TECHNOLOGICAL FIELD

The present invention relates to system and method for determining location of selected objects within selected environments, and specifically relates to techniques for determining object location in multi-room environment.

BACKGROUND

Indoor positioning systems provide data about exact location of one or more objects within a predefined area. Such location data may be used for various applications such as to locate specific items in a store or warehouse as well as various other applications.

Ultrawide band (UWB) relates to a radio frequency communication technology the allows transmission of information using wide bandwidth, typically over 500MHz. This technique enables transmission of relatively large amount of data for short ranges. This technique required relatively low energy use for the transmission and reception of UWB signals.

Typical UWB-based indoor positioning system may utilize triangulation for determining location data based on measurement of time of arrival of signals transmitted to or from an object to be located. This technique provides simple and accurate data using an arrangement of anchor units arranged in the space within relatively direct line of sight with the objects to be located.

GENERAL DESCRIPTION

There is a need in the art of a novel, simple and robust technique for determining location of selected objects within multi-room environment. Such multi-room environment may be formed by arrangement of a plurality of spaces separated by walls. The plurality of spaces may be directly connected between them or connected via one or more corridors leading to the plurality of compact spaces, or rooms.

In typical UWB -based location systems, the location of one or more tag unit is determined by triangulation or multilateration. To provide suitable triangulation data, the tag typically needs to communicate with 3-4 anchor units, having known location. Additionally, the time-of-flight of a signal between an anchor and the tag unit needs to reflect the actual distance between them. Alternative signal paths, that include reflection of the signal from walls or other objects, result in ambiguous time-of-flight data that may hinder the ability to triangulate tag position. Generally, to provide reliable location data by triangulation, the tag and anchors should preferably be in line-of-sight between them.

Pinpoint location data, with centimeter resolution may be required in certain situations and corresponding regions such as atrium, storage facilities etc. However, in certain regions, such as medical centers, hotels, schools, or other environments characterized by multi-room or multi-region arrangement, the typical layout requiring line-of-sight between each tag and 3-4 anchors to determine tag location may result in overcomplicated installation process and costs. More specifically, such layout may require installation of 3-4 anchors in each enclosed space (room), making installation of such a system to be highly complicated and pricey due to required number of anchors and wiring between them.

Moreover, in such multi-room environment, location data indicating a room in which a selected object is located may in some applications be sufficient. Such space arrangement may allow reduced accuracy and resolution in location data. More specifically, providing information that a specific item is located within certain room may be sufficient, as it may require low effort to locate an item within a small room or a couple of adjacent rooms. Accordingly, in multi-room environment various objects to be located may be found in different rooms or compact spaces and may also be located in one or more open regions (e.g., corridor area).

Utilizing conventional triangulation techniques for determining location of selected items in a multi-room environment typically requires a large number of anchor receivers/transmitters. This increases the costs and complexity of installation. This is while in a typical multi-room environment the location resolution requirement may often be reduced to indicate a specific room where items are located. Accordingly, multi-room arrangement of a space may limit line-of- sight between tag unit marking an item to be located, and anchor unit providing anchor points and enabling triangulation positioning.

The present disclosure provides a system and method for determining location of selected items within a multi-room space, comprising a plurality of one or more compact sub-regions (rooms) and may include one or more general areas (e.g., corridor) or not. The present technique may further operate in an arrangement of a plurality of such spaces (e.g., different floors of a building). This is while utilizing an arrangement of anchor transmitter or receiver units along the general area (corridor). The present disclosure utilizes training stage and processing of tag-anchor communication data to specify a room in which the item is located. To this end, the present disclosure utilizes a control system, comprising one or more processors and memory circuitry configured for receiving and processing data on location signals, including e.g., times of arrival, signal power, tag or anchor unit id etc., of collected location signals. The control system utilizes selected one or more processing modules (including e.g., artificial intelligence, machine learning or other suitable processing modules) pretrained for determining room identity in which an item is located based location signal data received by or transmitted from the anchor units, and intensity of the signal. Following the training stage, the control system comprises a trained machine learning module trained for determining item location differentiating between different rooms and open spaces of the region where the system is installed.

To this end, the control system, and/or the processing modules thereof, undergo a training stage, in which relations between location signal data and room data are recorded. This allows the control unit to process location signal data in accordance with training determining room location, to provide room resolution location data.

