FORMALIZATION OF AN RNA-INSPIRED MIDDLEWARE

FOR COMPLEX SMART OBJECT FEDERATION SCENARIOS

emie Julia

, Yuzuru Tanaka

and Nicolas Spyratos

Meme Media Laboratory, Hokkaido University, Sapporo, Japan

Laboratoire de Recherche en Informatique, Universit

e Paris-Sud 11, Paris, France

Keywords:

Ubiquitous Computing, Pervasive Computing, Service Federation, Smart Object.

Abstract:

This paper proposes a new approach to deal with the smart objects or smart mobile devices by proposing

a middleware framework inspired by RNA mechanisms in molecular biology. Our framework represents

complex application scenarios of autonomic federation of smart objects as catalytic reaction networks. Each

catalytic reaction is modeled as an RNA expressions from a DNA. Our framework is capable of dealing not

only with the two stereotyped scenarios of ubiquitous computing, i.e. location-transparent service continuation

and location-and/or situation-aware service provision, but also with much more complex federations scenarios

of smart objects.

1 INTRODUCTION

Nowadays, our environment is ﬁlled with computing

devices such as RFID tags, chips with sensors and/or

actuators, smart phones, PDAs, intelligent electronic

appliances, embedded computers, and access points

to network servers: computing devices called smart

objects (Tanaka, 2008)(Tanaka, 2010). However,

we are using only a tiny portion of their potential-

ities since we generally do not dynamically and/or

ﬂexibly connect them to create a federated complex

ubiquitous environment. Ubiquitous computing has

been addressing just two stereotyped kinds of sce-

narios: location-transparent service continuation, and

location- and/or situation-aware service provision. It

was pointed out that the absence of formal model is an

obstacle preventing us from ﬁnding new kinds of sce-

narios (Milner, 2004)(Henricksen et al., 2002). Some

researchers are trying to extend the application target

of ubiquitous computing by using formal computation

models of process calculi, e.g., Bigraphical Reactive

System (Milner, 2001). These trials mainly focus on

mathematical description and inference of the behav-

ior of a set of mobile objects, but not those of the dy-

namically changing interconnection structures among

mobile physical objects based on abstract description

of their interfaces. That is why they are not sufﬁcient

to describe new kinds of scenarios. A lot of other

studies have been conducted on the mathematical de-

scription of network topology for network reconﬁgu-

ration and rerouting. These studies focused on phys-

ical connectivity among nodes, but not on their logi-

cal or functional connectivity. Thus they cannot de-

scribe application frameworks. These models are not

sufﬁcient to provide a basis for application frame-

works that go beyond the existing stereotyped sce-

narios. We believe that any essential extension be-

yond these stereotypes will require formal modeling

that can describe more complex federation scenarios

among smart objects.

In conventional research studies, federation mech-

anisms were based on the matching of a ser-

vice requesting message with a service provid-

ing message. This matching was made by a

centralized-and-repository lookup service (as in the

case of Linda(Gelernter, 1985)), or by a distributed

repository-and-lookup service (as in the case of

Lime(Picco et al., 1999)). However, the messages

to be matched depend on the implementation, which

tells why these architectures cannot describe dynamic

changes of federation structures independently from

the implementation deﬁning the behavior of each

smart object. Our approach is focused on the feder-

ation of smart objects through their interfaces. Each

smart object has a set of ports corresponding to ser-

vice providing ports or to service requesting ports.

Thus the federation of the smart objects does not de-

pend on their implementation, nor on a repository-

and-lookup service. This means that, since our ap-

proach does not require any repository, we have a new

Julia J., Tanaka Y. and Spyratos N..

FORMALIZATION OF AN RNA-INSPIRED MIDDLEWARE FOR COMPLEX SMART OBJECT FEDERATION SCENARIOS.

DOI: 10.5220/0003818400960105

In Proceedings of the 2nd International Conference on Pervasive Embedded Computing and Communication Systems (PECCS-2012), pages 96-105

ISBN: 978-989-8565-00-6

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

federation model for purely P2P networks.

The formal models that we aim to provide deal

with the following three different levels:

• First Level: The port matching model is used

to describe federation and interoperation mecha-

nisms.

• Second Level: Graph rewriting rules are used to

describe dynamic change of federation structures.

• Third Level: The catalytic reaction network mod-

eling is used to describe complex application sce-

narios with mutually related more than one feder-

ation.

