio tags.

[0004] In existing systems, such as those using machine learning (ML), the association between antenna measurements and the location of radio tags is identified in a device consisting of one or more RFID gates and radio tags moving within them (the radio tags being attached to moving objects). In this case, a model is trained in a machine learning module using a training architecture consisting of antennas and radio tags. These models can then be used, for example, in different warehousing systems to similarly determine the location of radio tags.

[0005] However, in these systems, location and / or passage predictions are created based on data from a single gate; that is, only data received from the antennas of a single gate is considered for prediction. While a radio tag may also be "seen" by the antennas of other gates, or its radio signals may also be received by the antennas of other gates, these gates, at best, simply classify the radio tag as irrelevant and ignore it. At worst, a radio tag may be incorrectly associated with passage through adjacent gates, leading to data inconsistencies.

[0006] This can be addressed with more expensive hardware that provides more accurate data and / or higher data rates (even when there are many radio tags in the gate area). Additionally, the transmission range of the gate's antenna can be limited, for example, by using metal shielding, to avoid misclassification. Summary of the Invention

[0007] In this context, one objective of the present invention is to provide a way to improve the accuracy and robustness of location identification and / or passage identification in a device having multiple radio tag detection gates in a low-cost and simple manner.

[0008] This task is accomplished by a method having the features of claim 1 and an apparatus having the features of claim 12.

[0009] Advantageous embodiments and improvements of the invention are given in the dependent claims.

[0010] The proposed method is used to determine the passage of radio tags through radio tag detection gates within a monitored area. Multiple radio tag detection gates (also known as RFID gates or RFID-Gates) are installed within the monitored area. These gates have one or more antennas (e.g., four antennas) to receive radio signals from radio tags within the monitored area. The monitored area may be, for example, a warehouse, and the gates are used to detect entry into or exit from the warehouse. Radio tags (e.g., RFID tags) can be attached to various objects such as packages, containers, forklifts, etc.

[0011] According to the proposed method, when a radio tag moves across a monitored area, in the first step a), at least one of a plurality of radio tag detection gates receives the radio signal of the radio tag. The antennas of the radio tag detection gates can be mounted on the respective gates such that the radio signal is received at different intensities and / or at different times when the radio tag approaches the respective gate. Furthermore, due to the arrangement of each gate within the monitored area, and due to the movement and / or position of the radio tag, each gate also receives the radio signal of the radio tag at different intensities and / or at different times.

[0012] Therefore, in another step b), the method detects the signal strength curve of each radio signal received by at least one of a plurality of radio tag detection gates. For each gate, the signal strength curve may consist of the received signals from the respective antennas. Thus, for different locations of the radio tag, the signal strength (also known as the RSSI value: Received signal Strength Indicator) determined by the plurality of gates is detected.

[0013] Based on the detected signal strength curve, the method can then determine in step c) which of the multiple radio tag detection gates the radio tag passed through. In this case, it is advantageous to aggregate and use the information from the multiple radio tag detection gates together to determine whether the gate was passed through and which gate it passed through. For example, this can be determined from the signal strength curves of each gate, as will be explained in detail below.

[0014] If interference is detected at a radio tag detection gate, passage through that gate is determined based on the signal strength curves detected by other radio tag detection gates. Interference at a gate might occur, for example, due to the presence of a large metal object within the gate area, or it could be attributed to a significant reduction in data rate when a large number of radio tags pass through simultaneously. This could result in insufficient data available for predicting the passage through the faulty gate. It is also possible that no data is available at the gate with interference at all. A single model relying on the data from that gate would struggle to provide accurate and robust predictions in such cases. However, by combining data from multiple gates using the proposed method, data from the remaining (preferably adjacent) gates can be used to generate good predictions for passage through that gate even when interference is present within its area. In particular, data from those gates through which the radio tags have passed can be used here.

