RFID-assisted Product Delivery in Sustainable Supply
Chains: A Knowledge-based Approach
M. Ruta, F. Scioscia, E. Di Sciascio, F. Gramegna, S. Ieva and G. Loseto
Politecnico di Bari, via Re David 200, I-70125, Bari, Italy
Abstract. The paper proposes an integrated framework which uses knowledge
representation theory and languages to annotate relevant product information in
a semantically rich and unambiguous fashion so disclosing several added-value
services in multiple supply chain stages, without depending on a back-end in-
frastructure. EPCglobal RFID protocol standard has been extended and further
applied in an innovative supply chain model where production, packaging and
logistics have a reduced ecological impact. Particularly, on-product information
are leveraged in a more general Intelligent Transport System (ITS) framework
enabling advanced delivery scheduling and product tracking.
1 Introduction
Supply chains are articulated organisms involving organizations aimed at transferring
either products or services from a supplier to a customer. Nowadays, a successful supply
chain should more and more rely on an extended collaboration and integration among
component actors belonging to the productive and/or logistic network. Hence, infor-
mation has assumed an increasing strategic role in production, logistics and marketing.
From this standpoint, Radio-Frequency IDentification (RFID) is seen as a perspective
key technology. RFID enables radio interconnection of transponders –hosting informa-
tion associated to the goods to be identified– with interrogators able to extract carried
data. EPCglobal consortium
1
is one of the most active subjects involved in the mission
of a worldwide diffusion of RFID standards. Goods management during the delivery
stage of a supply chain is a complex process with criticisms tied to the packaging of
different and possibly incompatible products over multiple vehicles. Several systems
for product tracking which also enable integrity check have been proposed to improve
process quality in supply chains exploiting RFID technology [10]. Nevertheless, the
original identification mechanism –exclusively providing “true/false” replies– appears
as too restrictive for advanced applications. Furthermore, RFID-based technology usu-
ally relies on a stable and fixed back-end which makes every solution only partially
applicable to intrinsically volatile contexts such as the transportation and delivery ones.
On the contrary, given the increased storage availability (up to several kBs [1]) modern
transponders provide, RFID could provide further automation of actions and processes
1
EPCglobal, http://www.epcglobalinc.org
Ruta M., Scioscia F., Di Sciascio E., Gramegna F., Ieva S. and Loseto G.
RFID-assisted Product Delivery in Sustainable Supply Chains: A Knowledge-based Approach.
DOI: 10.5220/0003029100530065
In Proceedings of the 4th International Workshop on RFID Technology - Concepts, Applications, Challenges (ICEIS 2010), page
ISBN: 978-989-8425-11-9
Copyright
c
 2010 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved

related to item sorting and shipment, helping to prevent human errors and assist users
in decision making.
The paper presents a framework which exploits knowledge representation theoreti-
cal studies to annotate tagged objects which can so describe themselves in several sup-
ply chain stages, without depending on a back-end infrastructure and adhering to the
Ubiquitous Computing paradigm [22]. Canonical EPCglobal RFID identification has
been extended [16,7], providing semantic-based value-added services whose deploy-
ment and outcomes have been grounded in an advanced “green supply chain” setting.
Semantic Web languages such as OWL
2
and DIG [3] are used for building the lin-
guistic and semantic infrastructure underlying a networked and capillary exchange of
information among chain actors aimed at overcoming restrictions imposed by exist-
ing RFID solutions. Key features are: (i) to leverage a hybrid wireless communication
platform for enabling a distributed, backward compatible data management system; (ii)
to provide decision support exploiting semantic-based annotations accompanying and
describing goods aimed at optimizing the reliability and sustainability of the whole sup-
ply process. Case studies and simulations have evidenced usefulness and feasibility of
the framework. Distinctive features of this model enabled it to support several RFID-
based applications integrating Knowledge Representation techniques and technologies
to improve data management and business processes. The proposed approach has been
applied and tested in an innovative RFID-enabled supply chain model. On-product in-
formation have been exploited in a more general Intelligent Transport System (ITS)
framework enabling advanced context-aware delivery and tracking of fruits and veg-
etables. Integrity checks have been also introduced for process analyses, according to
the total quality management vision. Noteworthy is the capability to enable pollution-
aware dynamic path calculation, fully automated goods compatibility management in
delivery, truck fleet intelligent routing.
The remaining of the paper is structured as follows. In the next section relevant
related work is surveyed. Section 3 outlines the framework, explaining the followed
approach, Section 4 illustrates the system architecture thanks to a reference scenario
and finally conclusion and future work terminate the paper.
2 Related Work
This section surveysessential related work concerning RFIDtechnology in supply chain
management and ITS for building more efficient and sustainable goods delivery.