A catalytic network reaction is a network of catalytic

reactions in which a product of a reaction may work

as a source material for another reaction or as a cata-

lyst to enable (stimulus) or inhibit (inhibitor) another

reaction. In our research the materials are smart ob-

jects. To implement this system, we use what we call

nucleotide smart objects. The nucleotide smart ob-

ject framework is described at the second layer of our

formal model, i.e.; it describes the dynamic change

of federation structures. In this framework, smart ob-

jects have behaviors similar to the nucleotides in the

RNA world hypothesis. By simulating the replication

of RNA, we can describe all kinds of catalytic reac-

tion networks that we are interested in (i.e., federation

and unfederation with or without a stimulus as a cat-

alyst). This paper is focusing on the mapping from

the third level to the second level. In the next session,

we will deﬁne what a catalytic reaction network de-

notes. In the following section, we will shortly review

the deﬁnition of smart objects. After that, we will

show what nucleotide smart objects denotes and how

we can use them to model each reaction in catalytic

reaction networks. Then, we will present the global

state transition diagram that summarizes all the rules

of the nucleotide smart objects and helps us to explain

the whole process of implementing each catalytic re-

action. Finally, we will give an example application

of this new framework.

2 CATALYTIC REACTION

NETWORK

Stuart A. Kauffman believes that a collectively auto-

catalytic set is one of the essential mechanisms for the

self-organization of life (Kauffman, 2000). We be-

lieve that complex application scenarios of autonomic

federation of smart objects can be modeled as a cat-

alytic reaction network. A catalytic reaction network

is a set of reactions in which the product of each reac-

tion may work as a source material of another reaction

or as a catalyst to enhance or to repress another reac-

tion. Each reaction is either a composition to produce

a compound product from more primitive source ma-

terials, or a decomposition of a source material into

its component materials.

In our modeling, the materials are smart objects.

For example, Fig. 1 shows four reactions (R1, R2, R3

and R4).

All characters outside a circle represent a smart

object or a composite smart object. A reaction be-

tween them can happen with the inﬂuence of a stimu-

lus (for example the reaction R3 between PH and T to

produce PHT will happen only with the presence of

MI) or without the inﬂuence of a stimulus (for exam-

ple the reaction R1 between M and I to produce MI),

and with or without a speciﬁc context that is made of

smart objects and represented by an object in each re-

action circle. The result material of a reaction may

work as an input material (for example PH), or as a

stimulus (for example MI) of another reaction. Both

stimuli and contexts work as catalyst. Stimuli are mo-

bile, while contexts are immobile.

This application allows an automatic federation of

wireless headphones and PDA depending on the users

mobility in a museum. A user with a social ID card I

and a smart member card M passes a check-in gate G1

of an museum. This gate sets up a federation between

M and I to check if he is a registered member of this

museum. The user could pick up at the entrance of

the museum a headphone and a PDA. When he will

pass through a second gate G2 working as a context,

his PDA and headphone will be automatically con-

nected. Later when he will come close to a table T,

the federation MI works as a stimulus which may also

contain his preferences. The federated objects PDA-

headphone could become connected with the table to

retrieve the related information. The user will receive

sound and text contents in his favorite language and/or

in his favorite color contrast for the text. If the user

does not have a member card, he should have a ticket

u. If a user come close to a table T, through the inﬂu-

ence of u, his federated PDA-headphone object will

not be connected with T, but with T’ that provides

generic information about the table T for non mem-

ber of the museum. No user operation is necessary to

set up the necessary connections of this scenarios.

PHT

M: member card

I: social Id card

P: PDA

H: headphone

T: table

T': generic table

u: ticket

C1: first gate

C2: second gate

C3: table context

C4: generic table context

PHT'

Figure 1: Catalytic reaction network representing a scenario

of smart-object Federations.

FORMALIZATION OF AN RNA-INSPIRED MIDDLEWARE FOR COMPLEX SMART OBJECT FEDERATION

SCENARIOS

In addition, we have two kinds of reactions, com-

position and decomposition reactions respectively

corresponding to federating and defederating actions.

In this paper, we will focus on composition reactions

with a context and with or without a stimulus.

3 SMART OBJECT

In this section, we shortly review the deﬁnition of

smart objects (SOs).

3.1 Ports

Each port has:

• a type;

• a polarity, + or − (the + polarity means that the

port is a service-providing port; the − polarity

means that the port is a service-requesting port);

• an arity (a service-requesting port has an arity of

one; a service-providing port can have an arity

greater than one);

• a state, a port can be in one of the two states: Vis-

ible (V) or Hidden (H);

The type and polarity are permanent characteristics of

a port, whereas the state of a port can change.

A SO can have many ports but no two of them can

have the same type and the same polarity (but they can

have the same type if they have different polarities).

3.2 Connections

Here are described the characteristics of a connection

from a SO A to a SO B:

• There can be a connection from A to B only if A

has a requesting port and B has a providing port of

the same port type. The common port is deﬁned

to be the type of the connection.

• There can be at most one outgoing connection

from a given port of a smart object.

• There can be any number of incoming connec-

tions to a given port of a smart object.

• There can be many connections from A to B using

different ports.

3.3 Smart Object Graph (SOG)

A smart object graph is an edge-labeled directed

graph without self loops, in which:

• The nodes are smart objects;

• There is an edge labeled t from a node A to node

B if there is a connection of type t from A to B.

