A Ubiquitous Knowledge-based System to enable RFID

Object Discovery in Smart Environments

M. Ruta, T. Di Noia, E. Di Sciascio, F. Scioscia and E. Tinelli

Politecnico di Bari, via Re David 200, I-70125, Bari, Italy

Abstract. This paper presents an extended framework supported by a suitable

dissemination protocol to enable ubiquitous Knowledge Bases (u-KBs) in perva-

sive RFID environments. A u-KB is a distributed and decentralized knowledge

base where the factual knowledge (i.e., individuals) is scattered among objects

disseminated within the environment, with no centralized repository and coordi-

nation.

1 Introduction

In pervasive contexts intelligence is embedded into a physical environment by means

of a relatively large number of heterogeneous micro devices such as RFID tags and

wireless sensors, each conveying a small amount of useful information. Due to power

and cost constraints, they are usually endowed with very low storage space, little or

no processing capability and short-range, low-throughput wireless links. Each mobile

host in the area can access information only on micro devices in its communication

range. Consequently, approaches based on centralized control and information storage

are utterly impractical in such scenarios.

In our previous work [1], we devised solutions for the integration of semantic-

enhanced EPCglobal RFID into Mobile Ad-hoc NETworks (MANETs). Nevertheless,

we were still forced to use a ﬁxed central component for reasoning over a Knowledge

Base (KB). This led to expensive information duplication within the environment: se-

mantic annotations were placed simultaneously on tags and within the KB; moreover,

the reasoning engine was a single point of failure. Here we want to show how these

problems can be mitigated if a more distributed approach is followed. We present a

general framework to carry out an advanced matchmaking using metadata stored in

RFIDs lacking unique and ﬁxed knowledge bases. An advanced resource discovery is

supported by a dissemination protocol allowing to exactly locate suitable descriptions

directly on tags attached to objects. The proposed framework comprises the speciﬁca-

tion of components and operations of a u-KB (ubiquitous Knowledge Base), as well as

a distributed application-layer protocol for dissemination and discovery of knowledge

embedded within RFID tags in a MANET-based computing environment. RFID readers

are adopted as cluster-heads w.r.t. tags in their radio range and they are able to automat-

ically build up a multi-hop communication infrastructure when placed in the same area

[2]. For lower network layers we adopt IEEE 802.11 [3], IP and UDP, while for knowl-

edge representation we exploit DL-based ontology languages originally conceived for

Ruta M., Di Noia T., Di Sciascio E., Scioscia F. and Tinelli E. (2008).

A Ubiquitous Knowledge-based System to enable RFID Object Discovery in Smart Environments.

In Proceedings of the 2nd International Workshop on RFID Technology - Concepts, Applications, Challenges, pages 87-100

DOI: 10.5220/0001741900870100

 SciTePress

the Semantic Web effort, particularly DIG [4], which is a more compact equivalent of

OWL-DL

The remaining of the paper is organized as follows: in the next Section we present

motivation of our paper; in Section 3 some signiﬁcant theoretical aspects of the pro-

posed approach are revised; in Section 4 we outline the proposed data propagation and

retrieval framework, whereas in Section 5 our proposal is explained and motivated by

means of a case study. Section 6 closes the paper.

2 Motivation

Motivation for this paper stems from our experience in using Knowledge Representa-

tion (KR) approaches –based on Description Logics (DLs)– in pervasive contexts, and

particularly in semantic-based discovery frameworks for objects equipped with RFID

tags [1]. In traditional applications (see [5, 1.5] for a survey) a Knowledge Represen-

tation System (KRS) plays an architectural role very much like a Database Manage-

ment System (DBMS). Both are used as central repositories where knowledge of the

problem domain can be inserted (forming a KB) and upon which automated inference

procedures can be executed in order to extract implicit knowledge. Hence, in traditional

KRSs, a KB is seen as a single large entity which is immediately available, either in

local storage or via a high-throughput network link. This approach is effective only as

long as large computing resources and a stable network infrastructure are granted. A

different approach is needed to adapt KR tools and technologies to functional and non-

functional requirements of mobile and ubiquitous computing applications. They are

characterized by user (and device) mobility, dependency on context, severe resource

limitations. Hence, knowledge-based systems designed for wired networks are hardly

adaptable, due to architectural differences and performance issues.