2.1 RFID
Benefits of RFID technology in supply chain management include timeliness, accu-
racy and completeness [9]. In latest years, they are becoming widely acknowledged not
only in distribution and warehousing, but also in the retail and post-sales domains. The
largest retail groups are mandating the adoption of interoperable technology solutions
2
OWL Web Ontology Language, version 2, W3C Recommendation 27 October 2009,
http://www.w3.org/TR/owl2-overview/
4

to commercial partners [18]. De et al. [6] first introduced a system for real time tracking
of items in a ubiquitous context. That work can be considered a reference for current
technological architectures for supply chain management, endorsed by worldwide spe-
cial interest groups such as the EPCglobal consortium [19].
Nevertheless, RFID has received relatively little attention as an information con-
veyor that can directly increase business process awareness and thus improve both
performance analysis and support to decision processes. Most current approaches for
run-time processing of RFID data in supply chains [21,2] manage only very basic in-
formation, namely raw (EPC, location, time) triples produced by RFID readers, where
EPC (Electronic Product Code) is the unique product identifier, while location and time
mark each RFID reading event.
In order to improve decision-making capabilities and information sharing across
boundaries of partner organizations, it is important to enable RFID-based run-time ob-
ject/group discovery facilities in decentralized and pervasive contexts. In each supply
chain node, it should be possible to process expressive requests –in terms of shared
and formal domain vocabularies– without depending on a central fixed information in-
frastructure. In our proposal, we adapt Semantic Web techniques and formalisms to
RFID-based supply chains. The basic goal is to fully characterize products equipped
with RFID tags by means of annotations in semantic languages such as RDF
3
, OWL
and its equivalent DIG. By means of formal ontologies, knowledge about a specific do-
main can be modeled, shared and exploited to derive new implied information from the
one stated within metadata associated to each resource.
Our proposal for semantic-enabled RFID [16] allows tags to contain a structured
and detailed description of product features, endowed with unambiguous and machine-
understandable semantics. Semantically annotated information is encoded in a compact
way by means of an algorithm aimed at efficient compression of XML-based ontolog-
ical languages [17]. Tag memory structure is extended in order to store the additional
information required for semantic-based object discovery. Additional TID (Tag Iden-
tification) memory space stores a 16-bit word for optional protocol features (currently
only the most significant bit is used to indicate whether the tag is semantic-enabled or
not, other bits are reserved for future uses) and a 32-bit Ontology Universally Unique
Identifier (OUUID) marking the ontology with respect to the description stored in the
tag is expressed. The encoded product annotation and contextual parameters (depend-
ing on the specific application) are stored, instead, within the User memory bank. This
is possible by adopting increasingly available models of passive EPCglobal UHF Class
1 Generation 2 RFID tags with several kilobits of available memory, such as TEGOTag
4
or Intelleflex IF602
5
. The air interface protocol for Gen 2 RFID systems is exploited but
neither new commands nor modifications to existing ones are introduced, thus keeping
full backward-compatibility with current RFID readers. In our simulations, we obtained
a reading rate of about 10 tags/s, which is comparable with products Gen 2 standard.
Further efficiency considerationsin terms of data compression, response time and mem-
3
RDF (Resource Description Framework) Primer, W3C Recommendation, 10 February 2004,
http://www.w3.org/TR/rdf-primer/
4
TEGOTag, http://www.tegoinc.com/products/products tegotag.php
5
Intelleflex Integrated RFID Tags, http://www.intelleflex.com/pdf/datasheet.c1g2ic.pdf
5

ory usage are reported in [17,7]. In this way goods auto-expose their description to
any RFID-enabled computing environment is reached. This enables decentralized ap-
proaches for context-aware application solutions, based on less expensive and more
manageable mobile ad-hoc networks. Furthermore, by combining standard and non-
standard inference services in Description Logics [4], several semantic-based match-
making schemes can be designed to meet goals and requirements of specific processes
and applications within a supply chain.
Few other proposals for semantic-based annotation of physical products can be
found in the literature. In [20] the adoption of the Part Libraries (PLIB) standard ISO
13584 was proposed and an XML Schema for data-on-tag storage was developed. Even
though the benefits of standardization are clear, the chosen standard provides only a
very rudimentary taxonomy, lacking explicit semantics for product characteristics. A
solution based on Semantic Web languages was proposed in [12] for ubiquitous com-
merce environments.In that case, however,RFID tags stored only a productcode, which
was used as a key to retrieve the corresponding RDF annotation from a central back-end
information system. This approach, inherited from traditional RFID applications, poses
major architectural and organizational challenges for information sharing in complex
multi-party supply chains. Conversely, our core idea is that, as physical products flow
among supply chain partners, ipso facto relevant high-level information about them is
conveyed [16] and can be exploited for meaningful business analysis at different lev-
els [7]. Recent studies [8, 23] highlighted factors in favor of data-on-tag approaches
w.r.t. traditional data-on-network ones. Mainly, fast data access is essential when the IT
infrastructure must meet real-time requirements, avoiding bottlenecks related to back-
end queries. Other benefits include: avoiding single points of failure; additional security
controls over information by consumer and companies; cost reductions in establishing
and maintaining network infrastructures.