• Each node is associated with the following ﬁve

properties:

– the smart object identiﬁer denoted oid;

– the smart object type;

– the smart object state;

– the set of its visible ports;

– the set of its hidden ports;

The set of these properties is called the description

of the node. Note that, in a SOG, a path σ from a

source SO A to a target SO B is uniquely deﬁned

by the sequence of the labels of its edges. This

follows from the fact that a requesting port can

have at most one connection.

A SOG is accompanied by another graph called

the proximity graph (PG) deﬁned as follows:

• the nodes are those of SOG;

• there is an edge from A to B if B is in the scope of

where the scope of A is deﬁned to be the set of smart

object whose existence can be detected by A, say,

through wireless communication.

3.4 Graph Rewriting Rules

Graph rewriting rules are the second-level formal

modeling of proximity-based federation that was in-

troduced by (Tanaka, 2010). It focuses on the dy-

namic changes of interconnection structures among

smart objects in a single complex federation. This

model describes a rule as two SOGs. Both of them

are sharing the same nodes. The ﬁrst one, called the

condition part, represents the condition of the appli-

cation of the rule. The second, called the action part,

corresponds to the change result of the condition part

after applying some primitive actions. The case where

the rule can be applied is illustrated by showing the

condition on the left hand side. The result of the rule

application is speciﬁed by the SOG on the right hand

side. The condition part SOG can be extended with

edges from its proximity graph.

Primitive Conditions. Each primitive condition is

presented in Fig. 2. The gray node means the smart

object that checks each condition and executes the

rule. It is called the activated node. For each con-

dition, the path σ may be of length zero. A proximity

graph edge is represented by a dotted arrow.

Port Condition

: the σ object has a certain port p

in a certain state. The letter “v” is used for the visible

state, and the letter “h” for the hidden state.

PECCS 2012 - International Conference on Pervasive and Embedded Computing and Communication Systems

Port Connections: the σ object has a port with a

certain number of active connections (0 is the exam-

ple). Note: the notation -p:v(1) denotes that a port -p

is visible with 1 active connection.

Type Condition: the σ object is of type t.

State Condition: the σ object has the state S.

Scope Condition: the σ2 object is in the scope of

the σ1 object.

We deﬁned a list of commands that will be used

to check the previous conditions and allow to get the

oid, the type, the state, the state of one of the ports, the

number of active connections of one of the ports, the

presence of a σ2 object in the scope, and the presence

of an object with a certain port in the state visible in

the scope of a σ object.

We deﬁned a list of commands that will be used to

check the previous conditions and allow us to get for

a σ object its oid, its type, its state, the state of one

of its ports, the number of active connections of one

of its ports, the presence of a σ2 object in its scope,

and the presence of an object with a certain port in the

state visible in its scope.

Primitive Actions. Each possible action is shown

in Fig. 3.

Port Exposition: the activated node makes the port

-p of the σ object visible.

Port Hiding: the activated node makes the port -p

of the σ object hidden.

State Setting: the activated node sets the state of

the σ object to S.

Connection Breaking: remove the edge labelled p

from the σ object.

Channel Spanning: span an edge labelled p from

the σ1 object to the σ2 object.

Figure 2: Primitive conditions that can be used in graph

rewriting rules.

Figure 3: Primitive actions that can be used in graph rewrit-

ing rules.

Channel Spanning to Neighbor: span an edge la-

belled p from the σ object to an object with a -p port.

We deﬁned a list of commands that will be used

to execute the previous actions and allow us to get for

a σ expose of hide one of its ports, to set its state, to

span or to break a connection to a σ2 object or to an

object in its scope.

Complex Rules. As shown in Fig. 9, it is possible

to combine the different primitive conditions and ac-

tions to deﬁne such a complex rule as the rule 2. For

σ1 = nil, and σ2 = B[x], this rule check that the σ1

object and the σ2 object have the state a and 0, and

that the σ1 object is in the scope of the σ2 object. The

actions consists to set the state of the σ1 object and

the σ2 object to c and 2, and to span a connection

from the σ2 object to the σ1 object through the port

B[x].

4 NUCLEOTIDE SMART

OBJECTS AS ELEMENTS OF

CATALYTIC REACTION

NETWORKS

4.1 Nucleotide Smart Objects

Nucleotide smart objects (NSOs) are a solution to im-

plement the primitives of the third layer: the catalytic

reaction network. The idea is to emulate the RNA

replication process in the smart object world to design

reactions of a catalytic reaction network in a generic

way.

We use different types of NSOs corresponding to

different types of SOs. For example, PDAs, credit

cards, or earphones are different types of SOs. We

use a different type of NSOs for each different type of

them. A NSO of subtype t is a SO in which there are

exactly four ports:

• two ports with the same type L; one of them

is service-requesting and the other is service-

providing, hence denoted as -L and +L respec-

tively;

• two ports with the same type; the type of these

ports is dependent of the subtype of the NSO. It

is denoted as B[t] where t is the subtype of the

NSO; one of them is service-requesting and the

other is service-providing, hence denoted as -B[t]

and +B[t] respectively;

• each of these ports can have at most one connec-

tion at any given moment;

FORMALIZATION OF AN RNA-INSPIRED MIDDLEWARE FOR COMPLEX SMART OBJECT FEDERATION

SCENARIOS

Strand of Nucleotide Smart Objects. Before

deﬁning the inputs of a reaction, we deﬁne a strand

as an SOG, and some operations on a strand. A strand

is a maximal acyclic path such that:

1. all nodes are NSOs;

2. all edges have the same label L;

3. no two nodes in the path are connected by a B

edge.