To this end, the location system according to some embodiments of the present disclosure comprises a control system, comprising at least one processor and respective memory unit, a plurality of anchor units configured to be placed in selected locations in a selected region, and at least one tag unit to be positioned/mounted on at least one object to enable determining location of the at least one object.

The at least one tag unit, and the plurality of anchor units are configured to communicate between them and to provide location signal data to the control system. The control system is configured to receive the location signal data from the plurality of units and process the location signal data to determine location of the at least one tag unit and provide output data indicative thereof. To this end, the control system may determine location data using at least first and/or second processing techniques. In a first processing technique, the at least one processor may operate to triangulate tag location utilizing location signal data. In a second processing technique, the at least one processor operate to estimate tag location data using respective machine learning module being pretrained on estimating location in the region where the system is installed.

The use of machine learning processing, and proper training on a region where the system is installed enables to greatly reduce installation complexity and costs. As indicated above, the present technique enables reducing the number of anchor units installed in the region, obviating the need for line-of-sight between a tag unit and three- four anchor units required for general triangulation. Accordingly, for a general multi room environment, the present technique enables determining tag location in single room resolution, using an arrangement of anchor units that are mostly located in the open regions (e.g., corridor), while some anchors may be located in selected rooms to enhance accuracy.

Following installation of the plurality of anchor units, training stage of the system according to some embodiments of the present disclosure may utilize providing a training data set in the form of location signal data pieces associated with a selection of tagged items, where rooms in which the tagged items are located in provided as label data. More specifically, the training stage may involve the use of one or more tag units, and placement of the one or more tag units in different locations in the region, including different rooms and open regions. The location signal data pieces obtained by tag-anchor communication is labeled by actual position (in room- level resolution) of the tag unit(s). Training of the AI module is generally followed by testing and generating a confusion matrix. The confusion matrix may be collected based on similar tagged items used for training, or a set of one or more additional tagged items. The confusion matrix indicated number of instanced where the AI labeled items as located in the different rooms, over the actual location of the items based on labeling thereof. The so-generated confusion matrix enables to determine specificity and/or sensitivity of the location system. If the confusion matrix indicates a confusion rate exceeding a selected or predetermined ratio, the system may be installed including one or more additional anchors, located at selected positions to differentiate between the rooms associated with the high confusion rate.

Thus, the present technique utilizes machine learning and/or artificial intelligence processing, combined with reduced number of location anchors, for determining location of one or more objects within a space, where location data is needed at room resolution, i.e., indicating a room in which the items are located.

Thus, according to a broad aspect, the present disclosure provides an indoor location system comprising: one or more tag units associated with one or more items to be located, and one or more anchor units positioned within a selected indoor space; a control unit comprising at least one processor and memory and connectable to receive location signal data comprising at least one of time-of-arrival data and signal strength data, and for processing said location signal data and determining data on room-level location of said one or more tag units.

According to some embodiments, the control unit may comprise a communication module configured for communicating with said one or more anchor units for receiving location signal data indicative of location of at least one tag unit.

According to some embodiments, the memory unit may comprise prestored data indicative of statistical relations between location signal data and room level location of a respective tag unit.

According to some embodiments, the at least one processor may be configured and operable for receiving location signal data indicative of communication of a selected tag unit with one or more anchor units, and for processing the location signal data to determine room-level location of said selected tag unit; wherein said processing comprises estimating a triangulation score, indicative of at least one of accuracy and fidelity of location determined by triangulation of the location signal data, and determining operation in one or first and second processing modes in accordance with said triangulations core; said first processing mode comprises determining location of said selected tag unit by triangulation based on time-of-arrival data in said location signal data, said second processing mode comprises estimating room-level location in accordance with said statistical relations between location signal data and room level location of a respective tag unit.

According to some embodiments, the memory unit may comprise prestored data comprising computer readable instructions that when executed by said at least one processor cause the at least one processor to execute a pre-trained machine learning module configured for determining room-level location of said one or more tag units in accordance with location signal data associated with said one or more tag units. According to some embodiments, the at least one processor may be configured and operable for receiving location signal data indicative of communication of a selected tag unit with one or more anchor units, and for processing the location signal data to determine room-level location of said selected tag unit; wherein said processing comprises estimating a triangulation score, indicative of accuracy of location determined by triangulation of the location signal data, and determining operation in one or first and second processing modes in accordance with said triangulations core; said first processing mode comprises determining location of said selected tag unit by triangulation based on time-of-arrival data in said location signal data, said second processing mode comprises operating said machine learning module for determining room-level location of said one or more tag units in accordance with location signal data associated with said one or more tag units.

