Available online at www.sciencedirect.

com

Electronic Notes in Theoretical Computer Science 322 (2016) 169–180

www.elsevier.com/locate/entcs

Synchronous Robots vs Asynchronous

1
Lights-Enhanced Robots on Graphs

Mattia D’Emidio2
Gran Sasso Science Institute (GSSI), Viale Francesco Crispi, I–67100, L’Aquila, Italy.
Department of Information Engineering, Computer Science and Mathematics, University of L’Aquila,
Via Vetoio, I–67100, L’Aquila, Italy.

Daniele Frigioni3

Department of Information Engineering, Computer Science and Mathematics, University of L’Aquila,

Via Vetoio, I–67100, L’Aquila, Italy.

Alfredo Navarra4
Department of Mathematics and Computer Science, University of Perugia,
Via Vanvitelli, 1, I-06123, Perugia, Italy.

Abstract
In this paper, we consider the distributed setting of computational mobile entities, called robots, that have
to perform tasks without global coordination. Depending on the environment as well as on the robots’
capabilities, tasks might be accomplished or not.
In particular, we focus on the well-known scenario where the robots reside on the nodes of a graph and
operate in Look-Compute-Move cycles. In one cycle, a robot perceives the current configuration in terms
of robots positions (Look), decides whether to move toward some edge of the graph (Compute), and in the
positive case it performs an instantaneous move along the computed edge (Move).
We then compare two basic models: in the first model robots are fully synchronous, while in the second one
robots are asynchronous and lights-enhanced, that is, each robot is equipped with a constant number of
lights visible to all other robots. The question whether one model is more powerful than the other in terms
of computable tasks has been considered in [Das et al., Int.’l Conf. on Distributed Computing Systems,
2012 ] but for robots moving on the Euclidean plane rather than on a graph.
We provide two different tasks, and show that on graphs one task can be solved in the fully synchronous
model but not in the asynchronous lights-enhanced model, while for the other task the converse holds.
Hence we can assert that the fully synchronous model and the asynchronous lights-enhanced model are
incomparable on graphs. This opens challenging directions in order to understand which peculiarities make
the models so different.
Keywords: Distributed algorithm, Synchronicity, Mobile Robots, Luminous Robots

http://dx.doi.org/10.1016/j.entcs.2016.03.012
1571-0661/© 2016 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
170 M. D’Emidio et al. / Electronic Notes in Theoretical Computer Science 322 (2016) 169–180

1 Introduction
In the last few years a considerably large amount of research, in the area of dis-
tributed computing, has been devoted to the study of models and algorithmic ap-
proaches for the so-called robot-based computing systems, due to their importance
in a wide range of real-world applications. In this kind of systems a set of mobile
entities, usually referred as robots, have either to perform tasks and/or to achieve
goals under a variety of assumptions that depend on the considered scenario [6,14].
Particular efforts have been dedicated to models where robots are autonomous,
i.e. they act without a central control, and operate in a Look-Compute-Move (LCM)
operational mode (see [1,2,3,19,22] and references therein). In such a model, which
has become a prominent one in the area of distributed algorithms for robot-based
computing systems, robots operate in so-called LCM cycles. During each cycle,
a robot acquires a snapshot of the surrounding environment (Look phase), then
executes an appropriate algorithm, which is the same for all robots, by using the
obtained snapshot as input (Compute phase), and finally moves toward a desired
destination, if any (Move phase).
Several modeling assumptions have been also considered that can affect the com-
putational power of the robots. In particular, in some cases, robots may have distinct
identities, i.e. each robot is associated with a different identifier that can be used
during the Compute phase. If robots are without identifiers, they are said to be
anonymous. In some other cases, robots may have a finite but persistent memory
device whose content is preserved from one LCM cycle to the next; robots are said
to be oblivious if they do not, which means that they start each cycle without any
information about what happened in the past.
In this research area, many different problems and tasks have been studied:
robots might be asked to gather in certain specific locations [20] (also known as the
Gathering problem), to form a desired geometric pattern [21] (also known as the
Pattern Formation problem), or to explore an unknown area [17] (also known as the
Exploration problem). In addition, several different settings have been investigated.
Robots can move on a Euclidean plane [16], or they are constrained to move on a
given input graph, which can either be known in advance [9] or not [4]. Robots can
be able to communicate, e.g. by means of tokens as in [15], or not [18]. Finally,
there might exist or not an objective function to be optimized, associated with the
problem (see [3,11,12,13] and references there in). For instance, one may ask for
the minimum number robots, the minimum number of steps performed by all the
robots, or the minimization of the maximum number of steps performed by a single
robot, to achieve a certain goal.
Look-Compute-Move cycles might be subject to different temporal constraints