Given a strand V : v

→ v

→ ··· → v

, a comple-

ment of V is deﬁned to be any strand

V such that:

V : u

→ u

··· → u

2. type(u

) = type(v

n−i+1

), i = 1, ··· , n

4.2 Inputs Strands of a Reaction

The input materials of a reaction may be a set of k

input strands, V

·· ·V

; there is at most one distin-

guished input strand that plays the role of a stimu-

lus. We assume (without lost of generality) that, if a

stimulus is involved, the strand V

is the stimulus. We

also assume that each NSO of subtype t used in input

strands is already federated with a SO of type t. For

example in Fig. 4, you have an example of the input

materials of the context C3 presented in Fig. 1.

4.3 Context of a Reaction

Another element of a reaction is a strand called con-

text deﬁned as follows:

→ S

→

→ S

→ ··· →

k−1

→ S

k−1

→

V k

where type(S

) = I (Input separator) , i = 1, ··· , k − 1

and if V

is a stimulus, then type(S

) = S (Stimulus

separator).

Later, each

of a context is called a docking part.

For example in Fig. 5 is presented the implemen-

tation of the context C3 presented in Fig. 1.

Figure 4: Input materials of the reaction R3 in Fig. 1 using

C3 and MI as its context and stimulus.

Figure 5: The context C3 of the reaction R3 in Fig. 1.

4.4 Output of a Reaction

If V

is a stimulus, then the output is

→ V

→ ··· → V

; else, the output is

→ V

→ ··· → V

In Fig. 8 is presented the output of the reaction

that happens in the context C3.

4.5 Implementation of a Reaction

A context strand will work as a template deﬁning how

to federate the objects. For example C3 (Fig. 5) is a

context of a reaction accepting as input a stimulus MI

and two input materials PH and T. When the corre-

sponding SOs M, I, P, H and T attached to their NSOs

(Fig. 4) enter in the scope of the context C3, each

NSO of the input strands become automatically con-

nected to the corresponding docking part of the con-

text (Fig. 6). After this step, the connection among

the NSOs in the context through the ports L is repli-

cated in the opposite direction among the NSOs in the

stimulus and in the input materials (Fig. 7).

There, we ﬁrst need to break all the connections

between the context strand and both the stimulus and

the input materials, and then need to break the con-

nection between the stimulus and the object P. As a

result, we have the federation PHT (Fig. 8).

Figure 6: The docking of the stimulus strand and two input

strands.

Figure 7: Composition of an output strand from two input

strands.

Figure 8: The undocking ot the stimulus strand and the out-

put strand.

PECCS 2012 - International Conference on Pervasive and Embedded Computing and Communication Systems

100

Because T is indirectly connected to H and P

through the NSOs, some software that had been stored

in the context can be sent to T to establish arbitrary di-

rect connections among the three SOs P, H and T.

Suppose that the object P can be any kind of these

PDAs. We can consider for example a Microsoft Win-

dows Mobile PDA, and a Palm OS PDA. We can

develop a NSO, for the Microsoft Windows Mobile

PDA and another for the Palm OS PDA, having the

same subtype “N

PDA

”, and will communicate with

other NSOs with the same protocol. We will be able

to use all these different PDAs in the same federation

context without redesigning them.

All the reactions designed in the same way share

the same implementation structure. This structure is

composed of:

• the ﬁrst layer, that is called the ﬁrst strand or the

context;

• the second layer, that is called the second strand,

corresponding to the set of input strands;

• the third layer, corresponding to the SOs, work-

ing as a stimulus or as input materials, to be com-

bined.

The rest of the paper will focus on the ﬁrst and sec-

ond layers of the structure, i.e., the ﬁrst strand and the

second strand.

4.6 Graph Rewriting Rules for

Nucleotides Smart ObjectS

We assume that the context is manually constructed

as follows.

• The ﬁrst node of the context has the state 0 and

the ports +B[N

] : v(0), −B[N

] : h(0), +L : h(0)

and −L : h(1).

• The last node of the context has the state: 1 and

the ports +B[N

] : h(0), −B[N

] : h(0), +L : h(1)

and −L : h(0).

• Each intermediate node of the context has the

state 1 and the ports +B[N

] : h(0), −B[N

] : h(0),

+L : h(1) and −L : h(1).

The context with the above mentioned nodes is said

to be in the initial state. As we shall see, after obtain-

ing the output, the context is automatically reset to its

initial state.