According to some embodiments, the one or more anchor units may comprise anchor units installed in open space regions of said selected indoor space.

According to some embodiments, the at least one processor may be configured for performing a setup session, said setup session comprises:

(a) providing instructions to an operator to move at least one selected tag unit in said selected indoor space;

(b) collecting location signal data pieces indicative of communication between said at least one selected tag unit in a plurality of locations in said selected space;

(c) processing said location signal data pieces in accordance with path of move of said selected tag unit in said selected indoor space and determining at least one statistical relation between location of said at least one selected tag unit and respective location signal data.

According to some embodiments, the system may further comprise determining a training score indicative of specificity of said at least one statistical relation and comparing said training score to a respective selected threshold; and if said training score is below said selected threshold, generating a notification indicating a request to install at least one additional anchor unit in at least one additional location in said selected indoor space. Said at least one additional location may be in one or more rooms in said selected indoor space. In some embodiments, said at least one additional location may be in at least one room associated with said training score below a selected local threshold. According to some embodiments, at least some of (or each of) said one or more tag units may comprise a transmitter configured to periodically transmit beacon signal comprising at least tag unit identity data, said one or more anchors are configured to receive said beacon signal and determine time of reception of the beacon signal, thereby enabling detection of time difference for beacon signal reception by two or more anchor units.

According to some embodiments, the one or more anchor units may comprise a signal receiver configured to receive beacon signal from one or more tag units, a clock synchronized between the one or more anchor units for determining time difference of signal collection between the anchor units, and a communication module configured to transmit location signal data to the control unit.

According to one other broad aspect, the present disclosure provides a method for determining location data, the method comprising: providing an arrangement of two or more anchor units in a selected space for determining location of one or more tag units in said selected space; operating said anchor units and tag unit for exchanging location signal transmission and providing location signal data comprising at least time-of arrival data and signal strength data; training an artificial intelligence module on said location signal data for determining location of said one or more tag unit; utilizing said training and processing locations signal data for determining location of said one or more tag unit within room-level separation.

According to some embodiments, the method may further comprise: receiving said location signal data from said two or more anchor units and processing said location signal data; said processing comprises estimating a triangulation score, indicative of accuracy of location of said one or more tag units determined by triangulation of the location signal data, and determining operation in one or first and second processing modes in accordance with said triangulations core; said first processing mode comprises determining location of said selected tag unit by triangulation based on time-of-arrival data in said location signal data, said second processing mode comprises operating said machine learning module for determining room-level location of said one or more tag units in accordance with location signal data associated with said one or more tag units. According to some embodiments, the method may further comprise determining a training score indicative of specificity of said an artificial intelligence module following training on said location signal data and comparing said training score to a respective selected threshold; if said training score is below said selected threshold, generating a notification indicating a request to install at least one additional anchor unit in at least one additional location in said selected indoor space.

According to some embodiments, the notification may comprise, or be indicative of, instructions to install said at least one additional anchor unit in said selected indoor space anchor unit being at least one room associated with said training score below a selected local threshold.

According to yet another broad aspect, the present disclosure provides a computer implemented method comprising:

(a) obtaining location signal data indicative of at least time difference of arrival of location beacon from two or more anchor units placed is a selected space;

(b) using at least one processor and processing said location signal data for determining location of at least one tag unit transmitting said location beacon signal, said processing comprises operating a pre-trained machine learning module for inferring location of said at least one tag unit in accordance with the respective location signal data, and

(c) generating location data indicative of location of said at least one tag unit and storing said location data in a memory unit available to be obtained by an operator.

According to some embodiments, said processing comprises determining location of said at least one tag unit by triangulation using time difference of arrival in said location signal data, and if said triangulation is unsuccessful operating said machine learning for inferring said location.

According to some embodiments, the method may further comprise: receiving said location signal data from said two or more anchor units and processing said location signal data; said processing comprises estimating a triangulation score, indicative of accuracy of location of said one or more tag units determined by triangulation of the location signal data, and determining operation in one or first and second processing modes in accordance with said triangulations core; said first processing mode comprises determining location of said selected tag unit by triangulation based on time-of-arrival data in said location signal data, said second processing mode comprises operating said machine learning module for determining room-level location of said one or more tag units in accordance with location signal data associated with said one or more tag units.