1 The work has been partially supported by the Italian Ministry of Education, University, and Research
(MIUR) under national research projects: PRIN 2010N5K7EB “ARS TechnoMedia – Algoritmica per le
Reti Sociali Tecno-Mediate” and by the National Group for Scientific Computation (GNCS-INdAM).
2 Email: [email protected]
3 Email: [email protected]
4 Email: [email protected]
 M. D’Emidio et al. / Electronic Notes in Theoretical Computer Science 322 (2016) 169–180 171

dictated by the considered schedule. The most common models in the literature are
the following:
Fully-synchronous (FSYN C) [8]: The activation phase of all robots can be logi-
cally divided into global rounds, where in each round the robots are all activated,
obtain the same snapshot of the environment, compute and perform their move.
Notice that, this assumption is computationally equivalent to a fully synchronized
system in which robots are activated simultaneously and all operations happen
instantaneously.
Semi-synchronous (SSYN C) [10]: It coincides with the FSYN C model with the
only difference that not all robots are necessarily activated in each round.
Asynchronous (ASYN C) [5]: The robots are activated independently, and the
duration of each Compute, Move and inactivity phase is finite but unpredictable.
As a result, robots do not have a common notion of time. Moreover, they can be
seen while moving, and computations can be made based on obsolete information
about positions.
Recently, further models have been introduced by Das at al. in [6], extending the
aforementioned ones. In detail, given a model M ∈ {FSYN C, SSYN C, ASYN C},
the authors define models Mk , where each robots operating in the M model is
equipped with a light that is visible to itself and to the other robots during the
Look phase. The light associated with a robot can assume k different colors and
can be updated by a robot during its Compute phase. The light is assumed to be
persistent, i.e. despite robots can be oblivious, their lights are not automatically
reset at the end of a cycle. Thus, robots’ lights can be considered as a form of
external persistent memory. Light-enhanced robots are usually referred as luminous
robots. Note that, depending on the considered scenario, a robot might have visibility
of the lights of either all other robots of just of a subset of them. Another model,
introduced in [6], considers oblivious robots which are “slightly” empowered with
the additional capability of remembering a constant number of previously acquired
snapshots. More precisely, for some integer constant j > 0, each robot is allowed to
store in an internal memory at most j snapshots from previous Look phases. When
robots, operating in a certain model M, are endowed with the ability of remembering
k snapshots, we refer to the modified model as Mk and to the robots as partially
oblivious robots.

1.1 Motivation of the paper

Our work is inspired by the paper of Das et al. [6], which concerned with the problem
of comparing the computational capabilities of robots operating in LCM model and
moving in the Euclidean plane, under different levels of synchronization, with those
of both luminous robots and partially oblivious robots. In details, the authors of [6]
provided a series of results that prove relationships of dominance between classic
models and variations of them that consider the capability of either using lights or
snapshots, or a combination of them.
In Figure 1, we summarize the main contributions of [6] and show such relation-
172 M. D’Emidio et al. / Electronic Notes in Theoretical Computer Science 322 (2016) 169–180