As Figure 1 shows, in our vision a u-KB layer provides access to knowledge em-

bedded into semantic-enhanced EPCglobal RFID tags populating a smart environment.

Discoveryand reasoning tasks can then be performed either by hosts in the local MANET

or by a remote entity through a gateway exposing a high-level interface (e.g., Web

Sevices of RPC (Remote Procedure Call) or REST (REpresentational State Transfer)

type) and translating remote methods into operations on the u-KB. This paper focuses

on the deﬁnition of the components of a u-KB layer, while parallel research effort is

being spent into the adaptation of reasoning procedures to resource-constrained mobile

devices.

2.1 Application Scenarios

All application areas of RFID technology [6] can be enhanced by a semantically rich

description and discovery layer, without depending on a centralized infrastructure. In

the lifecycle of industrial products, manufacturing and quality control can exploit accu-

rate descriptions of raw materials, components and processes. Sale depots beneﬁt from

OWL Web Ontology Language, W3C Recommendation, February 10th 2004, available at

http://www.w3.org/TR/owlfeatures/

Fig.1. Architecture of the proposed approach.

easier inventory management and can introduce ubiquitous commerce [7] capabilities

like in [8], withouth expensive investments in infrastructure. Finally, smart post-sale

services can be provided to purchasers, by integrating knowledge discovery in home

and ofﬁce appliances [1].

Asset management is greatly improved in those scenarios where retrieval should

depend on object properties and purposes, rather than mere identiﬁcation codes. In

healthcare applications, equipment, drugs and patients can be thoroughly and formally

described and tracked, enabling to provide not only monitoring but also decision sup-

port to clinicians. Wireless pervasive technologies help break barriers between patient

management in the hospital and at home. Likewise, in museums and archaeological

sites, smart semantic-based content fruition can be granted to local visitors as well as to

remote clients connected through the Internet, leveraging the lightweight infrastructure

already deployed for internal inventory.

The u-KB approach can be extended to other monitoring and sensing technologies

beyond RFID. Wireless semantic sensor networks [9] are an emerging yet challeng-

ing technology. Semantic-based sensory data dissemination and query processing are

needed to enable advanced solutions for e.g., precision agriculture and disaster recov-

ery.

3 From KBs to u-KBs

3.1 KB Components and Operations

A DL knowledge base has two components [5]: a TBox, containing intensional knowl-

edge in the form of an ontology describing general concepts and properties of the ref-

erence domain; an ABox containing extensional knowledge that is speciﬁc to the indi-

viduals of a particular problem within the domain.

Current KRSs are characterized in terms of what functions they provide to appli-

cations, instead of exposing system data structures and low-level operations [10]. Two

basic functions were identiﬁed for KB management. (1) Tell: build the TBox and the

ABox by explicit assertions of terminological knowledge and information about indi-

viduals. (2) Ask: extracting (implicit) knowledge by using inference procedures that

determine if the meaning of the query is implied by the information that has been told

to the system. This paradigm has led to detailed and formal interface speciﬁcations

for Knowledge Representation Systems (such as KRSS [11] and, more recently, DIG

[4]), implemented by most KRSs. The ability to remove information from a KB is also

desirable. Due to technical reasons, most systems allow to Un-Tell (i.e., retract) only

information that has been previously told explicitly [12]. Nevertheless, experience has

shown that, for the vast majority of applications, in production environments the TBox

seldom or never changes after an initial knowledge acquisition phase (see [5, ch. 8] for

a review).

3.2 u-KB Components

In our approach we preserve the distinction between TBox and ABox. The TBox is

contained in an ontology ﬁle, which can be managed by one or more mobile hosts.