According to some embodiments, the method may further comprise determining a training score indicative of specificity of said an artificial intelligence module following training on said location signal data and comparing said training score to a respective selected threshold; if said training score is below said selected threshold, generating a notification indicating a request to install at least one additional anchor unit in at least one additional location in said selected indoor space.

According to some embodiments, the notification may comprise, or be indicative of, instructions to install said at least one additional anchor unit in said selected indoor space anchor unit being at least one room associated with said training score below a selected local threshold.

It should be noted that the above-described aspects of the present disclosure and respective embodiments may be implemented separately or combined. More specifically, features of the present disclosure described with respect to the system may be implements in a method thereof and features described in connection with the above-described method may be implemented by the system of the present disclosure.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to better understand the subject matter that is disclosed herein and to exemplify how it may be carried out in practice, embodiments will now be described, by way of non-limiting example only, with reference to the accompanying drawings, in which:

Fig. 1 exemplifies anchor arrangement in conventional indoor location system;

Fig. 2 exemplifies anchor arrangement in indoor location system according to some embodiments of the present invention;

Fig. 3 exemplifies room arrangement in exemplary deployment field illustrating operation of some embodiments of the present disclosure;

Fig. 4 illustrates a method for determining objects’ location using a location system according to some embodiments of the present disclosure; Fig. 5 illustrates a method for initial training according to some embodiments of the present disclosure;

Fig. 6 illustrates schematically a location system according to some embodiments of the present disclosure;

Fig. 7 exemplifies method for determining tag location according to some embodiments of the present disclosure; and

Fig. 8 shows decision tree for location technique according to some embodiments of the present invention.

DETAILED DESCRIPTION OF EMBODIMENTS

The present disclosure provides a system and respective method for use in indoor positioning of selected objects. The technique of the present disclosure generally utilizes ultra-wide band (UWB) communication between one or more anchors and one or more tag transceivers associated with items/objects to be located.

Generally, in the realm of Indoor-positioning, UWB technology is associated with highly accurate, positioning, as accurate as decimeter level. Further, Real-Time Location Systems (RTLS) based on UWB communication generally involve time of flight measurements. Accordingly, object location can be estimated by triangulation or multilateration of signals exchanged between the portable device (i.e., tag) and fixed, reference devices, often referred to as anchors.

In the conventional techniques, a portable UWB device (tag) generally needs to communicate with at least 3-4 anchors to provide sufficient data enabling to determine its location by triangulation. Further, as signal time of flight is the characteristic representing the distance between a tag and an anchor, and location is determined based on distances of a tag from 3-4 anchors, the anchors and tag should effectively be within line-of-sight (LoS) to provide reliable location data. Generally, UWB-based triangulation can provide location accuracy of up to a few centimeters when relying on proper arrangement of anchors and direct line-of-sight between the tag and anchors.

The optimal conditions for indoor location techniques are typical to large open spaces, such as atriums or hangars, where direct line-of-sight between anchors and a tag unit is relatively simple to achieve. However, various indoor spaces are formed of a plurality of sections (rooms), such rooms may, or may not, be connected by one or more open spaces such as corridor areas. Examples of such indoor spaces include hospitals, schools, office buildings, shopping malls, etc. In spaces having such multi-room configurations deployment of a triangulation-based indoor location system becomes more complex. This is associated with signal reduction when passing through walls, and more importantly, reflection of communication signals from walls, generating multi-path signal transmission.

It should be noted that in the context of the present disclosure the term multi-room arrangement relates to an internal arrangement of a space including a plurality of separated regions, where each “room” is defined by a sub-region separated from other sub-regions by radiation reflecting materials (wall, glass window, curtain, etc.). A room may be formed by a hatch defining a space separated by three or more “walls” having a further side open to an open space region. As indicated herein, a “wall” may be formed of various materials and is defined herein as separating sub-regions and generating certain reflection and/or absorbance to radiation impinging thereon.