In each input strand, if a strand contains a single

node, then the description of this node includes the

followings:

• the state a and the ports +B[N

] : h(0),

−B[N

] : v(0), +L : h(0) and −L : h(0);

else, the description of these nodes includes the fol-

lowings:

• The ﬁrst node of the input strand has the state

b and the ports +B[N

] : h(0), −B[N

] : v(0),

+L : h(1) and −L : h(0).

• The last node of the input strand has the state a and

the ports +B[N

] : h(0), −B[N

] : h(0), +L : h(0)

and −L : h(1).

• The last node of the input strand has the state

1 and the ports +B[N

] : h(0), −B[N

] : h(0),

+L : h(1) and −L : h(1).

We will make an assumption that the connections

are stable and not broken unless explicitly broken.

Furthermore, we will assume that we will get only

the expected input strand for each docking part. If not,

and if the input strand has already started to dock with

the docking part, we should undock the input strand,

and try to ﬁnd another one. The rules managing un-

expected input strands, however, will not be given in

this paper because of the page limitation.

In the Fig. 9 are presented the graph rewriting

rules of the NSOs. For the rules 1 to 15, when the type

of a SO is not speciﬁed, the type is N. In the rules 16

to 22, a smart object can be of type N, S or I. By fol-

lowing these rules, step by step, the input strings will

dock at the docking part of the context from left to

right, and then the different input strings will horizon-

tally be connected together. The NSOs are undocked

from the context from right to left. Then we get a new

federation of NSOs corresponding to the federation

of the different input strings. This new federation is

called the output material of the reaction and it can be

used as an input string of another reaction.

5 CORRECTNESS OF THE

GRAPH REWRITING RULES

FOR NSOS

It is difﬁcult to visualize and to verify the full pro-

cess that is obtained from these rules. To solve this

problem, here, we will introduce a Global State Tran-

sition Diagram (GSTD) to summarize the rule exe-

cution process. The ﬁrst part of this section presents

the graph rewriting rules as mathematical rules and

how to generate the global state transition diagram

corresponding to these mathematical rules. The sec-

ond part of this section is a description of the GSTD

that is obtained from the NSO graph rewriting rules.

And in the third part, we use this GSTD to explain the

whole process of a reaction.

FORMALIZATION OF AN RNA-INSPIRED MIDDLEWARE FOR COMPLEX SMART OBJECT FEDERATION

SCENARIOS

101

Rule 1(x)

Rule 3(x,y)

2 2

+B[x]:v(0)

-B[x]:v(0)

B[x]

B[y]

+B[y]:h(0)

-B[y]:h(0)

+B[y]:h(0)

-B[y]:h(0)

B[x]

Rule 5(x,y)

B[y]

B[x]

B[x] B[x]

2 1

Rule 7(x,y) (or7’ (x,y))

6 2

S (or I) S (or I)

B[x]

B[T

]

B[x]

c c

Rule 8(x)

B[x]

S (or I) S (or I)

B[x]

B[y]

B[x]

B[y]

8 2

f c e

B[y]

Rule 13(x,y)

B[y]

f f

Rule 14(x,y)

B[y]

f f

Rule 4(x,y)

2 2

B[x]

B[y]

2 2

B[x]

B[y]

Rule 10(x,y) (or 10’ (x,y))

S (or I) S (or I)

B[x]

B[y]

B[x]

B[y]

6 8

e c c

Rule 6(x,y)

B[y]

B[x] B[x]

B[y]

B[x] B[x]

S (or I)

B[x]

B[y]

f e

S (or I)

B[x]

B[y]

f f

Rule 2(x)

B[x]

Rule 9(x)

B[x]

B[x] B[x]

h g

Rule 21(x,y)

Rule 22(x)

Rule 19(x)

g b

h f

Rule 20(x,y)

g b

Rule 16(x)

4 5 5 1

Rule 17(x,y)

5 0

Rule 18(x)

B[y]

f f

Rule 15(x,y)

B[y]

f h

B[x]

+L:h(0) +L:h(0)

+L:h(0)

-L:h(0)

+L:h(0)

+B[y]:h(0) +B[y]:v(0)

+L:h(1)

+B[x]:v(0)

-L:h(0)

+B[x]:v(0)

+L:h(1)

-B[x]:v(0)

-L:h(0)

+B[x]:v(0)

+B[x]:h(0)

-B[x]:h(0)

-L:h(0)

+L:h(0) +L:h(0)

-L:h(0)

+L:h(0) +L:h(0)

+L:h(1)

+B[x]:v(1)

-L:h(0)

+B[x]:v(1)

Rule 11(x,y) (or 11’ (x,y))

+L:h(0)

+L:h(1)

Rule 12(x,y) (or 12’ (x,y))

-L:h(0)

+L:h(0)

-L:h(0)

+L:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[x]:h(0)

-B[x]:h(0)

-L:h(0)

+B[x]:h(0)

-B[x]:h(0)

-L:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[y]:h(0)

-B[y]:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[y]:h(0)

-B[y]:h(0)