We hypothesize that ontologies are deﬁned before object annotation and u-KB deploy-

ment and do not change during normal system activity. We adopt Ontology Universally

Unique Identiﬁers (OUUIDs) [13] to mark ontologies unambiguously and to associate

each individual to the ontology w.r.t. it is described.

The ABox is scattered within a smart environment, as KB individuals are physically

tied to micro devicesdeployed in the ﬁeld. In RFID-based scenarios, each individual is a

semantically annotated object/product description, stored within the RFID transponder

the object is clung to. Each annotation refers to an ontology providing the intensional

knowledge for a particular domain. In detail, each individual is characterized by: glob-

ally unique item identiﬁer (the EPC code in the case of RFID tags); OUUID; semantic

annotation, stored as a compressed document fragment in the DL-based language DIG;

a set of data-oriented attributes, which allow to integrate and extend logic-based reason-

ing services with application-speciﬁc and context-aware information processing. Since

several object categories can co-exist within the same physical space, multiple u-KBs

can actually populate the same environment and share the system infrastructure.

3.3 u-KB Operations

In our u-KB approach, we adhere to the classic Tell/Ask model. These functions, how-

ever, are implemented in a novel way, coping with the characteristics of pervasive com-

puting scenarios. Section 4 deﬁnes in detail the data structures and protocol devised to

build a u-KB system.

Tell/Un-tell operations are transparent to users, i.e., no explicit knowledge declara-

tion/retraction is required. The system allows autonomic creation and maintenance of a

“virtual” knowledge base, by means of a data alignment protocol between caches of the

various mobile hosts. Each host advertises individuals it sees in its proximity via RFID.

Other hosts store advertisements in their cache and forward them to nearby nodes. The

protocol keeps track of the freshness of advertised individuals through sequence num-

bers, so that only updates will be automatically propagated. Each individual has also

a limited Time-To-Live (TTL), so that it will be automatically removed (un-told) from

the u-KB if not renewed. The system tends toward a steady state where every host is

aware of all individuals in the environment.

Ask operations require a preliminary retrieval phase. The requester host speciﬁes

the ontology identiﬁer and a range for each particular attribute it is interested in. Ac-

cess to local cache tables provides IP addresses of hosts owning all the individuals that

meet the speciﬁed criteria. Pre-ﬁltered “on-demand” provisioning of KB individuals

avoids unnecessary data transfers, minimizes transmissions for data alignment and pre-

vents propagation issues in the case of description update or individual removal. The

requester can fetch ontology ﬁle and ﬁltered individuals from their respective providers,

so as to reconstruct a local subset of the whole KB, containing only the TBox and indi-

viduals which are actually needed. Then it is able to submit any Ask-type request to a

local or remote reasoning engine. In the present work we aim at assuring “on-demand”

knowledge availability, so we will not specify further how reasoning is carried out.

4 How to Build a u-KB

The proposed framework presents a two level infrastructure where RFID is exploited at

the ﬁeld layer (able to interconnect tags dipped in the environment and readers able to

receive the transmitted data) whereas the discovery layer is related to the inter-reader

ad-hoc communication (see Figure 2). The communication between the tag ﬁeld and

readers exploits the semantic-enhanced EPCglobal RFID protocol data exchange [1],

whereas the data propagation among readers is performed following a data dissemina-

tion paradigm in 802.11 [3] proposed here.

Thus, the resource discovery is based on three stages: (1) the extraction of good’s

parameters (for carrying object characteristics from ﬁeld layer to discovery one); (2)

resource data dissemination (to make the overall nodes fully aware of the “network

content”); (3) the extraction of resource annotations (for carrying semantic-based de-

scriptions from ﬁeld level to the discovery one) for the further matchmaking. Each

reader involved in the data propagation and/or in the object discovery, maintains a cache

containing the advertisements which will be matched against requests. It plays a central

role in the whole service oriented architecture as it advertises contextual parameters

discovery layer

802.11

RFID reader/Wi-Fi PDA

RFID tag

tag

RFID tag

tag

field layer

RFID EPCGlobal

RFID tag

field layer

RFID EPCGlobal

Fig.2. Field and Discovery layers in the proposed framework architecture.

referred to tags in its radio range (at the ﬁeld layer) and during the further phase (at the

discovery layer), it will receive requests from nearby nodes (via 802.11) and in case it

will extract semantic annotations from tags in range (via RFID EPCglobal) so replying

to the requester.