Reference is made to Fig. 1 exemplifying arrangement of anchor units ANT1- ANTn in a multi-room environment according to conventional indoor positioning techniques. In this example, the multi -room environment is a typical medical center layout, having long corridors and a large number of rooms along the corridors. As shown, to provide a tag unit with line of sight (LoS) to a number of 3-4 anchors, while allowing the tag unit to be located in the different rooms, the system needs to utilize a large number of anchors. As exemplified in Fig. 1, the conventional technique may require installing a plurality anchor units marked as ANl-ANn within each of the different rooms. As shown, the anchor units may generally need to be installed within the rooms, to enable LoS with items located within the rooms. In many cases, and specifically the hospital use-case, this substantial investment in infrastructure and setup, complicates and limits the proliferation of the solution.

The inventors of the present disclosure have found that in various scenarios associated with multi -room space, such as medical center or other such spaces, the required accuracy and resolution for location data may be reduced to room-level separation. More specifically, it is typically sufficient to indicate a specific room in which the item is located, and centimeter resolution location data is generally not required. Thus, the system may only be required to provide data on the room in which an item is located while identifying the item within the room is considered an easy task left for human personnel when the item is needed. Accordingly, the high number of anchors required for standard triangulation actually provides a more-than-required accuracy.

The present technique provides for room-level location data, using an arrangement of anchor units mounted generally within the space (typically along corridor sections), and a control unit comprising one or more processors and utilizing selected processing scheme for determining location of items based on location signals exchanged between the tag unit (associated with items to be found) and the anchor units. This is exemplified in Fig. 2 showing similar room arrangement as Fig. 1, including rooms numbered 1-6, and additional rooms that are not numbered. In this example, the location system of the present technique utilizes an arrangement of three anchor units AN1-AN5, and a control system (not specifically shown). Generally, the control unit may utilize artificial intelligence, or machine learning modules. In some configurations, for small spaces, the control unit may utilize statistical models allowing more direct processing for determining location of objects as described hereinbelow.

The present technique provides a hybrid implementation utilizing a relatively sparse arrangement of anchors installed in the space. To this end the present technique operates to determine location of a tag unit in at least one of two different techniques and may also utilize decision and scoring processing for determining which of the techniques is to be used. Accordingly, the technique of the present disclosure operates to determine location of tag units based on triangulation data when location signal data received from the anchor units is sufficient for determining time-of-arrival (ToA) data, and to determining tag location based on statistical analysis and/or artificial intelligence processing at times where location signal data is insufficient for triangulation. Accordingly, the present technique utilizes installation of a relatively small number of anchor units, typically arranged in spaces for general use, or within one or more rooms, and a training process that enables collection of location signal data, allowing to generate statistical model and train the AI processing to determine items’ location. The training process may include recording of location data from a plurality of tag unit positioned in known rooms within the space or moving a selected tag unit along different regions of the space while collecting location signal data and recording actual location along the path.

The training process includes processing of multiple data points, where each data point relates to a time instance when the training tag unit transmits a beacon signal. Accordingly, the data point includes location signal data formed of at least time of collection of the beacon signal by one or more anchor units. The location signal data may also include signal power as collected by the anchor units. Known location of the training tag unit is used as label data for each of the data points, enabling supervised training by one or more training techniques including e.g., using regression, classification, or other training techniques. During the training process, the one or more processors of the control unit, may operate to determine location of the tag unit in room-level separation, based on the received location signal data, e.g., including ToA and intensity of location data signals exchanged between the tag unit and the different anchor units. The one or more processors further compares the resulting location data to label of the respective data point to determine accuracy. This process is applied for a plurality of data points to achieve desired accuracy in determining location.

Generally, in building that include several floors, the system may be configured as two or more sub-systems, each assigned to a specific floor. In such environments, signal transmitted by a tag unit located in one floor, may be collected by anchors located in the same floor, as well as in other floors. In such environments, triangulation score may be determined under an assumption that the tag and anchor are in a common level (floor). Thus, triangulation score obtained by processing time of arrival of the signal to anchors of different floors may generally be lower than the triangulation score determined by time of arrival of the signal to anchors in the same floor. This is the distance between anchors of different floors and the tag may be increased indicating that the tag and respective anchor are in different levels. This enables selection of anchors within a common level with the tag to be used for determining location both using triangulation and using statistical model or machine learning processing as described herein. In various multi level arrangements however, signal reception by anchors located in different levels than the tag unit transmitting the signal, is generally relatively weak. This enables the system to operate as separated single level system.