[0015] By sharing data from multiple radio tag detection gates, defects or interference in the gates can be compensated for. In this case, data from other gates can be used to identify the gate with the defect. Therefore, the method presented here helps to provide reliable predictions for a large number of radio tags even when interference or capacity issues exist in individual radio tag detection gates.

[0016] According to one embodiment, each of the radio tag detection gates receives a radio signal from a radio tag and determines a signal strength profile for each received radio signal. If all gates are functioning correctly and there is no interference, the signal strength profiles will correspondingly indicate how close the radio tag is to the corresponding gate during its movement. If reception is interrupted at a gate, it can be inferred, for example, that the gate is defective. This information can then be used to determine passage through the radio tag detection gate.

[0017] According to another implementation, a machine learning module is used to determine passage through the radio tag detection gate.

[0018] Such machine learning modules may include or be based on artificial neural networks, recurrent neural networks, convolutional neural networks, LSTM networks, deep learning architectures, support vector machines, Bayesian neural networks, autoencoders, Gaussian processes, data-driven regression models, K-nearest neighbor classifiers, physical models, or decision trees. Typically, this machine learning module is first trained, for example, in the device to which it will subsequently be used (i.e., a combination of radio tag detection gates, radio tags, and other environmental elements).

[0019] During training, the radio tags attached to an object can be moved along with the object across the monitored area. This allows training to be conducted under conditions that closely resemble real-world scenarios. In particular, the potential impact of tagged objects on the signal strength curve can be considered during training.

[0020] Then, signals from the radio tag detection gate can be received, and additionally, for example, the position of the radio tag or object can be determined by a positioning sensor. Optical sensors, cameras, radio wave-based sensors, proximity sensors, and / or laser scanners can all be used as positioning sensors. In particular, the positioning sensor can be implemented based on a lidar system, radar system, real-time positioning system (RTLS) system, Bluetooth system, WLAN system, or 5G system. Preferably, multiple such positioning sensors, especially multiple cameras, can be used.

[0021] Subsequently, the machine learning module is trained based on the signal strength curves of the detected current location and the received signal to reproduce the corresponding current location of the radio tag based on at least one signal strength curve.

[0022] Here, training is generally understood as optimizing the mapping from the input signal to the output signal of a machine learning module. This mapping is optimized during the training phase according to one or more predefined criteria. In predictive models, prediction error can be used as a criterion in particular. Through training, for example, the network structure of neurons in a neural network and / or the weights of connections between neurons can be adjusted or optimized to satisfy the predefined criteria as well as possible. Therefore, training can be understood as an optimization problem. For such optimization problems in the field of machine learning, a large number of efficient optimization methods are available, especially gradient-based optimization methods, gradient-free optimization methods, backpropagation methods, particle swarm optimization, genetic optimization methods, and / or population-based optimization methods.

[0023] According to one implementation, this trained machine learning module can determine a single prediction for each radio tag detection gate based on the signal curves detected by the corresponding radio tag detection gates, and aggregate these single predictions to determine which of the multiple radio tag detection gates the tag traversed. Therefore, according to this implementation, a prediction of the radio tag's location / traversal is first created separately for each gate, and then these single predictions are aggregated and evaluated together. Alternatively, the machine learning module can also receive raw data (i.e., the signal curves detected by each radio tag detection gate) and aggregate it. Then, the machine learning module generates an overall prediction to determine which of the multiple radio tag detection gates the tag traversed.

[0024] According to another embodiment, determining which of a plurality of radio tag detection gates a radio tag has passed through includes comparing the signal strength of the detected signal strength curves. Even if each of these gates displays a signal strength curve individually, which curve only indicates passage with some uncertainty, it is possible to determine which gate the radio tag has passed through by comparing the signal strength curves of multiple gates. The result of the comparison here can indicate which gate the radio tag has passed through with less uncertainty than a single gate. Here, comparing the signal strength curves provides a simple and feasible approach to this determination.