We hypothesize each resource in the MANET is labeled by means of the triple

[SOURCE ADDRESS, OUUID, EPC], where the ﬁrst value is the IP address of the

RFID reader which has “seen” the resource, the second one marks the speciﬁc reference

ontology the resource is associated with, the last one is the Electronic Product Code.

Initially, each reader will advertise, for each resource, the managed reference OUUID

as well as some context-aware parameter (i.e., the resource life time). So the initial

selection allows to choose only semantically compatible services (OUUID matching)

which have suitable values for context aware parameters.

To perform the data dissemination, resource providers periodically send advertise-

ment packets also specifying the maximum number of hops for the advertisement

travel (MAX ADV DIAMETER). The cyclic advertisement diffusion allows to cope with

tag and reader mobility. During their travel, the advertisements are forwarded using

MAC broadcasts and can be stored in the cache memories of the nodes they go through.

When starting a resource matchmaking, a node generally attempts to cover the request

by using resource descriptions stored within its own cache memory. If some seman-

tically annotated description is missing, it can be retrieved in unicast using apposite

demand PDUs. On the contrary, if a requester has no resource descriptions in its cache

or if managed resources are considered insufﬁcient to satisfy the request, the node can

send a solicit PDU with a speciﬁed maximum travel diameter (MAX REQ DIAMETER)

in order to get new resource locators. A node receiving a solicit, replies (in unicast) pro-

viding cache table entries matching parameters contained within the solicit frame. If it

does not manage any information satisfying the solicit, it will reply with a “no matches”

message. During their travel, replies to the demand and solicit PDUs are used to update

the cache memory of forwarding nodes. Figure 3 and Figure 4 show the typical se-

quence and involved actors of the data dissemination phase and of semantic annotations

retrieval, respectively.

Fig.3. Tag data dissemination phase.

Fig.4. Retrieval of advertised semantic annotations from tags.

Table 1. PDU types exploited in the proposed framework.

TYPE BIT SET PDU

A 0 Advertisement

B 1 Cache entry

C 2 Solicit

D 3 Demand

E..L 4..7 reserved

4.1 Advertisement PDU

In Figure 5 the structure of an advertisement PDU is sketched. All the resources inven-

toried by a reader are advertised by means of a unique advertisement PDU and then the

size of the packet increases proportionally with the number of tags in the reader range.

In what follows we outline PDU ﬁelds.

– TYPE: the kind of PDU (see Table 1).

– FLAGS: it contains one status ﬂag to distinguish the kind of transmission (uni or broadcast);

the remaining ﬂags are reserved for future purposes.

– TRAVELED HOPS: the number of hops already traversed by the packet. A reader sets this

value to 1 and it is increased every time a node forwards the packet.

– NUMBER OF RESOURCES: how many tags are in the reader radio range.

– ADVERTISEMENT ID: the reader’s sequence number.

– NODE SEQUENCE NUMBER: the sequence number of the node forwarding the packet. If

the packet has been sent by a reader this ﬁeld value coincides with the previous one.

– SOURCE ADDRESS: the IP address of the reader.

– RESOURCE PARAMETERS: a composite, variable length ﬁeld depending on the number of

advertised resources. In particular, it contains the OUUID value, the remaining life time of a

resource, the maximum hops number for the advertisement travel, and ﬁnally the EPC code

of the single tag.

Table 2. Constant values used in the framework.

