The amount of data output into our environment is increasing each day, and the development of new technologies constantly redefines how we interact with this information. It is therefore necessary to control the different ways information is diffused. As a result, a data dissemination methodology in the framework of the Supply Chain Management is introduced. A specific stage of the methodology is detailed in this paper which aims at storing directly on the product, relevant data to the subsequent users. To do so, a new type of product is presented referred to as ”communicating material”.
Andr´e Thomas1
New challenges and opportunities arise with concepts such as
Internet of Things (IoT), Ubiquitous/Pervasive Computing
[
Any product, during its life, passes through numerous
companies and undergoes various operations (manufacturing,
transportation, recycling. . . ). Technical, semantic and
organizational interoperability between these companies is not
always ensured, thus, conducing to information loss. If one
considers the product as an information vector (on which
information could be stored), it would contribute to improve
interoperability all along its life cycle. Meyer et al. [
As a result, we propose a new concept referred to as communicating material, which considers the material as intrinsically communicating. First, recent works carried out on this concept are discussed in section 2. This corpus of works mainly focus on the development of a data dissemination process to identify what the relevant information to users is and where it should be stored: on databases or on the product themselves? One important step of this framework deals with the storage/retrieval of data on/from the communicating materials, which is the subject of the paper. An appropriate architecture of communication is developed in section 3 which aims at splitting data over the communicating material and, at determining where this information is (or will be) located. An applicative scenario is finally presented in section 4. 2
As said previously, the product passes through numerous
companies during its life cycle. Each actor/operation requires
product related information (e.g., for decision-making,
production orders) which is not always available (inaccessible
information, e.g. owned by another supplier) and is not always
up-to-date [
➫ Process step 2 aims at selecting context-sensitive
information. When users want to write information on the product
at a given time (e.g. before the product leaves the company),
it is necessary to identify information that is relevant (e.g.
for subsequent users). This identification is achieved thanks
to the process step 2. Our approach, detailed in [
MDWP0 MDP-B
3 (✩)
Wood plank with a nominal 3/4”... 4m of...
Textile with a high developed pol... 3mm...
MD06 0.6 Textile which is provided with..0..05 15◦C.0...41 MDH-V1
Vehicle headrests which conform... 2 × ...
(b) Relational table Material & Data item relevance
item on the product according to the user concerns, the
environment details, etc. One data item corresponds to a cell
of a relational table (i.e. the intersection between a column,
named ”attribute”, and a row, named ”tuple”) as
emphasized in Figure 1(b). For instance, the data item located at
row 3-column 1 in table Material, noted TMat{3,1}, has the
value MD06. Only data items which have an interest of being
stored on the product are assessed (such a selection is also
proposed in [
➫ Two things are needed in process step 3: (i) first, the
product must be instrumented in order to carry such
information. As introduced previously, [
Data fragment storage/retrieval using
communicating materials
Process step 3 of the data dissemination process deals with
the storage of data items (identified across the process step 2)
on the product. In this section, an appropriate architecture
combined with a protocol of communication is developed in
order to split data items over the material and then, to rebuild
2 Designed in collaboration with the CETELOR laboratory.
RFID µtag
them. Subsequently, we develop a method to determine where
data are located on the communicating material.
First, it is necessary to have a communicating material, as
the communicating textile designed in our previous work [
Network munteicraiatilng comma 1. Protocol (8 bits): Integer from 0 to 255 which enables to know which fields compose the packet. The value 255 is defined in our application which refers to the frame structure defined in Figure 3(b) (specific to our application), 2. Size (8 bits): Integer from 0 to 255 which indicates the size of data included into the field Data (7th field), 3. Seq Num (8 bits): Integer from 0 to 255. It provides the sequence number of the current frame (1 frame/tag). The sequence number is used to determine the order of the different frames that have been written over the RFID tags (needed to rebuild the set of data items), 4. Last Num (8 bits): Integer from 0 to 255. It provides the sequence number of the last frame which contains data related to the same writing phase. The notion ”same writing phase” is important because one data item may be written at two different times in its life cycle and then, data inconsistency/conflict may occur, 5. ID Phase (64 bits): Integer from 0 to 264 which is the identifier of the writing phase (the date of writing is currently used). Several frames may have the same ID Phase but a couple ID Phase/Seq Num is unique. 6. Checksum (32 bits): Integer from 0 to 232. Used for data error-checking (it does not detect errors in the header), 7. Data (n − 128 bits): The content of the data item is added in this field, which is a string. This string may or may not be stored integrally in a unique RFID tag according to the technology (where n is the number of writable data bits in one RFID tag). Let us note that an index is added for locating each data item in the database (i.e. the table name, the attribute name and the instance concerned). In our method, the index is coded as follows: Tablename.Attributename.PrimaryKeyValue. For instance, when the data item TMat{3,3} is written (see Figure 1(b)), the index Material.Value.MD06 is added. Then, the data item content is added (i.e. 15◦C...).