ASYN C O(1) ≡ SSYN C O(1)

O(1) ?
ASYN C O(1)

FSYN C SSYN C
Figure 1. Relationships among models.

ships of dominance. We denote by a direct arrow between two models, say A and B
the fact that every problem solvable in A is also solvable in B. Moreover, we denote
by a triple bar symbol the fact that every problem solvable in A is also solvable in
B and viceversa. In particular, the authors of [6] showed that:
• ASYN C O(1) ≡ SSYN C O(1) → SSYN C: robots that operate in ASYN C and are
endowed with a constant number of lights (i.e. that operate in ASYN C O(1) ) are
computationally as powerful as robots that operate in SSYN C and are endowed
with a constant number of lights (i.e. that operate in SSYN C O(1) ); moreover,
both ASYN C O(1) and SSYN C O(1) models are computationally more powerful
than SSYN C.
O(1)
• ASYN C O(1) → ASYN C O(1) : robots that operate in ASYN C and are endowed
with both a constant number of lights and the ability to remember a constant
O(1)
number of snapshots (i.e. that operate in ASYN C O(1) ) are more powerful than
O(1)
robots that operates in ASYN C O(1) . Moreover, ASYN C O(1) is more powerful
than FSYN C.
In Figure 1 we also report relationships of dominance that were already well known
from the literature, like e.g. FSYN C → SSYN C. Note that, in [6] the relationship
between FSYN C and ASYN C O(1) has not been established.

1.2 Contribution of the paper

In this paper, we try to answer to the problem of finding the relationship, in terms of
computable tasks, that exists between FSYN C and ASYN C O(1) . To this regard, we
provide two different distributed tasks of robots moving within a graph environment,
and not in the Euclidean plane as in [6], and show that one task can be solved in
FSYN C but not in ASYN C O(1) , while for the other the viceversa holds. Hence, we
provide a proof that FSYN C and ASYN C O(1) are incomparable on graphs. This
opens challenging directions in order to understand which peculiarities make the
models so different.

1.3 Structure of the paper

In Section 2, we describe the distributed system in which the robots operate. In
Section 3, we define a distributed problem which can be solved in FSYN C but not
 M. D’Emidio et al. / Electronic Notes in Theoretical Computer Science 322 (2016) 169–180 173

in ASYN C O(1) . In Section 4, we define a distributed problem which can be solved

in ASYN C O(1) but not in FSYN C. Finally, in Section 5, we conclude the paper
and provide some possible future research directions.

2 Preliminaries

In this paper, as already mentioned, we consider a system composed of a team

of mobile entities, called robots, that operates in Look-Compute-Move cycles. In
particular, each robot is modeled as an independent computational unit with its
own local memory and capable of performing local computations. The robots are
placed in a spatial environment, which can be assumed to be an undirected graph
G = (V, E), i.e. robots are placed on the nodes of the graph. Therefore, each
robot has its own local perception of the surrounding environment, which means
it can detect a graph isomorphic to G and understand whether a node is occupied
by a robot or not. Each robot is equipped with sensing capabilities that return a
snapshot of the relative positions of all other robots with respect to its location on
the perceived graph.
In the remaining of the paper, we assume that robots are anonymous and iden-
tical, i.e. they are indistinguishable by their appearance, and execute the same
algorithm. They are oblivious, i.e. they have no memory. Moreover, we consider
that robots act without a central control, i.e. they are assumed to be autonomous
and are not able to directly communicate information (e.g. by a wireless interface)
with other robots, i.e. they are silent. Each robot is endowed with motor capa-
bilities and can freely move on G. However, the movement along one edge of G is
considered instantaneous, so that each time a robot perceives the snapshot of the
current configuration, it sees all other robots always on the nodes of G. We will
specify different assumptions when it is not clear from the context.
At any point in time, a robot is either active or inactive. When active, a robot
executes a Look-Compute-Move (LCM) cycle performing the following three opera-
tions in sequence, each of them associated with a different state:
Look: The robot observes the world by activating its sensors, which return a snap-
shot of the positions of all robots with respect to its own perception.
Compute: The robot executes its algorithm, using the data sensed in the Look
phase as input. The result of this phase is a target (destination) node.
Move: The robot moves toward the computed target: if the destination is the
current position, the robot simply stays still, i.e. it performs what we call a null
movement.
When inactive, instead, a robot is idle. All robots are initially inactive. The
amount of time to complete a full LCM cycle is assumed to be finite but unpre-
dictable.
174 M. D’Emidio et al. / Electronic Notes in Theoretical Computer Science 322 (2016) 169–180