Based on the training of the AI module, the system may operate to provide output data e.g., including confusion matrix data. The output data is indicative of efficiency of the training and provides data on cases where item location is determined different than its labeled actual location. Table 1 shows a confusion matrix determined experimentally for room arrangement as shown in Fig. 3. The confusion matrix indicates probability for identifying location for items located in specific rooms, and the probability for any room to be indicated based on the system training. In some embodiments, training of the system according to the present disclosure may include determining a training score, e.g., in the form of a confusion matrix, and, in accordance with error levels of the training score, the system may request operators to install one or more additional anchor units in reference to selected rooms.

Table 1

In this connection, reference is made to Fig. 3 illustrating an arrangement of rooms in a selected indoor space. In this specific and non-limiting example, the indoor space relates to a medical facility including rooms A-G and a general corridor. Three anchor units AN1-AN3 are illustrated being installed along the corridor and in room C to enable determining location of one or more tag units in the space.

Training session of the location system was conducted using the anchor layout exemplified in Fig. 3. In this training session, a tag unit was used. The tag unit was carried along while walking in the different rooms of the space, operated to periodically transmit beacon signals. Following the training session, the control system generated a confusion matrix shown in table 1. The confusion matrix provides indication to accuracy and efficiency of the training. Indicating the number of instances where location of a certain item is determined, and the determined location with respect to actual, known, location of the item. Generally, the confusion matrix provides indication on ability of the machine learning module to identify location accurately. If the confusion matrix includes high level of confusion between rooms, illustrated by non-diagonal elements exceeding selected thresholds, the system may be improved by addition of anchors positioned to differentiate between signals arriving from the rooms having high confusion rate. For example, an anchor may be positioned within a specific room, to remove confusion. To this end, the system may operate to test training accuracy following a training session and determine a training score. Such training score may be indicative of non-diagonal confusion probability, illustrated by magnitude of non-diagonal elements of the confusion matrix. Additionally, or alternatively, the training score may be indicative of lowest diagonal elements, indicating rooms in which the system did not identify the tag, e.g., room C showing detection accuracy of 0.56. Upon determining that the training score is below selected threshold, the system may generate a request to install one or more additional anchor units and may also indicate one or more rooms where the additional anchor units are to be installed. Selection of the rooms may be determined based on rooms showing higher confusion (or lower detection accuracy).

In this connection reference is made to Figs. 4 and 5 exemplifying methods for installation and training of the system according to the present disclosure. As shown in Fig. 4, system installation may include installing selected number of anchor units, typically in open regions (along corridors) of the selected space 4010. Following installing and connecting the anchor units to a suitable control system, one or more selected tag units are used for generating training data by travelling along throughout the space 4020. While the tag unit is moved around in the space, the method include registering travel path of the tag unit 4030 including locations in open spaces (e.g., corridor) as well as location within the rooms. To provide training data set, the method includes collecting location signal data 4040, typically including time of collection of the beacon signals by the anchor units, and signal strength when collected by anchors, and generating training data set 4050. The training set may be generated by labeling each location signal data instance (collection of each beacon signal by the different anchor units) by data on respective location of the anchor. The training set is used for training the machine learning module 4060 to determine location of the tag based on location signal data. As indicated above, following a training session, the resulting training data may be analyzed to determine its accuracy and statistical relations between the machine learning module output and actual location of the tag 4070. Such analysis is briefly exemplifying above with reference to Fig. 3 and table 1 exemplifying a confusion matrix. In response to limited accuracy (or high confusion) the technique may suggest installing additional one or more anchor units in selected regions when confusion matrix data is below desired accuracy and providing the machine learning module ready to infer location of tag unit in response to location signal data 4080. At this stage the machine learning module is ready to be used 4090.