2.2 Supply Chain and ITS
One of the most important factors in supply chain management is efficient control of
product delivery through vehicle routing. Vehicle routing problem (VRP) has always
been a fundamental problem in network optimization research and it is an important
objective of Intelligent Transport Systems (ITS). It was first introduced by Dantzig and
Ramser [5] and it can be defined as the problem of finding the fleet route planning that
serves as many jobs as possible at the least cost. Since that, many VRP types were
studied and classified based on additional problem constraints required by different
applications, and many algorithms and methods have been developed.
Ramachandran [15] developed an integer linear programming (ILP) model for the
design of routes that satisfy the load compatibility constraints for a fleet of vehicles,
transporting different types of goods. ILP and MILP (mixed integer linear program-
ming) are –along with Markov processes and genetic algorithms– the most widespread
approaches to solve this kind of optimization problems. Nevertheless, these methods
require that modeling and execution are performed in off-line mode. It is hard to adapt
the model to real-time applications and whenever constraints change.
In [11] an optimal solution was proposed for a VRP based on real-time environ-
ment information of perishable goods, exploiting continuous monitoring of RFID tags
6

located in the refrigerated cargo and of road traffic flow information. The approach,
however, requires large resources to compute the optimal solution. A further common
shortcoming of purely mathematical and statistical methods is that they cannot provide
explanation for outcomes, which is important to strengthen user trust in the system.
Mahr et al. [13] introduced a distributedarchitecture based on truck agents andorder
agents (containers) in a real world scenario constrained by uncertain job arrival time.
The system does not need any central monitoring and delegates decision making at local
level. Benefits of the distributed approach include better handling of local information
and quicker adaptive reaction to unpredictable order arrivals. They apply to our current
proposal as well.
Approaches at semantic-enhanced route planning have recently been attempted.
Niaraki et al. [14] introduced a technique which uses ontologies to annotate each road
segment with both user preferences and context point of view, in order to find cus-
tomized routes that best match specific needs. The solution was focused on personal
route planning, hence the model and the process are not easily adaptable to the require-
ments of supply chain VRP scenarios.
3 Framework and Approach
We propose an RFID-based delivery managementsystem extending classic supply chain
organization and shipment models using techniques and technologies for smart tagging
[16] and a semantic-based decision support [4]. Due to space constraints, the reader
is referred to cited works for an explanation of used Descritpion Logic languages and
algorithms. Product packages (tagged with RFID transponders conforming to the EPC-
global Class 1, Generation II UHF standard) and vehicles are accompanied by an OWL-
DL compressed annotation [17] which include relevant object properties/purposes and
vehicle features, according to a reference ontology. Product characteristics comprise
relevant item properties, delivery requirements (e.g., micro-climate and storage needs
or possible security measures) and potential incompatibility constrains referred to other
nearby products. Vehicle descriptions will contain general truck specifications along
with freight equipment information and the remaining load availability.
Starting from a single logistic unit where delivery of available goods has to be
planned, a semantic-based matchmaking process allows a smart allocation process aim-
ing at maximizing vehicle carrying capability also minimizing travel distance. In order
to take into account item-item and item-vehicle constraints in shipment schedule we
make use of a lightweight version of non-monotonicinferences[4], which automatically
detect the best load-truck associations by taking into account semantic compatibility de-
gree of goods among them and with vehicles. Such a system aims to reduce response
times in delivery decisions and improve efficiency of product allocation. Particularly,
the proposed approach can be effectively integrated into existing supply chain man-
agement systems extending already supported technology. It allows to solve on-truck
product allocation issues, ensuring products quality to destination (minimizing risks
due to incompatibility among closely delivered goods and reducing delivery times).
A system architecture schema is shown in Figure 1. Each tag is tracked within a
warehouse by either fixed RFID readers deployed in strategic locations or handheld
7

Fig.1. System Architecture.
ones. Truck informationcan be read when vehicles arrive at the stocking center, whereas
product tags are scanned individually before storage. Data extracted from RFID tags are
decompressed and sent to a mobile matchmaker which performs inference services.
Given a shipment order,the warehouse unit can automatically build a set of products
to be sent to each customer by means of a on-the-fly semantic matchmaking between
product information and truck features. The matchmaking process has to satisfy the
following goals.
– Products can be allocated only to trucks fulfilling their transportation requirements,
in terms of needed loading/unloading equipment, containers and internal environ-
mental conditions (e.g., temperature, humidity, lighting).