3 ASYN C O(1) is not more powerful than FSYN C

In this section, we define a distributed problem, namely the Pattern Sequence
Chasing (P SC) problem, and show that it can be solved in FSYN C but not in
ASYN C O(1) . The P SC problem can be thought as a variation of the Series of
Patterns Formation problem, defined in [7], where: i) no pattern is repeated in the
provided sequence; ii) robots move on a graph instead of a Euclidean plane.
The problem is defined as follows: as an input, we consider a set of k robots
disposed on a non-anonymous undirected and complete graph G (i.e. each node of
G is associated with a unique identifier), and an array A of dimension n whose entries
are pairwise distinct patterns. A pattern P is an ordered array of k distinct identifiers
of nodes of G that represents a placement of the k robots on the nodes identified
by P . Initially, the robots are disposed according to pattern A[0]. Subsequently,
the k robots have to move in order to form pattern A[1], and so forth, moving from
pattern A[i] to pattern A[(i + 1) mod n], see for instance Figure 2.
Note that, since the robots are identical and anonymous, assuming that the
nodes of G can be uniquely identified is crucial to the feasibility of the problem. In
fact, in anonymous graphs, if two robots occupy the same node, then the adversary
can make both robots always move concurrently. In this way they will never reach
different destinations, hence the problem becomes unsolvable.
For the sake of simplicity, we consider that A has finite size n. Therefore, without
loss of generality, robots are asked to move from a pattern A[i mod n] to a pattern
A[(i + 1) mod n].
In the remaining of the paper, we denote by A[i][j] the j-th entry of pattern A[i].

Pattern Series Chasing (P SC) Problem

Input: A non-anonymous undirected and complete graph G. An array A of n
patterns, each involving k nodes of G, such that A[i] = A[j], for every
0 ≤ i = j < n. A set of k robots forming A[0] in G.
Solution: A distributed algorithm that ensures robots to form pattern A[(i + 1)
mod n] after A[i mod n], i ∈ N.

In what follows, we provide an algorithm, namely Algorithm PatternChaser

(see Algorithm 1), which solves P SC in FSYN C by exploiting the uniqueness of
the current pattern in the sequence defined by the entries of the array A. The
algorithm works as follows: during each Look phase, each robot perceives a snapshot
s of the positions (i.e., ids of nodes of the graph) of other robots. Such snapshot
corresponds to a certain pattern A[i] in the sequence which can be easily found by
scanning A (initially the snapshot is clearly equal to A[0]). Then, given A[i], each
robot compares the id of the node of the graph he is located at, say v, with those
in A[i] by sequentially scanning it. Eventually, each robot finds the index j such
that A[i][j] = v, and can perform its move from position A[i][j] to position A[(i + 1)
mod n][j].
 M. D’Emidio et al. / Electronic Notes in Theoretical Computer Science 322 (2016) 169–180 175

Algorithm 1 Algorithm for solving P SC under FSYN C.

1: procedure PatternChaser(Snapshot s)  Snapshot s acquired during the
Look phase
2: i := 0;
3: while A[i] = s do
4: i + +;
5: end while
6: Let myp be my position within pattern A[i];
7: j := 0;
8: while myp = A[i][j] do
9: j + +;
10: end while
11: The new position is A[(i + 1) mod n][j];
12: end procedure

Theorem 3.1 Algorithm PatternChaser correctly solves P SC problem in

FSYN C.