As indicated, operation for training the machine learning modules may generally include a testing session that may optionally result in generating a request for installation of one or more additional anchor units. Fig. 5 exemplify training process according to some embodiments of the present disclosure. As shown in Fig. 5, training the machine learning module 5010 based on collected training data set, e.g., collected as indicated in 4020-4050 in Fig. 4. Following a training session, the method includes determining a training score 5020 indicative of level of accuracy and specificity of location determined using the machine learning module based on the training and available input data. Determining training score may include generating a confusion matrix 5025 as exemplified in table 1. Actual training score may be determined to indicate average or minimal accuracy values of the confusion matrix, level of diagonality of the table etc. Alternatively, other techniques for determining training score may be used. The training score is compared to a selected threshold 5030, associated with required/desired accuracy level. If the training score is determined to be below the selected threshold, the method may operate to select one or more rooms associated with lowest accuracy 5040. This is exemplified in table 1 indicating 56% accurate detection of items in room C. Accordingly, to improve determining of tag location, the method may generate a request to install one or more additional anchor units in the selected one or more rooms 5050. At this stage, an additional training data set may be generated 5060, however the additional training data set may be slightly more limited and include regions surrounding the selected one or more rooms where additional anchor units are installed. The updated training data set is used update training of the machine learning module 5070 and repeat the process of determining training score 5080 until the training score is sufficient.

Reference is further made to Fig. 6 schematically exemplifying a system for determining location of one or more tag units in a selected space. The system includes a plurality of anchor units AN1-AN3 configured to be installed in selected location in the space. The anchor units include at least UWB receiver, synchronized clock, and communication port for transmitting beacon collection data and time to the control system 600. The anchor units may communicate with the control system via wires or using wireless communication techniques. The tag unit generally transmits beacon signals using UWB frequencies, where a beacon signal generally includes signal ID and tag ID. The control system 600 may be configured as a computer system or server, and includes one or more processors 650, communication controller/port 610, memory 620, and may also include certain user interface 630. The user interface 630 may be provides as one or more terminals, module for web-based interface or any other platform allowing operator to receive output data from the location system, and optionally provide input operation data, or general queries.

The at least one processor 650, and memory 620 are operatively connected to communication controller 610, typically being hardware-based communication port, for receiving and processing location signal data. The at least one processor is configured to provide processing necessary for operating the control system as detailed herein. The at least one processor 650 can be configured to execute several functional modules in accordance with computer readable instructions that are generally stored in the memory 620, and/or in any selected non-transitory storage devices. Such functional modules are referred to hereinafter as comprised in the at least one processor 650, including for example: decision module 651, triangulation module 652, statistical location module 654 and/or machine learning modules 656.

Following training of the system, data on relation between parameters of location signal data and actual location of tag units may be stores in the memory as statistical data and/or as operational parameters of the machine learning module 656. For example, is small spaces, tag location may be determined using straight forward statistical relations, and machine learning processing may be not required. However, typical spaces that require indoor location system operation, may be rather large requiring higher processing capabilities that can be provided using machine learning techniques. Accordingly, the machine learning modules 656 and statistical location module 654 may replace each other’s operation in accordance with size of the space and number of parameters or anchors units providing location signal data.

Generally, during operation of the system for determining location of one or more tag units (TAG) generally associated with selected items to be located, the tag units periodically transmit beacon signals including at least tag ID and signal ID data. The beacon signal may be collected by one or more anchor units AN1-AN3, located in selected positions in the space. When an anchor unit receives a beacon signal, the anchor units register time of arrival of the beacon signal, and preferably certain data on power of the signal as collected. The anchor units transmit data on time of arrival and reception power of the beacon signal to the control system 600.

The control system operates to receive input data including location signal data indicative of time of arrival of the beacon signals to number of anchors and respective collection power and operates for processing the location signal data to determine location of the respective tag. The processing may generally include a decision stage including initial processing of the location signal data to determine if the data is sufficient for triangulation or if additional, machine learning processing, is required.

The initial processing may be provided by the decision module 651 and may include estimating of possible triangulation score, indicating of level of accuracy that can be obtained by triangulation using time of arrival differences of a beacon signal to two or more, or three or more, anchor units. Triangulation score may be determined based on quality of the location signal data including one or more parameters such as: power of beacon signal collected by the anchor units, signal quality parameters including e.g., existence of echo signals, etc. If the triangulation score is sufficient, i.e., exceeds a selected threshold, the system may proceed for determining location of the tag unit by triangulation using time-of-arrival difference, and provides data on location of the tag unit. Generally, triangulation may typically be used for objects located in open regions of the space such as corridors, main hall etc. However, when the tag is located within a room, or closed region of the space, location signal data may be insufficient for efficient triangulation. In that case the location data processing may be based on the machine learning or statistical relations to determine location of the tag unit with room-level resolution. As indicated above, for small spaces including limited number of anchor units, statistical data on location signal data in relation to actual location of the respective tag may be sufficient to infer tag location using a look-up table formed based statistical analysis 654 of the training data. For larger spaces, including e.g., 5 or more room sections, and many anchor units (e.g., 5 or more) the machine learning module 656 may be needed for determining location.