– Different products cannot travel together if they have negative mutual effects. A
typical case concerns climateric fruits (e.g., apples), which can influence ripening
of other fruits and vegetables.
– Product destinations must be taken into account, in order to avoid inefficient route
planning.
We suppose all goods which will compose a shipment are available in a single ware-
house and a delivery request has been addressed to retail endpoints, including different
items to be delivered with related destination and required quantity. A semantic-based
product allocation process is performed in the warehouse. Elements of the optimization
problem can be formalized as follows:
– The set of stored products P = {p
1
, p
2
, . . . p
n
}. In addition to semantically anno-
tated description referring to an ontology T , quantity q
i
and destination location l
i
are associated to each product, based on orders received from partner supply chain
nodes (e.g., retail stores).
– The set of available vehicles V = {v
1
, v
2
, . . . v
m
}. Besides the semantically anno-
tated description, the freight capacity is associated to each vehicle.
– l
0
is the warehouse location.
– A set T = {θ
1
, θ
2
, . . . θ
n
} is derived from P . For each product p
i
∈ P , the approx-
imated angle θ
i
of the line connecting the warehouse to the destination is computed
as θ
i
= ar ctan(
lat(l
i
)−lat(l
0
)
lon(l
i
)−lon(l
0
)
).
8

– A threshold semantic distance value 0 < s < 1 and a threshold angle θ
max
are de-
fined. The latter allows to avoid grouping products that have to be delivered toward
very different directions.
Algorithm 1: Greedy algorithm for product clustering.
Algorithm: clustering (hP, T, L, T i)
Require: L Description Logic, acyclic T , p
i
∈ P, i = 1, 2, . . . n concept expressions in L satisfiable in T .
Ensure: G = {G
1
, G
2
, . . . G
k
} set of product compatibility groups.
1: G := ∅
2: k := 0
3: while P 6= ∅ do
4: k := k + 1
5: pick p
i
∈ P
6: G
k
:= {p
i
}
7: C
k
:= p
i
8: for all p
j
∈ P, i 6= j do
9: δ
θ
= min
(p
q
∈G
k
)
(|θ
j
− θ
q
|)
10: if δ
θ
<= θ
max
AND (p
j
⊓ C
k
) is satisfiable in T AND
rankPotential (
h
L,p
j
,C
k
,T
i
)
rankPotential (
h
L,p
j
,⊤,T
i
)
<= s then
11: G
k
:= G
K
∩ {p
j
}
12: C
k
:=
d
p
q
∈G
k
(p
q
)
13: end if
14: end for
15: P := P \ G
k
16: G := G ∩ {G
k
}
17: end while
18: return G
The optimization strategy divides the problem into the following steps.
1. Product Clustering. For all available goods, the system verifies compatibility be-
tween product descriptions to cluster items in different compatibility groups. All prod-
ucts within each group present similar storage requirements and no compatibility con-
straints with other items are violated. The greedy bottom-up clustering Algorithm 1 is
exploited. Remarks follow.
– As a preliminary route optimization feature, a product cannot join a group if its
delivery direction is too different from all the other group elements. This check is
implemented by comparing the minimum angle distance δ
θ
to the threshold θ
max
.
– The algorithm exploits rankPotential [4] to discover semantic conflicts between
productcharacteristics/requirementsand to evaluate product similarity among mem-
bers of each cluster. It is important to note that the description of a group is seman-
tically expressed as the logical conjunction of individual item descriptions (which
is satisfiable by construction). Compatibility level is calculated by normalizing the
semantic distance between each new element p
j
and the current product group, as
returned by rankPotential , w.r.t. the maximum possible rankPotential value for p
j
,
which is rankPotential (hL, p
j
, ⊤, T i) and depends only on axioms in the reference
ontology. The interested reader is referred to [4] for details.
– It is easy to see that the algorithm requires O(n
2
) rankPotential calculations.
2. Cluster Allocation. Properties of product groups are subsequently matched with de-
scriptions of the delivery trucks, in order to find the most suitable vehicles for transport:
rankPotential inference service is exploited again to measure the semantic distance of
9

Table 1. Orders list.
Orders Delivery Distance (km) Quantity (kg)
Golden Delicious Barletta 53.75 500
Altamura Bread Foggia 117.06 300
Ostrich Egg Brindisi 105.30 200
Navelina Lecce 139.67 300
Cavendish Taranto 79.37 300
descriptions. At most nm rankPotential calculations are performed at this stage. The
product group with lowest score (i.e., the most compatible) is selected and inserted in
the suggested vehicle.
3. Storage Optimization. Now a refinement process can be performed to maximize ve-
hicle carrying capacity. If a vehicle is overfull (e.g., because the product group total
volume is larger than the vehicle storage space), the system splits the group and re-
allocates the exceeding portions to the vehicle with second best match score. Dually,
the most empty vehicles are gradually unloaded and products are reallocated to par-
tially full trucks, according to compatibility scores. Processing ends when all goods are
allocated and vehicles have minimum carrying space waste.