Proof As robots are fully synchronous, every time a new Look-Compute-Move cycle
starts, all robots perceive the same snapshot s (at the beginning, s is equal to A[0]).
Therefore, during each Compute phase, each robot can: i) uniquely find the current
pattern in the sequence A, say A[i]; ii) find the index j within the array A[i] such
that A[i][j] matches its position. Since all such entries are distinct, each of them
uniquely defines the next position at which robots should be placed in the subsequent
step, namely the j-th entry of array A[(i + 1) mod n], i.e. A[(i + 1) mod n][j]. The
movement is accomplished by letting all robots move simultaneously to reach the
desired placement. 2

On the contrary, in general, operating under ASYN C O(1) does not permit to
solve P SC, unless the number of provided lights is O(n), which is not necessarily a
constant number.

Theorem 3.2 Problem P SC cannot be solved in ASYN C O(1) .

Proof To prove the statement, we have to show that there is at least one case where
a constant number of lights does not suffice to robots to infer the current pattern
A[i], and therefore to move to the correct subsequent pattern A[(i + 1) mod n].
In particular, we prove that any algorithm requires robots to be equipped with a
number of lights that is not constant with respect to the size n of the input array A,
thus making the problem unsolvable in ASYN C O(1) . In fact, if the number of lights
permits to encode the current index i, then the robots may synchronize themselves
to obtain A[(i + 1) mod n] from A[i].
To prove the statement true for the general case of k robots, we make use of an
example that can be easily extended to higher dimensions. As input we consider
a set of k = 3 robots, a non-anonymous complete graph G = (V, E) with V =
{v1 , · · · , v7 }, and a sequence of patterns A (part of which is shown in Figure 2). We
176 M. D’Emidio et al. / Electronic Notes in Theoretical Computer Science 322 (2016) 169–180

A[0] A[1] A[2] A[3] A[4]

 v1   v4   v1   v4   v1 
v2 → v5 → v5 → v2 → v2
v3 v6 v3 v3 v6
A[5] A[6] A[7] A[8] ···
 v4   v1   v4   v1 
v5 → v5 → v2 → v5 → ···

v3 v6 v6 v7

Figure 2. Example of an instance that is not solvable in ASYN C O(1) by k = 3 luminous robots.

now show that such an instance of P SC cannot be solved by asynchronous robots

equipped with a constant number of lights.
Let us assume that, at step i = 0, the 3 robots are correctly placed on the input
graph G, i.e. each one of them is located exactly at one of the nodes of the required
pattern, say A[0] = {v1 , v2 , v3 }, as shown in Figure 2.
From A[0] robots have to move to A[1]. Lights do not permit to make all robots
move concurrently, and by hypothesis lights do not suffice to encode index i. Three
cases may arise concerning the movement of robots from A[0]: (i) all robots move;
(ii) only one robot moves; (iii) only two robots move. In case (i), configuration A[1]
is correctly reached. In case (ii), configurations A[2], A[3] or A[4] can be reached.
In case (iii), configurations A[5], A[6] or A[7] can be reached.
It follows that, if robots cannot encode index i, then in some cases they can
“get confused” about which pattern must be realized. Lights might be used to infer
whether all robots have moved or not. However, for instance, from pattern A[6] it
is not sufficient for the third robot to know that the other two robots have moved
in order to understand whether the previous pattern was A[1] or A[7].
A possible approach to overcome this limitation is that of using moves to inter-
mediate patterns, that could help to infer the target pattern. However, also this
approach is not effective. In fact, since there are no restrictions on the size n of A,
we may consider A composed of all possible patterns, and therefore each interme-
diate pattern coincides with an entry of A. Again, robots can “get confused” about
which pattern must be realized. 2