+B[x]:h(0)

-B[x]:h(0)

+L:h(0)

-L:h(0)

+B[x]:v(0)

-B[x]:h(0)

+L:h(0)

+B[x]:h(0)

-B[x]:h(0)

+L:h(0)

-L:h(0)

+B[x]:h(0)

-B[x]:v(0)

+L:h(0)

-L:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[y]:h(0)

-B[y]:h(0)

+L:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[y]:h(0)

-B[y]:h(0)

+L:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[y]:h(0)

-B[y]:h(0)

+B[x]:h(0)

-B[x]:h(0)

+B[y]:h(0)

-B[y]:h(0)

+B[x]:h(0)

-B[x]:h(0)

-L:h(0)

+B[x]:h(0)

-B[x]:v(0)

-L:h(0)

+L:h(0)

-L:h(0)

+L:h(0)

+B[y]:h(0)

-B[y]:h(0)

+B[y]:h(0)

-B[y]:h(0)

+L:h(1)

-B[x]:v(0)

-L:h(0)

+B[x]:v(0)

Figure 9: Graph rewriting rules for nucleotides smart object.

5.1 Graph Rewriting Rules and their

Dependencies

5.1.1 Precondition and Postcondition

The precondition of a rule is the SOG in the condition

part of the rule, where we neglect all edges from the

proximity graph. The postcondition of a rule is the

SOG in the action part of the rule.

We deﬁne the following functions:

PreCd(R): to return the precondition of the rule R;

PostCd(R): to return the postcondition of the rule R.

5.1.2 Injective SOG Homomorphism

An injective SOG homomorphism (ISOGH) from an

SOG X to an SOG Y is an injection f from the nodes

of X to the nodes of Y such that:

• For every node n of X , n and f (n) have the same

type and the same state.

• For every node n of X , if both n and f (n) have

ports of the same port type and of the same polar-

ity, then these ports must have the same state and

the same number of active connections.

• For every edge in X labelled t from a node n

to a

node n

, there exists in Y an edge labelled t from

f (n

) to f (n

X is embedded in Y (denoted X ≺ Y ) if and only if

there exists an ISOGH from X to Y .

5.1.3 Joinability between Two SOGs

It is possible to join two SOGs X and Y through an

SOG Z if and only if it exists an ISOGH f

from Z to

X and another ISOGH f

from Z to Y such that:

• for every edge labelled t of X from a node n, is

there exists a node m in Z such that f

(m) = n,

them there is an edge with the same label t in Z

from the node m, or there is no such an edge in Y

from f

(m);

• for every edge labelled t of X to a node n, is there

exists a node m in Z such that f

(m) = n, them

there is an edge with the same label t in Z to the

node m, or there is no such an edge in Y to f

(m);

A join consists of replacing the nodes of X corre-

sponding to the nodes of Z by their respective cor-

responding nodes in Y to obtain a SOG W , as shown

in Fig. 10. The joinability function takes as its inputs

three SOGs X, Y and Z, and returns 1 as its output if

the two SOGs X and Y can be joined through a com-

mon SOG Z, or 0 otherwise.

In the example shown in Fig. 10 f

and f

are

two ISOGHs from Z to X, f

is one ISOGH from Z

to Y and the node ids are written in the black box.

PECCS 2012 - International Conference on Pervasive and Embedded Computing and Communication Systems

102

Figure 10: Example joinability between two SOGs.

For example, let us consider the two SOG homo-

morphisms f

and f

in Fig. 10. It is not possible to

join X and Y through Z because there is an edge la-

belled L in X from the node f

(5) (corresponding to

the node 1), but there is no edge with the same label

L in Z from the node 5, and furthermore, there is an

edge with the same label L in Y from f

(n) (corre-

sponding to the node 7).

However, if we consider the two ISOGHs f

and

, it is possible to join the two SOGs X and Y , be-

cause,

• for every edge labelled t of X from a node f

(n),

either there is an edge with the same label t in Z

from the node n, or there is no edge with the same

label t in Y from f

(n) , and

• for every edge labelled t of X to a node f

(n),

either there is an edge with the same label t in Z

to the node n, or there is no edge with the same

label t in Y to f

5.1.4 Modiﬁed Part of a Postcondition

The modiﬁed part of a postcondition of a rule R is a

subgraph of this postcondition that is obtained by re-

moving all the edges that keep the same description

and/or keep the same adgency nodes with the same

edges, between the precondition of R and the post-

condition of R.

We deﬁne a function ModPostCd(R) that returns

the modiﬁed part of the postcondition of the rule R as

an SOG. For example, in Fig. 11 ModPostCd(R) is

the SOG within an oval.

Figure 11: The modiﬁed part of the postcondition of an ex-

ample rule.

5.1.5 Rule Dependency

Let SOGs be the set of all the possible SOGs.