4. Route Planning. Detailed route planning is executed at the beginning of the deliv-
ery trip. Mobile device on board of a vehicle acts as a navigation system. Since the
previous step has grouped products with destinations along the same direction, in this
step the route is planned by means of a very simple algorithm: delivery point are sorted
according to increasing distance from the starting point. This approach resembles the
well-known SCAN algorithm for hard disk drive access scheduling: it is very quick and
simple to implement and, in our case, its approximation w.r.t. the optimal static route
improves when destinations span a narrower angle w.r.t. the source.
More sophisticated route planning algorithms can be plugged into the framework,
in order to minimize a cost estimation function that considers the overall economic
and environmental impact of route distance, trip duration, road type, traffic estima-
tion and truck load. Inference services upon semantically annotated information can be
leveraged also in this step. Using openly available and editable map data such as Open-
StreetMap
6
, maps can be customized with additional metadata about road segments and
points of interest, in order to better suit supply chain goals.
4 Case Study: Do not Compare Apples and Oranges!
In order to better explain the proposed framework and algorithms, let us consider a
practical example. A food cooperative in Apulia has adopted semantic-based supply
chain management process. Each production center ships goods to a distribution center,
located in Bari. From there products have to be delivered to shopping centers located
in different towns. The orders list is showed in Table 1. Figure 2 provides a visual
representation of the geographic area.
The problem is determining the most efficient method to ship the products so that
they are delivered without quality loss. As described previously, our proposed frame-
work uses mobile devices equipped with RFID reader. Warehouse operators, endowed
6
OpenStreetMap project, http://www.openstreetmap.org/
10

Fig.2. Map of supply chain nodes in our case study.
High Temperature ⊑ Controlled Temperature Medium Temperature ⊑ Controlled Temperature
Low
Temperature ⊑ Controlled Temperature Controlled Temperature ⊑ Temperature
Room
Temperature ⊑ Temperature High Humidity ⊑ Controlled Humidity
Medium
Humidity ⊑ Controlled Humidity Low Humidity ⊑ Controlled Humidity
Controlled
Humidity ⊑ Humidity Natural Humidity ⊑ Humidity
High
Oxygen ⊑ Controlled Oxygen Medium Oxygen ⊑ Controlled Oxygen
Low
Oxygen ⊑ Controlled Oxygen Controlled Oxygen ⊑ Oxygen
Natural
Oxygen ⊑ Oxygen Direct Lighting ⊑ Lighting Source
Indirect
Lighting ⊑ Lighting Source ISO Pallet Rack ⊑ Pallet Rack
Pallet
Rack ⊑ Stocking Equipment Plastic Shelving ⊑ Shelving
Shelving ⊑ Stocking
Equipment Stocking Equipment ⊑ Equipment
Hydraulic
Drill ⊑ Transport Equipment Transport Equipment ⊑ Equipment
High
Maturity ⊑ Maturity Low Maturity ⊑ Maturity
High
Fragrance ⊑ Fragrance Low Fragrance ⊑ Fragrance
disj(High
Oxygen, Medium Oxygen, Low Oxygen) disj(Controlled Temperature, Room Temperature)
disj(Controlled
Humidity, Natural Humidity) disj(Controlled Oxygen, Natural Oxygen)
disj(Direct
Lighting, Indirect Lighting) disj(Shelving, Pallet Rack)
disj(High
Maturity, Low Maturity) disj(High Fragrance, Low Fragrance)
Ripe
Product ⊑ ∃Has Climateric Maturation Degree ⊓ ∀Has Climateric Maturation Degree.High Maturity
Unripe
Product ⊑ ∃Has Climateric Maturation Degree ⊓ ∀Has Climateric Maturation Degree.Low Maturity
disj(High
Temperature, Medium Temperature, Low Temperature)
disj(High
Humidity, Medium Humidity, Low Humidity)
Fig.3. Axioms in the food transport ontology used in the case study
with RFID-enabled handheld devices, can locally manage the allocation problem and
they can also monitor automated system process behavior. For each tagged product,
the following information is retrieved via RFID: EPC code, unique identifier of the ref-
erence ontology [16], semantic-based annotation in compressed OWL-DL format and
delivery information, expressed with geographic coordinates.
In our supply chain case study, goods and vehicles are described according to an
example ontology devised for product management. Figure 3 shows a relevant excerpt
of it. Each RFID tag contains a semantic description w.r.t. the reference ontology,
summarizing both quality characteristics and storage and transport requirements.