NAME MEANING VALUE

DEFAULT RTIME Time interval between two consecutive advertisement packet

transmissions

30000 ms

POLLING TIME Time a reader node waits for the echo of the advertisements 7500 ms

MAX ADV JITTER Maximum value for random time waited when forwarding

advertisement packets

600 ms

ONE HOP WAIT Timer set by a requester node after sending a solicit packet, waiting for

cache contents reception

2000ms

HOP TRAVERSAL TIME Time a node needs to process and forward a solicit packet sent by a

neighbor

50ms

ACK RTT Timer set by a requester node waiting for ack after a solicit has been

sent

50 ms

DISCOVERY DIAMETER Current search diameter (in hops) during discovery phase 4

MAX RETRIES Maximum number of retransmissions before a reader assumes there are

no neighbors

A reader which has inventoried a tag series, broadcasts an advertisement every

DEFAULT RTIME milliseconds (see Table 2 for constant values). Nearby nodes for-

ward the packet by broadcasting it to their neighbors; as a consequence, the reader lis-

tens to the echo of the advertisement packet it originally transmitted. Thus, it can obtain

a conﬁrmation of the presence of other nodes in its neighborhood. If the reader does not

receive any echo within POLLING TIME milliseconds (less than DEFAULT RTIME),

it will retransmit the advertisement, assuming that a collision or a transmission error

has occurred. After MAX RETRIES retries it can be assumed there are no neighbors, so

the transmission of the advertisement can be scheduled after a longer timeout in order

to reduce power consumption.

When a node receives an advertisement, it extracts information about the resources

and, in case of “new” items, it adds cache entries; otherwise, before updating stored

data, the node veriﬁes if the received information is more recent or has ran across a

shorter path than the existing one. This happens also when multiple readers attempt

to scan the same tag. If the cache is updated and the maximum advertisement diam-

eter has not been reached, then the advertisement is forwarded; otherwise the whole

packet is silently discarded. This simple mechanism grants each mobile node in the

network sends the same advertisement at most once. Furthermore, in order to reduce

the collision probability (recall that MAC 802.11 protocol does not provide any ac-

knowledgment frame for broadcast transmission), each host waits a random time t

(t ∈ [0, MAX ADV JITTER]) before transmitting.

We developed an analytical model of the data dissemination protocol (not reported

here due to lack of space) to estimate whether the inclusion of the 96-bit EPC code

in each resource advertisement could lead to an unacceptable network overhead. Let

us consider a typical scenario with a partition-free network of 30 readers and 1000

tags per reader. Since typical read rates are not above 100 tags/s for EPCglobal Gen-2

UHF RFID [14], we deem that reasonably the advertisement period DEFAULT R TIME

should not be set below 30 s; moreover, let us set MAX ADV DIAMETER=4. In that

case, total trafﬁc generated by data dissemination has an upper bound of ≈ 590 kB/s,

i.e., ≈ 20 kB/s per reader. Furthermore, generated trafﬁc is a linear function of tag

population size, thus granting a theoretically acceptable scalability to the system.

Fig.5. Advertisement PDU.

Fig.6. Demand PDU.

4.2 Demand PDU

When starting a matchmaking, a node must ﬁrstly look within its cache table for entries

compatible with the request and, in case, it must require in unicast the corresponding

semantically annotated descriptions from their respective owners. A demand PDUs as

the one sketched in Figure 6 is used. In what follows the meaning of introduced PDU

ﬁelds is summarized.

– TYPE: it is set to 3.

– FLAGS: analogous to the corresponding ﬁeld of the advertisement PDU.

– TRAVELED HOPS: how many hops the frame has already gone across.

– LAST HOP SEQUENCE NUMBER: the sequence number of the last node processing the

request.

– DESTINATION SEQUENCE NUMBER: the sequence number of the destination node.

– DESTINATION ADDRESS: the address of the last node processing the request.

– PROVIDER ADDRESS: is the address of the destination node.

– OUUID: the ontology’s unique identiﬁer.

– NUMBER OF REQUESTS: the number of requested resource descriptions.