[0025] According to one implementation, if the signal strength curve detected by the corresponding radio tag detection gate is unknown in the previous data, interference is detected in the radio tag detection gate. The previous data could be, for example, training data for a machine learning module. Even if no radio signal is detected at all, but a neighboring gate receives a radio signal, interference can be inferred in the radio tag detection gate.

[0026] According to another embodiment, the arrangement of radio tag detection gates within the monitored area is known, and passage through one of the radio tag detection gates is determined by analyzing the temporal correlation between the signal strength curves detected by adjacent radio tag detection gates. Background knowledge of the radio tag detection gate arrangement is utilized in this case. For example, if the gates are arranged such that passing through the first and third gates necessarily involves passing through the second gate, then the temporal correlation between these passages can be used. In this example, if the signal strength curves of the first and third gates indicate passage, it can be inferred that the second gate was also passed through. This is true even if the gate has no data or very little data. In this way, the method can also be used in environments where, for example, not all gates are equipped with antennas for receiving radio signals from radio tags.

[0027] According to another embodiment, steps a) to c) are performed during the training phase of the machine learning module, wherein the signal strength curve is input into the machine learning module as training data. As described above, the machine learning module can be trained in a device consisting of a radio tag detection gate. For this purpose, it is preferable to perform the above steps a) to c) to train the machine learning module to reliably predict passage through the gate. Furthermore, the location information of the radio tags can also be used as training data, as described above.

[0028] Furthermore, the arrangement of the radio tag detection gates can also be input into the machine learning module as training data. This makes the machine learning module's predictions of gate passage more robust and accurate during use, as data from adjacent gates are used in both training and execution. Even when the data rate of individual radio tags drops significantly (e.g., when multiple radio tags are simultaneously in the receiving area of ​​a gate), the number of received radio tag signals can increase due to the availability of more antennas (i.e., antennas from multiple gates). Although the signals from different gates at different times have different meanings, this pattern is already included in the training data through the pre-defined (fixed) arrangement structure and can be learned by the machine learning module.

[0029] A machine learning module is applied to at least one radio tag detection gate to determine passage through that gate. In this case, a machine learning module can be provided for each gate to create a single prediction. Advantageously, the machine learning module can be used in multiple gates. Furthermore, a common machine learning module for receiving and processing the individual predictions can be additionally provided; or only a common machine learning module for receiving the (raw) signals from multiple gates can be provided.

[0030] According to another embodiment, steps a) to c) above are performed to retrain the machine learning module. This retraining can be performed during device operation to improve predictions and adapt the machine learning module to real-world operation. This retraining can also be used to equip machine learning modules for gates that do not yet have such detection capabilities. This modified machine learning module can be equipped with existing training data from other gates and can be used directly. Subsequently, the module can be fine-tuned through retraining.

[0031] For example, in a scenario where doors are arranged sequentially and traversed, information about the continuous passage through a series of doors can be collected with time-limited data and tagged passages. This data can then be used to recreate or fine-tune models for the doors involved. This, for example, allows in a factory with numerous doors the option of not initially equipping all facilities with (pre-trained) models, but rather creating models adapted to the actual real-world situation subsequently or incrementally. The data and models obtained in this way can then be used to adjust all machine learning modules within the facility based on real-world conditions. In this manner, for example, specialized sensor devices (such as cameras or photodiodes) can be eliminated to label specific passage scenarios.

[0032] According to another aspect, an apparatus for determining the passage of radio tags through radio tag detection gates within a monitored area is proposed. This apparatus is specifically configured to perform the method steps described above.

[0033] As described above, the device may have multiple radio tag detection gates disposed within the monitoring area, and each radio tag detection gate is configured to receive radio signals from radio tags and determine the signal strength curve of the received radio signals. Here, each radio tag detection gate has one or more antennas that receive the radio signals from the radio tags.