Descriptions corresponding to Table 1 follow:
Golden Delicious: A pple ⊓ ∀ Has Colour.Y ellow ⊓ ∃ Has Colour ⊓
∀ Has Quality.Ordinary Quality ⊓ ∃ Has Quality ⊓
∀ Storage T emperature.Room T emperature ⊓ ∃ Storage T emperature ⊓
∀ Storage Humidity.M edium Humidity ⊓ ∃ Storage Humidity ⊓
∀ Storage Oxygen.N atural Oxygen ⊓ ∃ Storage O xygen ⊓ U nripe P roduct.
Altamura Bread: Bread ⊓ ∀ Has Quality.Ordinary Quality ⊓ ∃ Has Quality ⊓
∀ Storage T emperature.Room T emperature ⊓ ∃ Storage T emperature ⊓
11

∀ Storage Humidity.M edium Humidity ⊓ ∃ Storage Humidity ⊓
∀ Storage Oxygen.N atural Oxygen ⊓ ∃ Storage O xygen ⊓ Low F ragrance.
Ostrich Egg: Egg ⊓ ∀ Has Col our.W hite ⊓ ∃ Has Colour ⊓ ∀ Has Quality.T op Quality ⊓
∃ Has Quality ⊓ ∀ Storage T emperature.Low T emperature ⊓ ∃ Storage T emperature ⊓
∀ Storage Humidity.Low Humidity ⊓ ∃ Storage Humidity ⊓ ∀ Storage Oxygen.Low Oxygen ⊓
∃ Storage Oxygen ⊓ Low F ragrance.
Navelina: Orange ⊓ ∀ Has Co l our.Or a nge ⊓ ∃ Has Colour ⊓
∀ Storage T emperature.Room T emperature ⊓ ∃ Storage T emperature ⊓
∀ Storage Humidity.Medium Humidity ⊓ ∃ Storage Humidity ⊓ High F rag rance.
Cavendish: Banana ⊓ ∀ Has Co l our.Y ellow ⊓ ∃ Has Colour ⊓∀ Has Quality.Ordinary Quality ⊓
∃ Has Quality ⊓ ∀ Storage T emperature.Room T emperature ⊓ ∃ Storage T emperature ⊓
∀ Storage Humidity.M edium Humidity ⊓ ∃ Storage Humidity ⊓
∀ Storage Oxygen.N atural Oxygen ⊓ ∃ Storage O xygen ⊓ Ripe P roduct.
Warehouse operator uses the mobile logic-based matchmaker embedded in her de-
vice to identify groups of compatible products. As previously said, each cargo should
be composed of products that do not interfere with each other, causing a general quality
loss. In order to accomplish this, the system adopts a greedy approach, applying the
algorithm presented in the previous section. We better explain this process seeing how
it is applied to our example.
1. At the beginning, Golden Delicious is added to cargo
1
group.
2. Its semantic description is compatible with Altamura Bread, because all atomic
concepts, universal quantifiers and unqualified number restrictions on roles are
compatible. Furthermore, products are also compatible in terms of truck direction,
because the angle difference between delivery locations is about 10 degrees (with
a supposed threshold of 60 degrees). For this reason it is added into cargo
1
group.
3. Then Ostrich Egg is matched against cargo
1
: a semantic incompatibility is re-
turned, because some storage requirements are in conflict, e.g., storage temperature
is different.
4. Next, cargo
1
is semantically compatible with Navelina, but they are not compati-
ble w.r.t. truck direction, because the angle between destinations is more than 160
degrees. Finally, it is not compatible with Cavendish, because both are climacteric
and so they can travel in the same cargo only if they are in the same ripening stage,
but in this case apples are ripe while bananas are unripe.
If needed, the system can show inconsistencies to operator by means of the Concept
Contraction inference service offered by the reasoning engine [4]. Outcome explanation
is a very important feature and a unique advantage of approaches based on knowledge
representation.
The same process is repeated with remaining products to build other groups.
The system finally returns the following groups: cargo
1
= { Golden Delicious, Alta-
mura Bread} ; cargo
2
= {Ostrich Egg}; cargo
3
= {Navelina, Cavendish}.
Next phase consists of the allocation of each cargo on a compatible truck. The
RFID reader retrieves the semantic descriptions stored on tags associated with trucks
and begins a matchmaking process. For example, let us consider the following truck
descriptions in the warehouse:
12

Table 2. Matchmaking results between cargoes and trucks.
cargo
1
cargo
2
cargo
3
truck
1
0.132 n.c. 0.284
truck
2
0.154 n.c. 0.324
truck
3
n.c 0.432 n.c
Truck 1: ∀ Storage T emperature.Room T emperature ⊓ ∃ Storage T emperature ⊓
∀ Storage Humidity.M edium Humidity ⊓ ∃ Storage Humidity ⊓
∀ Storage Oxygen.N atural Ox ygen ⊓ ∃ Storage Oxygen ⊓
∀ Storage Equipment.Hydraulic Drill ⊓ ∃Storage Equipment ⊓
∀ Storage Lighting Source.Indirect Lighting ⊓ ∃ Storage Lighting Source.