Regarding the application layer, 128 bits (32 × 4) out of
n useful are used in each tag for the application header.
The application logic rebuilds each data item thanks to the
Seq Num and the Last Num, which indicate in what order data
items have been split. In section 4, the tag memory is
itemized after having been written. Let us note that according to
the set of data items carried by the product, queries may be
answered or unanswered. Nevertheless, some methods can be
deployed to know in advance (i.e. before data items will be
written on the product) if queries may be answered or not
(e.g. by transforming queries to corresponding bitmaps [
In this section, we develop an approach to know where information will be written on the material (in case of writing) or where information is located (in case of reading). In certain production processes like textile manufacturing, materials are transformed with highly automatised machines, at high manufacturing speed. However, some defects may occur on specific material zones (e.g., holes, stains). Because machines are not aware of these defaults, end products could not be sold. In our vision, materials are consider hard disks, able to store information about themselves. If a defect is detected, this information is then stored on the material and can be reused ad libitum when needed. The type of data manipulated is arearelated : consider a grease spot on the textile. Information related to the grease spot must be stored as close as possible to the real grease spot. As a result, information must be located. Our method is designed to help locating information on the material, thus enabling machines not to be blindfolded and to adapt their behavior. To do so, a more complex architecture is required than previously. Indeed, in the previous section, the data location over the material was not taking into account and therefore, data was written/read anywhere (’on the fly’) as illustrated in Figure 3(a). However, if the user desires to know where a specific information is located over the material, he needs to use a specific method which is detailed hereafter.
First, let C be the set of RFID tags present in the material. Let R be the set of RFID readers which are aligned as depicted in Figure 4 (e.g. on a ramp), with R = {r1, r2, r3}). Each RFID tag and RFID reader has a reference number respectively noted IDc and IDr with c ∈ C and r ∈ R. A reader r generates an event er,c if the presence of the tag c is detected. This event is made up of {IDr, IDc, tr,c}, with tr,c the acquisition time of the event. All the events form the set E.
Let us focus now on the algorithm which calculates the
theoretical positions of the tags in the material (i.e. ∈ C).
First, it is necessary to model the reading zone (i.e. zone
in which a RFID reader and a RFID tag can communicate)
which depends on the RFID technology. This reading zone
must be modeled as a cylinder3 as depicted in Figure 4 (the
three RFID readers have the same cylindrical shape). Once
the reading zone is modeled, the algorithm for computing the
location 2D of tags takes as input the set of events E. Since the
reading zone is modeled as a cylinder, it is possible to calculate
3 A methodology to model the reading zone of a given RFID
technology is presented in [
Storage of: TMat{3,1}, 3 TMat{3,2}, TMat{3,3} yct = yr ± rad2 − 1
Database system tr,cL − tr,cF 2 × v 2
In this scenario, we use the communicating textile reel
designed in [
2 xct , yct theoretical coordinates x-y of the tag c (c ∈ C) xr, yr coordinates in x and y-axis of the reader (r ∈ R) tr,cF first date to which the reader r detected the tag c tr,cL last date to which the reader r detected the tag c v conveyor speed (or material speed) ts textile detection date (thanks to a presence sensor) rad radius of the reading zone
The error made on the chord computation depends on both the acquisition time cycle and the conveyor speed. In fact, the higher the acquisition time cycle, the lesser the error. Obviously, the error depends also on the modeled reading zone (modeled as a perfect cylindrical shape). Figure 4 illustrates that the computed chords do not exactly overlap with the RFID tag c1 and c2. By using the chord approach, two possible chords are determined, which means that two possibilities are returned. If one tag is detected by only one RFID reader, it implies two possible coordinates yct (impossible to know if the real tag is located on the left or right chord). This is illustrated in Figure 4 with the tag c1 (only detected by r3). This problem does not occur when a tag is detected by several readers, as illustrated with c2 (detected by r1 and r2). Indeed, the tag might most likely be located between chr1r and chr2l , as shown in Figure 4. The higher the number of readers detect one tag, the higher the precision.
Retrieval of: TMat{3,1},
TMat{3,2}, TMat{3,3} 3 + location of these data items on the textile Writer device
Reader device
Activity X Storage of data
Activity T Retrieval of data then starts to write TMat{3,1}, TMat{3,2} and TMat{3,3} over the communicating textile thanks to the protocol of splitting described in section 3.1, as depicted in Figure 5.