Triangulation score may be determine based on agreement, or consensus, between tag location as determined based on different pairs of anchors. Accordingly, is sufficient number of anchors agree on a specific location, the triangulation score indicates successful triangulation. However, in some situation, different pairs of anchors indicate different locations of the tag, rendering triangulation unsuccessful. In such situation, the system utilizes machine learning or statistical data to determine location of the tag.

Accordingly, the location signal data, including data on time of collection of a beacon signal by two or more (generally several) anchor unit is used as input data to the machine learning modules, while output data indicates a room in which the tag is location based on the training session. In some embodiments, the output data may also include a level of certainty, e.g., in the form of probability that the tag in is each of the different rooms as visible by lines of table 1.

Operational method is further illustrated in Fig. 7. As shown for operation of the location system, one or more tag units, associated with items to be located, are configured to communicate by periodically transmitting beacon signals 7010. The beacon signals are collected by one or more anchors 7020 by reception of beacon signals. Thus, the one or more tag units may periodically transmit location signals, the location signals are received by the anchor units and data on reception of the location signals is transmitted 7030 to the control system for processing. Alternatively, in some optional embodiments, the anchor units may transmit location signal, collected by the tag unit, which transmits data on the location signals to the control system.

The processing may include an initial processing including determining triangulation score 7040. If the triangulation score is sufficient, location of the tag may be determined by triangulation 7050. If the triangulation score is below a selected threshold, indicating insufficient data quality for triangulation, the processing proceeds to determining location by operation of the machine learning module 7060. Following determining location by machine learning and/or by triangulation, the location data is generated 7070 and stored indicating location of the respective tag unit. Accordingly, location data of the one or more tag units may be periodically stored in respective data base in the memory unit, including data on identity of the associated items. The so-stored location data is generally available to generate output data upon request, including transmission of location data to one or more user terminals or any other output module.

In some further embodiments, the control system may operate to receive location signal data and operate to determine location of the tag unit by triangulation. Fig. 8 illustrates a decision tree for processing location data and determining location of a tag unit according to some further embodiments. As shown, in response to location data, including data on To A of location signals exchanged between tag unit and one or more, or generally some, anchor units, the processing operates to determine location using triangulation technique 8010 (attempt triangulation). In the arrangement of the system, including sparse arrangement of anchor unit, triangulation attempt is tested 8020 (triangulation successful?), and if triangulation works, location of the tag is determined. Testing of triangulation is based on reliability of triangulation results, and level of ambiguity that may appear in the location signal data.

In a typical case, where items are within rooms, and limited LoS with the anchor units, location data determined by triangulation may be inconsistent or not converge properly to provide reliable indication. In this case, the processing according to the present invention comprises determine a level of sensitivity or accuracy obtained by triangulation data, e.g., determining a success score for location. If the success score is below a predetermine threshold, the processing automatically switches to determining location data using AI processing as described above. Accordingly, the location signal data in transmitted to the AI processing module to determine item location based on training of the AI module 8040, in accordance with ToA and signal intensity data. Based on the AI processing, location of the tag unit is determined and suitable indication, e.g., item is in room ##, is generated to operators 8030.

Thus, the present technique provides an indoor location system and corresponding method, for determining location of selected objects in generally room location resolution, utilizing reduced number of location anchors. The present technique generally utilizes pre-trained artificial intelligence processing utilizing time of arrival and signal strength data for estimating room identity and providing corresponding location data of selected objects.

It should be noted that the various features described in the various embodiments can be combined according to all possible technical combinations. It should also be understood that the present invention is not limited in its application to the details set forth in the description contained herein or illustrated in the drawings. The invention is capable of other embodiments and being practiced and carried out in various ways. Hence, it is to be understood that the phraseology and terminology employed herein are for the purpose of description and should not be regarded as limiting. As such, those skilled in the art will appreciate that the conception upon which this disclosure is based can readily be utilized as a basis for designing other structures, methods, and systems for carrying out the several purposes of the presently disclosed subject matter. Those skilled in the art will readily appreciate that various modifications and changes can be applied to the embodiments of the invention as hereinbefore described without departing from its scope, defined in and by the appended claims.