– DATA: the size of this ﬁeld depends on the number of requests; it contains the EPCs of the

tags whose descriptions are required.

4.3 Solicit PDU

Basically, the soliciting mechanism is analogous to the advertising one. Figure 8 shows

the format of a solicit packet. Not reported PDU ﬁelds are analogous to the previous

ones.



0 4 10 148 byte18 19 23 35

source

address

OUUID timestamp

resource descriptionsequence

number



address

size lifetime

traveled

hops

EPC

number

Fig.7. The structure of a record in the reader cache.

– TYPE: it is set to 2.

– FLAGS: it maintains the ordinary structure and functionality.

– TRAVELED HOPS: hops the packet has already gone across.

– TOTAL HOPS: total hops the PDU has to skip. Together with the TRAVELED HOPS ﬁeld,

it regulates the frame travel.

– REQUEST ID: unambiguously labels the PDU in order to distinguish different solicit re-

quests.

A node generating a solicit packet waits for an acknowledgment from each neigh-

bor for ACK RTT seconds. By means of this frame the requester elicits information

about nearby nodes: it can exactly know the number of neighbors. Each node located

at DISCOVERY DIAMETER hops from the requester, after receiving a solicit PDU,

replies with a cache content PDU in unicast toward the node the solicit cames from.

Readers receiving a cache content PDU update their own cache and recursively sends

back a cache content PDU, till the original requester node receives information it needs.

4.4 Cache Table Management

Each reader manages a cache table where it stores characteristics of both tags in its

radio range and tags it has “seen” in the network. Figure 7 shows the structure of a

typical entry. Here the content of each ﬁeld is reported.

– Source address: the address of the resource provider.

– Size: the description size (in byte).

– OUUID: a numeric identiﬁer for the speciﬁc ontology.

– Lifetime: the remaining time to live of a resource/tag.

– Timestamp: it marks the last reference to the entry (read/write).

– Traveled hops: distance (hops number) between provider and cache holder.

– Sequence number: it is referred to the last resource provider.

– EPC: the Electronic Product Code of a speciﬁc resource/tag.

– Resource description: the semantically annotated description of a resource. It will

have a variable length, but in some case there could be a pointer to the compressed ﬁle

containing the DIG description.

An entry could be added to the cache table whenever the node receives an adver-

tisement or a cache content frame arrives. Note that if a node receives an advertisement

frame, the corresponding entry of the cache table is updated either if the sequence num-

ber within the packet is higher than the one stored or if the route the PDU suggests is

shorter than the previously stored one.

Fig.8. Solicit PDU.

Fig.9. Cache content PDU.

A cache content packet (whose format is depicted in Figure 9) has a variable length

according to the number of resource handles the PDU transports. PDU ﬁelds are out-

lined in what follows.

– TYPE: it is set to 1.

– FLAGS: it maintains the ordinary meaning.

– N: the number of resources handles (and then cache tuples) the packet transports.

– REQUEST ID: is the identiﬁer of the original request.

– OUUID: the identiﬁer of the reference ontology.

– LAST HOP SEQUENCE NUMBER: the sequence number of the node sending the packet.

– DESTINATION ADDRESS: requester IP address.

Last ﬁelds are the resource records.

5 Case Study

5.1 RFID-based u-KB

The proposed framework has been studied speciﬁcally in pervasive computing envi-

ronments where a wide range of objects/products are endowed with RFID transpon-

ders conforming to the EPCglobal standard for class I - second generation UHF tags

[15]. Mobile RFID readers equipped with IEEE 802.11 wireless connectivity (hereafter

hosts) are responsible for u-KB creation and management. Tagged objects represent

KB individuals in our system. In our previous work [1] the EPCglobal standard was en-

hanced to support storage and retrieval on RFID tags of an ontology identiﬁer, a set of

attributes and a compressed semantic-based annotation, in addition to the EPC identi-

ﬁer. These basic elements recur in our deﬁnition of an individual in a u-KB, as explained

above. For the sake of conciseness, here we omit details of the semantic-enhancedRFID

protocol proposed in [1] so assuming the reader be familiar with it.