[0034] Furthermore, the device includes a processing unit configured to receive signal strength curves from each radio tag detection gate and determine, based on the detected signal strength curves, which of the multiple radio tag detection gates the radio tag traversed. For example, the processing unit may include a machine learning module as described above. This processing unit may be provided as a central unit or may be formed by multiple sub-units provided on different gates.

[0035] The methods and apparatus described above can be executed or implemented, for example, using one or more processors, computers, application-specific integrated circuits (ASICs), digital signal processors (DSPs), and / or so-called "field-programmable gate arrays" (FPGAs). Furthermore, the method can be executed at least partially in cloud and / or edge computing environments. The processing unit of the apparatus can be implemented using hardware and / or software technologies. When implemented using hardware technologies, the processing unit can be constructed as a device or part of a device, such as a computer or microprocessor. When implemented using software technologies, the corresponding unit can be constructed as a computer program product, function, routine, part of program code, or executable object.

[0036] The implementation methods and features described for the proposed method are also applicable to the proposed apparatus, and vice versa.

[0037] Other possible implementations of the invention include combinations of features or implementations not explicitly mentioned in the preceding or following descriptions of the embodiments. Those skilled in the art will also add individual aspects as improvements or additions to the corresponding basic forms of the invention. It should be noted that, regardless of the grammatical gender of a particular term, persons of all genders are included. Attached Figure Description

[0038] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings. Here, they are illustrated in schematic diagrams: Figure 1a shows a facility with multiple radio tag detection gates; Figure 1b shows the signal curve of the radio signal received by the antenna of the radio tag detection gate; Figure 2 An apparatus for determining passage through one of the radio tag detection gates in Figure 1 is shown; Figure 3 The signal curves of multiple radio signals from the radio tag in Figure 1 are shown; and Figure 4 A facility with multiple radio tag detection gates arranged in sequence is shown.

[0039] Where the same or corresponding reference numerals are used in the accompanying drawings, these reference numerals denote the same or corresponding entities, which in particular can be realized or designed as described in conjunction with the relevant drawings. Detailed Implementation

[0040] Figure 1a illustrates, for example, a setup in a warehouse consisting of multiple radio tag detection doors A, B, C, D, E, F, and G. Here, in this example, door AG separates the inner warehouse area (above door AG) from the outer area (below door AG). As an example, a package or other object OBJ carrying a radio tag RF (e.g., an RFID tag) is transported from location 1 to location 5. During this process, it passes through door C.

[0041] Each gate in gate AG is equipped with a certain number of antennas A1-A4 (e.g., four antennas, in...). Figure 2 (As shown in Figure 1b as an example), these antennas continuously receive signals from all radio tags in the environment (i.e., within the surveillance area UB). This signal is shown in Figure 1b and... Figure 3 The following is illustrated as an example. Figure 1b shows the radio signals of the radio tag RF received by antennas A1-A4 of gate C over a time period of 150 discrete time steps. In this case, antennas A3 and A4, facing inward, will first identify the radio tag RF with high signal strength, followed by antennas A1 and A2, facing outward. Figure 1b only shows the package traversal from position / time point 4 to position 5. In the time steps before and after, the signal strength of the radio tag shown will either rise from 0 or fall back to 0.

[0042] As package OBJ passes through multiple doors F, E, and D within the monitored area UB, and then exits the warehouse through door C, each door A, B, C, D, E, F, and G receives the radio signal of the radio tag RF within its corresponding reception range. This data can be collected and further processed, as described below.

[0043] The received radio signals are in Figure 3The following diagram is shown as an example. To simplify the illustration, the signal data of antennas A1-A4 for each of gates A, B, C, D, E, F, and G are summarized on a single curve. The X-axis plots the marker positions 1-5 that the radio tag RF passes through over time t. From left to right in time, it can be seen that gates E, F, and G receive the signal from the radio tag RF first. In subsequent times, gates A, B, C, and D also receive the signal from the radio tag RF. Since the package OBJ passes through gate C (i.e., antennas A1-A4 closest to gate C), the signal curve for gate C has the highest signal value, as seen through the signal value S on the Y-axis.