As said before, the supplier in Activity X does not mind where information is written over the textile. As a result, we use the architecture given in Figure 3(a) (i.e. data is written ’on the fly’). Figure 6(a) shows the log events generated when writing the three data items over the textile. It can be observed that 4 tags are required for splitting the three data items. However, 5 tags have been written because one write operation failed (tag e007000002199de8). This highlights that all data items are stored on the material even if writing errors occur4. Now, let us focus on the datagram content of a given RFID tag. Figure 6(b) details the datagram contained in the tag e007000002199ddd after having written the three data items. It is important to note that a RFID tag memory (whatever the RFID technology) consists of several Data Blocks (denoted DB in Figure 6(b)). The RFID tags disseminated in our textile are the Omron’s V720-D52P03 and their memory is divided in 64 blocks of size 4 bytes5. Accordingly, the datagram header occupies the four first bytes of each tag memory as highlighted in Figure 6(b) (DB0 to DB3). Then, the remaining DB are exclusively reserved to the content of data items. However, let us remind that one index is added at beginning of each data item in order to locate it in the database (cf. section 3.1). Figure 6(b) shows the index related to TMat{3,2}, which is Material.Description.MD06. Then, the content of this data item (string of value: Textile which is provided with two layers of protective coatings) is added (cf. Figure 6(b)).
Then, the textile reel arrives at Activity Y in which the machine requires data carried by that one and must identify where it is located over the textile. 4.2
Since information must be located over the textile, the
architecture described in Figure 4 is implemented. Then, the
communicating textile passes under the ramp of RFID
readers and is read. The three data items are retrieved and their
content is displayed via an application software (JAVA
programming) as shown in Figure 7. Then, services can be
programmed: for instance, queries may be directly performed via
the JAVA software based on the set of data items retrieved
from the communicating textile (unanswered queries could
happen). Many services may therefore be imagined as the
example of Ley [
Datagram header Data item index: TMat{2,1} Data item content
Until this point, data items are retrieved from the communicating textile. To locate them over the textile, we implement the algorithm described in section 3.2. To assess the algorithm precision, the theoretical results will be confronted to the real tag positions. The virtual map given in Figure 8(a) details the real tag position as well as those returned by the algorithm (two possible solutions for one tag). These results have been obtained when the textile is read with a distance of 60mm from the ramp of readers6. If we look at the detailed case in Figure 8(a), the euclidean distance between the real tag position and those computed by the algorithm is equal to 7.01mm for the nearest position (chord 2) and 33.44mm for the furthest (chord 1). The focus in Figure 8(a) emphasized where the tags e007000002199de6, e007000002199de7, e007000002199ddd and e007000002199dde (which have been used for storing the 3 data items) are located on the textile.
For each tag, the minimal error is measured between the real tag position and the nearest position among the two computed by the algorithm (euclidean distance). The set of all the computed errors obtained is synthesized in a box-and-whisker diagram in Figure 8 (the reader is positioned 60mm above the textile). We can see that the minimal error (i.e. the minimal euclidean distance) is 7mm. 25% of the errors are inferior to 8mm (cf. the 1st quartile) and 25% are superior to 31mm (cf. the 3rd quartile). In average, the error is about 15mm with a reader positioned 60mm above the textile, with the conveyor speed equals to 4m/s and with an acquisition time 6 The maximum distance depends on the modeled reading zone.
the reader
33.44mm 7.01mm 45 40 5 cycle fixed to 5ms (errors depending on these 3 parameters). 5
New challenges and opportunities arise with concepts such as Internet of Things, Ubiquitous Computing and Artificial Intelligence. Nowadays, products are more and more fitted with electronic devices (e.g. sensors, RFID tags) which give them abilities such as data storage, decision making, monitoring. Some authors argued the usage of intelligent products in the framework of the supply chain management. Indeed, these products are able to control their own life, evolution and could serve as an interoperability hub between the supplier members. However, most of the time, products are only given an identifier (e.g. via a RFID tag) which provides a network pointer to a linked database and decision making software agent. As a result, a new kind of material is discussed in this paper referred to as ”communicating material” which allows to embed significant proportion of data directly on the manufactured product. In order to answer the question of what information is relevant to store on the product during its lifecycle, a data dissemination process (consisting of three steps) is presented and makes reference to previous works. One important step deals with the storage and retrieval of data on/from the communicating material. In this paper, an appropriate architecture of communication is developed which aims at splitting information over the ”communicating material” and at determining where this information is located.