As shown in the proof of the above theorem, problem P SC cannot be solved

in ASYN C O(1) because there is no mean to encode the current configuration by
exploiting a constant number of lights nor some strategic positioning of the robots.
However, if we assume robots empowered also with a constant persistent memory
able to store one snapshot, then the index of the current pattern is deducible from
the snapshot, and hence the following corollary can be stated.
O(1)
Corollary 3.3 Problem P SC can be solved in ASYN C O(1) .
 M. D’Emidio et al. / Electronic Notes in Theoretical Computer Science 322 (2016) 169–180 177

Similarly to the above corollary, instead of adding persistent memory, if we in-

crease the number of lights to be enough to encode the index of the current pattern,
then the following corollary can be stated.
Corollary 3.4 Problem P SC can be solved in ASYN C O(log n) .

4 FSYN C is not more powerful than ASYN C O(1)

In this section, we define a distributed problem, namely the Forth and Back (F B)
problem, and show that it can be solved in ASYN C O(1) but not in FSYN C.

Forth and Back (F B) Problem

Input: Two robots, named r1 and r2 , and a graph G, which is a path P of at
least six nodes. The two robots reside at distinct nodes of P , which are
neither the endpoints of P nor neighbors of the endpoints of P . Let d
be the initial distance, i.e number of edges, between the two robots.
Solution: A distributed algorithm that ensures robots to alternate their distance
between d + 2 and d.

The F B problem has some similarities with our previous P SC problem, but the
sequence of patterns (distances) is not defined according to the ids of the nodes
but on the initial configuration where robots always start lying in distinct internal
nodes. Moreover, notice that, in this case, the graph is assumed to be anonymous.
Theorem 4.1 Problem F B cannot be solved in FSYN C.

Proof Whenever a Look-Compute-Move cycle starts, robots cannot infer whether

the current distance must be enlarged or restricted. In fact, this information requires
to know either how many times robots have executed their running cycles, or at least
whether such a number is odd or even. These information simply cannot be deduced
from the snapshot acquired during the Look phase. 2

In what follows we provide an algorithm, namely Algorithm ForthBack (see

Algorithm 2), which assumes that robots are empowered with two different lights,
each of them assuming two different colors. Algorithm 2 makes use of colors white,
black for light 1, and red, blue for light 2, with the following meanings:
• white: indicates that the robot is ready to move for the next step;
• black: indicates that the robot has moved and it is ready for the next step;
• red: indicates an even step where the previous distance must be increased;
• blue: indicates an odd step where the previous distance must be decreased.
At the beginning, both robots start with the lights set to white and red. As
we will see, algorithm ForthBack ensures that, whenever both robots have the
first light set to white, they also have the second light concordant, which means
they are currently synchronized. That is, Algorithm 2 solves F B in ASYN C O(1) by
178 M. D’Emidio et al. / Electronic Notes in Theoretical Computer Science 322 (2016) 169–180

exploiting the parity encoding of the current step by means of lights. Of course, the
same result can be obtained by making use of only one light assuming four different
colors.

Algorithm 2 Algorithm for solving F B under ASYN C O(1) .

1: procedure ForthBack(Colors l1  and l )  Lights of the other robot acquired
2
during the Look phase
2: Let l1 and l2 the current colors of my two lights;
3: Let d be the number of edges from the other robot;
4: if l1 = white ∧ l2 = red ∧ l2 = red then
5: l1 := black;
6: Let v be the neighbor at distance d + 1 from the other robot;
7: The new position is v;
8: Exit;
9: end if
10: if l1 = white ∧ l2 = blue ∧ l2 = blue then
11: l1 := black;
12: Let v be the neighbor at distance d − 1 from the other robot;
13: The new position is v;
14: Exit;
15: end if
16: if l1 = black ∧ l2 = red ∧ ((l1 = white ∧ l2 = blue) ∨ (l1 = black ∧

l2 = red)) then
17: l1 := white;
18: l2 := blue;
19: Exit;
20: end if
21: if l1 = black ∧ l2 = blue ∧ ((l1 = white ∧ l2 = red) ∨ (l1 = black ∧

l2 = blue)) then
22: l1 := white;
23: l2 := red;
24: Exit;
25: end if
26: end procedure

Theorem 4.2 Algorithm ForthBack correctly solves FB problem in

ASYN C O(1) .