[0044] In conventional systems, determining which gate AG a radio tag (RF) has moved through is based on data from one of those gates, A, B, C, D, E, F, or G. In contrast, the method performed by the apparatus shown in this paper does not use data from just one gate (A, B, C, D, E, F, or G). Instead, it uses data from all of these gates (see...) Figure 3 All of these were taken into account and included in the evaluation.

[0045] like Figure 2 As shown in the partial diagram, antennas A1-A4 of gates C and D (and the remaining gates A, B, E, F, G shown in Figure 1a) receive radio signals from the radio tag RF. These radio signals, or their signal strength curves, are then forwarded to the processing unit VE. This processing unit VE may have a machine learning module CNN or be directly implemented as such a machine learning module.

[0046] Subsequently, the machine learning module CNN not only independently evaluates the signals of the gate AG, but also determines the passage to the gate based on all available data of the gate AG, in this example, the passage to gate C. The benefit of this is that even if the prediction of the actual passage to one of the gates (e.g., gate C) has high uncertainty, this uncertainty can be eliminated by further information from the other gate AGs. Therefore, the overall prediction is more reliable and robust, thus enabling a higher degree of determinism in the output.

[0047] For example, the machine learning module CNN can compare the signal strength curves of different gate AGs. The gate C showing the relatively highest signal strength curve (see...) Figure 3 ) will be predicted to be the gate through which the radio tag RF passes.

[0048] The machine learning module CNN can also be used to determine passage through door B, which is arranged in series with doors A and C, such as... Figure 4 As shown.

[0049] In this example, an object OBJ with a radio tag RF passes through two or more antenna gates A, B, and C in sequence. If gate B is interfered with or lacks antennas A1-A4, the timing correlation between passing through gates A and C can be used to determine the passage through gate B. Therefore, the time points of passing through gates A and C can serve as the boundaries of a time interval, and the passage through gate B must occur within that time interval. By taking the corresponding structural conditions as a premise, it is possible, for example, to rule out bypassing gate B.

[0050] In this way, passage through door B can be identified without the need for additional sensor devices (such as optical cameras or photoelectric gratings), and door B does not need to perform passage determination on its own.

[0051] Data collected in this way can also be used as training data for a machine learning module (CNN) to identify passage through a middle door (B). This is achieved by associating known information (i.e., an object coming from direction A and leaving the door in direction C) as a label with the data recorded within a time window. In this way, training data for detecting doors using radio tags can be collected during factory or logistics center operations to (automatically) recreate a new, adjusted model for doors not yet equipped with this machine learning module (CNN), or to fine-tune an existing model.

[0052] Although the present invention has been described in conjunction with embodiments, it has many modifications.

[0053] List of reference numerals 1-5 Time Points / Locations A1-A4 antennas A Radio Tag Detection Gate B Radio Tag Detection Gate C Radio Tag Detection Gate CNN Machine Learning Module D Radio Tag Detection Gate E-radio tag detection gate F Radio Tag Detection Gate G Radio Tag Detection Gate OBJ objects RF radio tags S signal t time UB monitoring area VE processing unit.

Claims

1. A method for determining the passage of radio tags (RF) through radio tag detection gates (A, B, C, D, E, F, G) within a surveillance area (UB), wherein a plurality of radio tag detection gates (A, B, C, D, E, F, G) are disposed within the surveillance area (UB), wherein each radio tag detection gate (A, B, C, D, E, F, G) has one or more antennas for receiving radio signals from radio tags (RF) within the surveillance area (UB), wherein the method comprises the following steps: a) Receive radio signals from a radio tag (RF) through at least one of the plurality of radio tag detection gates (A, B, C, D, E, F, G). b) Detect the signal strength curve of each radio signal received by at least one of the plurality of radio tag detection gates (A, B, C, D, E, F, G). c) Based on the detected signal strength curve, determine which of the plurality of radio tag detection gates (A, B, C, D, E, F, G) the radio tag (RF) traversed, and d) If interference is detected at a radio tag detection gate (A, B, C, D, E, F, G), then the passage through that radio tag detection gate (A, B, C, D, E, F, G) is determined based on the signal strength curves detected by other radio tag detection gates (A, B, C, D, E, F, G).