Table 3. SELECT command able to detect only semantic enabled tags.

PARAMETER Target Action MemBank Pointer Length Mask

VALUE 100

000

00010101

00000010

DESCRIPTION SL ﬂag set in case of match, EPC memory initial address number of bits bit mask

clear otherwise bank to compare

Table 4. READ command able to extract OUUID from the TID memory bank.

PARAMETER MemBank WordPtr WordCount

VALUE 10

000000010

00000010

DESCRIPTION TID memory bank starting address read up to 2 words (32 bits)

5.2 Interactions with EPCglobal RFID Technology

1. Dissemination. After each advertisementperiod DEFAULT TIME, a host scans RFID

tags in its range. Only semantic enabled tags are preselected, by means of a Select com-

mand with parameters as shown in Table 3.

EPC codes of semantic based tags are then scanned individually and TTL of cor-

responding cache table entries are refreshed. If the host detects a new EPC code not

present in its cache table, it will read its OUUID and contextual attributes, exploiting

two Read commands, as shown in Table 4 and Table 5 respectively. Data extracted from

the RFID tag will be stored in a new cache entry with a fresh sequence number. Con-

versely, if the EPC code is not detected for an existing local entry in the cache table, the

host will wait for the TTL to expire before removing the entry. This prevents the well-

known issue of “RFID event ﬂickering”

[16] from causing incorrect removal/addition

of u-KB individuals. At the end of the loop, the cache table is fully updated and the host

can issue an advertisement PDU to notify individuals to neighboring hosts.

2. Discovery. When a host receives a demand PDU for which it is the destination, it

starts an RFID scan of semantic enabled tags only, as seen above. During inventory, for

each detected EPC among those listed in the PDU payload, it reads the semantically

annotated compressed object description stored in the User memory bank of the tag,

with a Read command as in Table 6. Finally, the host replies to the requester.

3. Ontology Provisioning. As already suggested in [1], the EPCglobal Object Naming

Service (ONS)

can be used as a fallback mechanism for ontology support in ubiquitous

computing contexts if an Internet connection is available.

4. System Evaluation. After a feasibility study using IBM WebSphere RFID middle-

ware, we are currently performing an extensive simulation campaign in ns-2

of the full

protocol stack, in scenarios with both ﬁxed and moving readers. We expect valuable

insight for a thorough evaluation of the proposed approach, in particular w.r.t. network

load, duration of service discovery sessions, hit ratio (percentage of successful resource

retrieval) and sensitivity to topology variations (due to RFID reader and tag mobility).

Due to lack of space, we are not able to report here our early ﬁndings.

Due to collisions, a tag might not be detected in every consecutive scan. This phenomenon can

trigger spurious leave-enter event pairs.

Object Naming Service (ONS - v. 1.0), EPCglobal Ratiﬁed Speciﬁcation, Oct. 4, 2005,

http://www.epcglobalinc.org

ns-2, the network simulator, http://www.isi.edu/nsnam

Table 5. READ command to extract contextual attributes from User memory bank.

PARAMETER MemBank WordPtr WordCount

VALUE 11

000000000

00001000

DESCRIPTION user memory bank starting address read up to 8 words (16 bytes)

Table 6. READ command to extract the compressed semantic annotation from User memory

bank.

PARAMETER MemBank WordPtr WordCount

VALUE 11

000001000

00000000

DESCRIPTION user memory bank starting address read up to the end

6 Conclusions

Building on our previous work, we have presented an approach to carry out an ad-

vanced matchmaking using semantic metadata stored in RFID tags without unique and

ﬁxed knowledge bases. An advanced resource discovery framework is supported by a

knowledge dissemination protocol, so allowing an “on-demand” retrieval of suitable

descriptions directly from tags located on the objects. An analytical model has been de-

veloped for a preliminary assessment of resource requirements, while functionality tests

have been carried out in a reduced simulation environment. We are currently working

on a thorough evaluation of the approach using ns-2 simulation environment.