Truck 2: ∀ Storage T emperature.Room T emperature ⊓ ∃ Storage T emperature ⊓
∀ Stor a ge Humidity.Controlled Humidity ⊓ ∃ Storage Humidity ⊓
∀ Storage Equipment.P lastic Shelving ⊓ ∃Storage Equipment ⊓
∀ Storage Lighting Source.Indirect Lighting ⊓ ∃ Storage Lighting Source.
Truck 3: ∀ Storage T emperature.Controlled T emperature ⊓ ∃ Storage T emperature ⊓
∀ Storage Humidity.Low Humidity ⊓ ∃ Storage Humidity ⊓
∀ Storage Oxygen.Controlled Oxygen ⊓ ∃ Storage Oxygen ⊓
∀ Storage Equipment.ISO P allet Rack ⊓ ∃Storage Equipment ⊓
∀ Storage Lighting Source.Direct Lighting ⊓ ∃ Storage Lighting Source.
Only if a truck is compatible, for each cargo the semantic distance is evaluated,
exploiting the rankPotential algorithm. The results obtained are ranked using the utility
function:
Rank(cargo
i
, truck
i
) =
rankPotential (cargo
i
, truck
i
)
rankPotential (cargo
i
, ⊤)
res idual space(truck
i
)
total space(truck
i
)
where: rankPotential (cargo
i
, truck
i
) is the semantic distance from cargo
i
to truck
i
;
rankPotential (cargo
i
, ⊤) is the maximum semantic distance from cargo
i
, which
depends on axioms in the domain ontology; residual spa c e(tru ck
i
) is the avail-
able space in the truck
i
after the allocation of the cargo
i
and total space(truck
i
)
is the overall space in truck
i
. The pair with lowest score will be selected. In
this way each cargo will be allocated to the truck that better satisfies its trans-
port requirements and that maximizes the truck load. Table 2 shows the results
of the computation, in case that each truck has a maximum capacity of 1 ton.
Results denote that, for example, cargo
1
is not compatible with truck
3
, due to
storage requirements not fulfilled by the truck. Also in this case, an explana-
tion of the causes for incompatibility can be obtained exploiting the Concept
Contraction inference service (give up: ∀Storage Oxygen.N atural Oxygen ⊓
∀Stora ge T emperature.Room T emperature).
At the end of the matchmaking process the cargoes will be arranged on the trucks
as follows: truck
1
: {Golden Delicious, Altamura Bread}; truck
2
: {Ostrich Egg};
tru c k
3
: {Navelina, Cavendish}.
The mobile system now performs optimization of storage space on different trucks.
Referring to previous example, truck
2
results partially empty. In this case, the sys-
tem can rearrange products moving them in other vehicle after checking compatibility
13

with truck features and already stored product descriptions. A possible approach could
exploit rankPartial and Concept Contraction algorithms to establish contrasting char-
acteristics. In this way, weakly incompatible products can be delivered on the same
truck to minimize unused carrying space.
Finally, products are loaded and a delivery schedule is planned for each truck.
Using the simple algorithm outlined in the previoussection, vehicle routes are computed
as follows: truck
1
: (Barletta, Foggia); truck
2
: (Brindisi); truc k
3
: (Taranto, Lecce).
5 Conclusions
The paper presented a novel supply chain model where semantic-enhancements to
RFID allows information exchange among actors involved in various stages of the good
life cycle. Benefits deriving from the adoption of such an approach have been proved
with reference to the environmental sustainability of products delivery in a generic fruit
and vegetable market. An RFID-assisted ITS relies on tagged goods information to per-
form: (i) fully automated goods compatibility management in load composition, (ii)
pollution-aware routing, (iii) intelligent delivery. Further work will be performed to
extend and improve the proposed approach. Future developments include studying so-
lutions to security issues specific of data-on-tag RFID approaches, whereas the current
framework uses the simple security methods provided by EPCglobal standards.
Acknowledgements
The authors acknowledge partial support of Apulia Region Strategic Project PS 025 -
Processes and technologies supporting quasi-markets in logistics.
References
1. B. R. Ayoub Khan M., Manoj S. A Survey of RFID Tags. International Journal of Recent
Trends in Engineering, 1(4):68–71, 2009.
2. Y. Bai, F. Wang, P. Liu, C. Zaniolo, and S. Liu. RFID data processing with a data stream
query language. In Proceedings of the 23nd International Conference on Data Engineering,
ICDE, pages 1184–1193, 2007.
3. S. Bechhofer, R. M¨oller, and P. Crowther. The DIG Description Logic Interface. In Pro-
ceedings of the 16th International Workshop on Description Logics (DL’03), volume 81 of
CEUR Workshop Proceedings, Rome, Italy, September 2003.