2. The method of claim 1, wherein, Each of the radio tag detection gates (A, B, C, D, E, F, G) receives radio signals from the radio tag (RF) and determines the signal strength curve of each received radio signal.

3. The method according to any of the preceding claims, characterized in that, The machine learning module (CNN) determines a single prediction for each radio tag detection gate (A, B, C, D, E, F, G) based on the signal curves detected by the corresponding radio tag detection gates (A, B, C, D, E, F, G), and aggregates the single predictions to determine which of the multiple radio tag detection gates (A, B, C, D, E, F, G) was traversed.

4. The method according to any of the preceding claims, characterized in that, The machine learning module (CNN) aggregates the signal curves detected by each radio tag detection gate (A, B, C, D, E, F, G) and produces an overall prediction to determine which of the multiple radio tag detection gates (A, B, C, D, E, F, G) was traversed.

5. The method according to any of the preceding claims, characterized in that, Determining which of the multiple radio tag detection gates (A, B, C, D, E, F, G) the radio tag (RF) traverses includes comparing the signal strength of the detected signal strength curves.

6. The method according to any of the preceding claims, characterized in that, If the signal strength curve detected by the corresponding radio tag detection gate (A, B, C, D, E, F, G) is unknown in the previous data, then interference is detected at the radio tag detection gate (A, B, C, D, E, F, G).

7. The method according to any of the preceding claims, characterized in that, The arrangement of the radio tag detection gates (A, B, C, D, E, F, G) within the surveillance area (UB) is known, and the passage through one of the radio tag detection gates (A, B, C, D, E, F, G) is determined by determining the temporal correlation between the signal strength curves detected by adjacent radio tag detection gates (A, B, C, D, E, F, G).

8. The method according to any of the preceding claims, characterized in that, Steps a) through c) are performed during the training phase of the machine learning module (CNN), wherein the signal intensity curve is input into the machine learning module (CNN) as training data.

9. The method of claim 8, wherein, The machine learning module (CNN) receives the location information of the radio tags (RF) and / or the arrangement of the radio tag detection gates (A, B, C, D, E, F, G) as training data.

10. The method according to claim 8 or 9, characterized in that, The machine learning module (CNN) is applied to at least one of the radio tag detection gates (A, B, C, D, E, F, G) to perform the determination of passage through the radio tag detection gates (A, B, C, D, E, F, G).

11. The method according to any one of claims 8 to 10, characterized in that, Perform steps a) through c) to retrain the machine learning module (CNN).

12. An apparatus for determining a radio tag (RF) crossing radio tag detection gates (A, B, C, D, E, F, G) within a monitored area (UB), characterized in that The apparatus is configured to perform the method steps according to any one of claims 1 to 11.

13. The apparatus of claim 12, wherein, The device has multiple radio tag detection gates (A, B, C, D, E, F, G) located within a monitored area (UB) and configured to receive radio signals from radio tags (RF) and determine the signal strength curves of the received radio signals. The device also has a processing unit (VE) configured to receive the signal strength curves from the radio tag detection gates (A, B, C, D, E, F, G) and determine, based on the detected signal strength curves, which of the multiple radio tag detection gates (A, B, C, D, E, F, G) the radio tag (RF) traversed.

14. The apparatus of claim 13, wherein, The processing unit (VE) has a machine learning module (CNN).

Images