Acknowledgements

We wish to acknowledge support of Apulia project PE 074 “IC Technologies for track-

ing of agricultural and food products equipped with RFID tags”.

References

1. Ruta, M., Di Noia, T., Di Sciascio, E., Scioscia, F., Piscitelli, G.: If objects could talk: A novel

resource discovery approach for pervasive environments. International Journal of Internet

and Protocol Technology (IJIPT), Special issue on RFID: Technologies, Applications, and

Trends 2 (2007) 199–217

2. Ramanathan, R., Redi, J.: A brief overview of ad hoc networks: Challenges and directions.

IEEE Communications Magazine 40 (2002) 20–22

3. 802.11, I.: Information Technology Telecommunications and Information Exchange be-

tween Systems Local and Metropolitan Area Networks Speciﬁc Requirements Part 11:

Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Speciﬁcations.

ANSI/IEEE Std. 802.11, ISO/IEC 8802-11. First edn. (1999)

4. Bechhofer, S., M¨oller, R., Crowther, P.: The DIG Description Logic Interface. In: Proceed-

ings of the 16th International Workshop on Description Logics (DL’03). Volume 81 of CEUR

Workshop Proceedings. (2003)

5. Baader, F., Calvanese, D., Mc Guinness, D., Nardi, D., Patel-Schneider, P.: The Description

Logic Handbook. Cambridge University Press (2002)

6. Weinstein, R.: Rﬁd: A technical overview and its application to the enterprise. IT Profes-

sional 07 (2005) 27–33

7. Watson, R., Pitt, L., Berthon, P., Zinkhan, G.: U-Commerce: Expanding the Universe of

Marketing. Journal of the Academy of Marketing Science 30 (2002) 333–347

8. Di Noia, T., Di Sciascio, E., Donini, F., Ruta, M., Scioscia, F., Tinelli, E.: Semantic-based

bluetooth-rﬁd interaction for advanced resource discovery in pervasive contexts. Interna-

tional Journal on Semantic Web and Information Systems (IJSWIS) 4 (2008) 50–74

9. Ni, L., Zhu, Y., Ma, J., Li, M., Luo, Q., Liu, Y., Cheung, S., Yang, Q. In: Semantic Sensor

Net: An Extensible Framework. Springer Berlin / Heidelberg (2005) 1144–1153

10. Levesque, H.: Foundations of a Functional Approach to Knowledge Representation. Artiﬁ-

cial Intelligence 23 (1984) 155–212

11. Patel-Schneider, P., Swartout, B.: Description-Logic Knowledge Representation System

Speciﬁcation. (KRSS Group of the ARPA Knowledge Sharing Effort)

12. Brachman, R., McGuinness, D., Patel-Schneider, P., Resnick, L., Borgida, A.: Living with

CLASSIC: When and how to use a KL-ONE-like language. Principles of Semantic Networks

(1991) 401–456

13. Ruta, M., Di Noia, T., Di Sciascio, E., Donini, F.: Semantic-Enhanced Bluetooth Discovery

Protocol for M-Commerce Applications. International Journal of Web and Grid Services 2

(2006) 424–452

14. Kawakita, Y., Mistugi, J.: Anti-collision performance of Gen2 air protocol in random error

communication link. In: Proceedings of the International Symposium on Applications and

the Internet Workshops - SAINT 2006. (2006) 68–71

15. Traub, K., Allgair, G., Barthel, H., Bustein, L., Garrett, J., Hogan, B., Rodrigues, B., Sarma,

S., Schmidt, J., Schramek, C., Stewart, R., Suen, K.: EPCglobal Architecture Framework.

Technical report, EPCglobal (2005)

16. R¨omer, K., Schoch, T., Mattern, F., D¨ubendorfer, T.: Smart Identiﬁcation Frameworks for

Ubiquitous Computing Applications. Wireless Networks 10 (2004) 689–700

100