The rule dependency (RD) is a directed graph

= (V

, E

) where the set V

of vertices is the set

of rules and the set E

of edges is deﬁned as the fol-

lowing set:

= {(R, S)|(R, S) ∈ V

∧

∃F ∈ SOGs (∃G ∈ SOGs

(G ≺ F ∧ G ≺ ModPostCd(R))∧

joinability(PostCd(R), PreCd(S), F) = 1))}

(1)

An edge from a rule R to a rule S denotes that

the execution of the rule R may lead to the trigger-

ing of the rule S. the execution of a rule R may trig-

ger the execution of a rule S if and only if it is possi-

ble to join PostCd(R) and PreCd(S) through an SOG

F, and if this SOG F contains at least one node of

ModPostCd(R).

5.1.6 Extended Rule Dependency

The extended rule dependency (ERD) is a directed

graph where the set of vertices is the set of precon-

ditions and postconditions of the rules. It has an solid

edge from the precondition to the postcondition of the

same rule. It has also an dotted edge from a postcon-

dition of one rule to a precondition of another rule if

there exists an edge in the rule dependency from this

postcondition this precondition.

5.2 Global State Transition Diagram

The GSTD is deﬁned based on the simpliﬁed ERD di-

agram. The GSTD shown in Fig.12 summarizes only

the rule 1 to 15 presented in Section 4.6. Each vertex

of this diagram represents an SOG (either a precondi-

tion, or a postcondition). Here, types and ports are not

speciﬁed. In each vertex, all the horizontal connec-

tions are made through ports -L and +L, and all the

vertical connections through ports −B[x] and +B[x].

We will use capital letters associated with each vertex

to refer to its SOG. Each solid edge between vertices

represents an execution of a rule, and is called a “tran-

sition”. Each number used as a label to a transition

corresponds to the number of the rule to be executed.

Each dotted edge from a postcondition of a rule R to

a precondition of a rule S denotes that, after the ex-

ecution of the rule R, an SOG corresponding to the

precondition of the rule S may be obtained and the

rule S may be executed. For example, in Fig. 13, by

making the transition C to D, the SOG D is obtained.

At the same time, D also creates I. Thus, dotted edges

are called “transition dependencies”. Each transition

FORMALIZATION OF AN RNA-INSPIRED MIDDLEWARE FOR COMPLEX SMART OBJECT FEDERATION

SCENARIOS

103

Figure 12: A part of the global state transition diagram for the graph rewriting rules in Fig. 9.

is identiﬁed by the concatenation of the source SOG

name and the destination SOG name. For example,

the transition from C to D is called the transition CD.

Figure 13: Transition dependency.

When in the RD diagram we have a sequence of

rules (R

, R

, ...R

) such that each rule (except R

)

has only one outgoing edge to its following in the

sequence, and each rule (except R

) has only one in-

coming edge from its predecessor in the sequence, we

deﬁne in the GSTD only the precondition of the ﬁrst

rule of this sequence, and the postcondition of the last

rule of this sequence. For example, from the RD dia-

gram, we could see that after the execution of the rule

7 (or the rule 7’), only the rule 8 can be executed, and

after the rule 8, only the rule 9 can be executed. Thus,

the transition IJ corresponds to the execution of the

sequence of rules 7 (or 7’), 8 and 9.

When there exists a dotted edge from a postCd(R)

to a preCd(S), and if postCd(R) ≺ preCd(S) and

preCd(S) ≺ postCd(R), we may merge the postcon-

dition of the rule R with the precondition of the rule S.

For example, postCd(3) ≺ preCd(4) and preCd(4) ≺

postCd(3). The only difference is that there are two

primitive conditions in 4 (two ports of type L should

be in their hidden state with 0 active connection, see

Fig. 9), that are not in 3. Thus, the transition CD

represents the execution of the rule 3, and the transi-

tion DE represents the execution of the rule 4. D is

the postcondition of the rule 3 and also the precondi-

tion of the rule 4. We may add the two extra primi-

tive conditions to the transition DE, the white node in

state 2 in the SOG D should be the last nodes of the

ﬁrst strand and the white node in state c in the SOG D

should be the last node of the second strand. Because

the ports and the types are not speciﬁed in the GSTD,

the port and type conditions may be expressed as extra

conditions on the transitions. These extra conditions

are directly inherited from the graph rewriting rules.

There are three different types of basic conditions that

are used and combined to specify each transition con-

ditions:

Type Conditions: They check the type of a node.

For example in the transition IJ, the condition “gray

node 1 M” means that the gray node in state 1 should

be of type M (input Material separator);

Subtype Condition: They compare the subtypes

of two NSOs. For example in the transition IJ, the

condition “white node 1 = white node a” means that

the white node in state 1 should have the same subtype

as the white node in state a;

Position Conditions: They check the position of

a node in the context. For example, in the transition

AB the condition “black node 1 and black node a ﬁrst

node”. There are only four positions to be speciﬁed.

The ﬁrst node of the context strand (a node with a

+L:0 port), the ﬁrst node of an input strand (a node

with a -L:0 port), the last node of the context strand

(a node with a -L:0 port) and the last node of a input

strand (a node with a +L:0 port).