4. S. Colucci, T. Di Noia, A. Pinto, A. Ragone, M. Ruta, and E. Tinelli. A non-monotonic
approach to semantic matchmaking and request refinement in e-marketplaces. International
Journal of Electronic Commerce, 12(2):127–154, 2007.
5. G. B. Dantzig and J. H. Ramser. The truck dispatching problem. Management Science,
6(1):80–91, 1959.
6. P. De, K. Basu, and S. Das. An ubiquitous architectural framework and protocol for object
tracking using RFID tags. In The First Annual International Conference on Mobile and Ubiq-
uitous Systems:Networking and Services (MOBIQUITOUS 2004), pages 174–182, August
2004.
14

7. R. De Virgilio, E. Di Sciascio, M. Ruta, F. Scioscia, and R. Torlone. Semantic-based RFID
Data Management. In D. Ranasinghe, Q. Sheng, and S. Zeadally, editors, Unique Radio
Innovation for the 21st Century: Building Scalable and Global RFID Networks. Springer,
2010. to appear.
8. T. Diekmann, A. Melski, and M. Schumann. Data-on-network vs. data-on-tag: Managing
data in complex rfid environments. In HICSS ’07: Proceedings of the 40th Annual Hawaii In-
ternational Conference on System Sciences, page 224a, Washington, DC, USA, 2007. IEEE
Computer Society.
9. M. Karkkainen. Increasing efficiency in the supply chain for short shelf life goods using
RFID tagging. International Journal of Retail & Distribution Management, 31(10):529–536,
2003.
10. F. Li. Impact of rfid technology on supply chain: A simulation approach. In Management
and Service Science, 2009. MASS ’09. International Conference on, pages 1 –4, sept. 2009.
11. F. Li and Y. Wei. A real-time vehicle routing system for rfid-tagged goods transportation. In
Service Operations and Logistics, and Informatics, 2008. IEEE/SOLI 2008. IEEE Interna-
tional Conference on, volume 2, pages 2892 –2897, oct. 2008.
12. W. Maass and A. Filler. Towards an infrastructure for semantically annotated physical prod-
ucts. Informatik, 94:544–549, 2006.
13. T. Mahr, J. Srour, M. M. de Weerdt, and R. Zuidwijk. Agent performance in vehicle routing
when the only thing certain is uncertainty. In Proceedings of the workshop on Agents in
Traffic and Transportation (ATT), 2008.
14. A. S. Niaraki and K. Kim. Ontology based personalized route planning system using a multi-
criteria decision making approach. Expert Systems with Applications, 36(2, Part 1):2250 –
2259, 2009.
15. P. Ramachandran. Vehicle routing problem with load compatibility constraints. In Industrial
Engineering and Engineering Management, 2009. IEEM 2009. IEEE International Confer-
ence on, pages 339 –343, dec. 2009.
16. M. Ruta, T. Di Noia, E. Di Sciascio, F. Scioscia, and G. Piscitelli. If objects could talk:
A novel resource discovery approach for pervasive environments. International Journal of
Internet and Protocol Technology (IJIPT), 2(3/4):199–217, 2007.
17. F. Scioscia and M. Ruta. Building a Semantic Web of Things: issues and perspectives in
information compression. In Semantic Web Information Management (SWIM’09). In Pro-
ceedings of the 3rd IEEE International Conference on Semantic Computing (ICSC 2009),
pages 589–594. IEEE Computer Society, 2009.
18. M. Tajima. Strategic value of RFID in supply chain management. Journal of Purchasing and
Supply Management, 13(4):261–273, 2007.
19. K. Traub, G. Allgair, H. Barthel, L. Bustein, J. Garrett, B. Hogan, B. Rodrigues, S. Sarma,
J. Schmidt, C. Schramek, R. Stewart, and K. Suen. EPCglobal Architecture Framework.
Technical report, EPCglobal, July 2005.
20. C. Tribowski, K. Spin, O. G¨unther, and O. Sielemann. Storing Data On Rfid Tags: A
Standards-Based Approach. In Proceedings of the 17th European Conference on Information
Systems (ECIS 2009), 2009.
21. F. Wang, S. Liu, P. Liu, and Y. Bai. Bridging physical and virtual worlds: complex event
processing for RFID data streams. Lecture Notes in Computer Science, 3896:588, 2006.
22. M. Weiser. The computer for the 21st century. SIGMOBILE Mob. Comput. Commun. Rev.,
3(3):3–11, 1999.
23. K. Werner, E. Grummt, S. Gro, and R. Ackermann. Data-on-tag: An approach to privacy
friendly usage of rfid technologies. In Proceedings of the 3rd European Workshop on RFID
Systems and Technologies (RFID SysTech ’07), Duisburg, Germany, June 2009.
15