5.3 Outline of the Correctness Proof of

our NSO Rules

The whole process is divided into four stages: dock-

ing, output strand formation, undocking and reinitial-

ization. Each stage consists of “propagation of SOGs”

over the global structure. Basically, for each transi-

tion, other SOGs reached through dotted arrows may

be also obtained as explained previously. This corre-

sponds to the “SOG propagation”.

Docking: This stage connects different input

strands to the context strand.

Initialization: By the transition AB, we get either

PECCS 2012 - International Conference on Pervasive and Embedded Computing and Communication Systems

104

C or I.

Propagation: From C (, or I), it is only possible to

perform CD (, or IJ). Both of these transitions connect

a node of the context to a node of an input strand .

After the transition CD (, or IJ), if the white node in

state 2 is not the last node of the ﬁrst strand, we will

obtain either C or I, and, hence, we will repeat the

same step. Else we can only obtain the transition DE

(, or JK), to obtain a E (, or K).

Input Strand Interconnection: When the last

node of the last input strand is connected to the last

node of the context, all the input strands are correctly

docked (because the direction of the docking is from

the left to the right, and the last node of the context is

its right most node).

Initialization: By the transition DE (, or JK), we

will get E (, or K).

Propagation: From E (, or K), it is only possible

to perform the transition EF (, or KL). Both of these

transitions change the state of the black node to the

state d (JK also connect the white node in e to the

black node in c). After the transition EF (, or KL),

if the black node in 2 is not the ﬁrst node of the ﬁrst

strand, we will obtain either E or F, and we can only

repeat the same step. Else we can only perform the

transition FG (, or LM), and we will obtain G (, or

M).

Undocking: This stage undocks the output strand

and the stimulus by breaking the connections between

the nodes in the ﬁrst strand and the nodes in the sec-

ond strand from right to the left.

Initialization: By the transition FG (, or LM), we

will get G (, or M).

Propagation: From G, we can only perform the

transition GH (if the black node in 3 is not a sepa-

rator). From M, if the grey node in 2 is a stimulus

separator or an input separator, we can perform either

the transition MN or MO. Therefore we can perform

either the transition GH, MN or MO. If the white node

in 3 is not the last node of the ﬁrst strand, we will ob-

tain either G or M, and will repeat the same step.

Reinitialization: When the undocking stage is

terminated, we have a pair of a context and a output

strand, or a triple of a context, an output strand and a

stimulus. They must be reinitialized to be able to par-

ticipate in another reaction. This is done by the rules

from 16 to 18 for the context and by the rules from 19

to 22 for the stimulus and the output strand.

6 CONCLUSIONS

This paper has proposed a framework based on the

RNA replication mechanism to design the third level

of our formal modeling. With this framework, it is

possible to implement any reactions of a catalytic re-

action network by using nucleotide smart objects in

a generic way. In order to prove the validity of such

an implementation, we have introduced a global state

transition diagram that resumes the graph rewriting

rules. The NSOs in the same strand need not to be in

the same place. They may exist at different locations

in a physical environment and be connected through

the Internet. This modeling opens a new vista for new

application scenarios of proximity-based federation

of smart objects. With this new approach based on the

RNA replication mechanism, and with the representa-

tion of the interactions among smart objects based on

catalytic reaction networks, we will become able to

consider, to describe, and to implement new innova-

tive scenarios beyond the current scope of stereotyped

applications of ubiquitous, pervasive, and/or mobile

computing.

REFERENCES

Gelernter, D. (1985). Generative communication in linda.

ACM Transactions on Programming Languages and

Systems, 7:80–112.

Henricksen, K., Indulska, J., and Rakotonirainy, A. (2002).

Modeling context information in pervasive computing

systems. in F. Mattern and M. Naghshineh (Eds.):

Pervasive 2002, LNCS, 2414:167–180.

Kauffman, S. A. (2000). Investigations. Oxford University

Press.

Milner, R. (2001). Bigraphical reactive systems: Basic the-

ory. In Proceedings of the International Congress of

Mathematicians, pages 155–169.

Milner, R. (2004). Theories for the global ubiquitous com-

puter. Foundations of Software Science and Computa-

tion Structures, LNCS, 2987:5–11.

Picco, G. P., Murphy, A. L., and Roman, G.-C. (1999).

LIME: Linda Meets Mobility. In Proc. of ICSE’99

Int’l Conference, pages 368–377.

Tanaka, Y. (2008). Proximity-based ad hoc federation

among smart objects and its applications. Proc. of

CSTST’08 int’l.

Tanaka, Y. (2010). Proximity-based federation of smart ob-

jects: Liberating ubiquitous computing from stereo-

typed application scenarios. Proc. of KES’10 Int’l

Conference, LNCS, 6276/2010:14–30.

FORMALIZATION OF AN RNA-INSPIRED MIDDLEWARE FOR COMPLEX SMART OBJECT FEDERATION

SCENARIOS

105