Proof Let us denote as l1 , l2 (l1 , l2 , respectively) the lights of r1 (r2 , respectively).
If l1 is white, then r1 is ready to accomplish a movement, which either must
enlarge the distance of the previous placement or restrict it. This depends on l2 ,
whether it is red or blue, respectively. Moreover, in order to understand the right
movement, r1 considers also the two lights of r2 , namely l1 and l2 . If they are the
same, i.e. l1 = l1 and l2 = l2 , then this is the case where robots are currently
synchronized. If l1 = l1 =black and l2 is discordant with respect to l2 , then r1
 M. D’Emidio et al. / Electronic Notes in Theoretical Computer Science 322 (2016) 169–180 179

waits for r2 in order to get synchronized with it. If l1 is black and l2 = l2 , then
r1 concludes that r2 has already moved, so it is now its turn. The movement to
be done is evaluated according to the current lights of r2 . In details, if l2 is red,
then r1 must move away from r2 of one edge. In fact, as we noticed, r2 has already
terminated its movement. Hence, it is waiting for r1 . If d was the distance between
the two robots before both robots moved, then d + 2 must define the final placement
of the two robots in the current step. Since they both contribute of one edge, when
only one robot moves, the current distance d is d = d + 1, Therefore, by moving of
one edge, the two robots will reach distance d + 2. Similarly, If l2 is blue, then r1
must move closer to r2 of one edge.
Finally, if l1 is black, then r1 has terminated a step and either it must start
the next one, or it has to wait for r2 in order to temporarily get synchronous with
it. This is realized by considering the lights of r2 . If l1 is white and l2 = l2 , then
r1 changes l1 to white and l2 to l2 . If l1 is white and l2 = l2 , then r1 has to wait
for r2 to move. If l1 is black and l2 = l2 , then r1 changes l1 to white and makes
l2 discordant with l2 . The case where l1 is black and l2 = l2 cannot occur. This
realizes the synchronization among robots. 2

5 Conclusion
We have shown the incomparability of two important models studied in robot-based
computing systems. In particular, we have considered FSYN C, where all robots
execute their algorithm synchronously, and ASYN C O(1) , where robots are asyn-
chronous but they are empowered with a constant number of lights. The proof has
been pursued by providing two different distributed tasks of robots moving within
a graph environment, and show that one task can be solved in FSYN C but not in
ASYN C O(1) , while for the other the viceversa holds.
The case where robots move on the Euclidean plane remains open. Actually,
arguments of Theorems 3.1, 4.1 and 4.2 can be extended quite easily in order to
hold in the Euclidean case as well. The main challenge remains Theorem 3.2 where
the difficulty mainly resides in proving that stigmergy strategies cannot be effective.
That is, in the Euclidean plane, robots may compute intermediate positions rather
than straightly move toward the position dictated by the current pattern. This
opens challenging directions in order to understand which peculiarities make the
FSYN C and the ASYN C O(1) models so different.
Another interesting direction is that of revisiting the whole picture of Figure 1
when robots move on a graph environment.

References
[1] Chatterjee, A., S. G. Chaudhuri and K. Mukhopadhyaya, Gathering asynchronous swarm robots under
nonuniform limited visibility, in: 11th International Conference on Distributed Computing and Internet
Technology (ICDCIT), Lecture Notes in Computer Science 8956 (2015), pp. 174–180.

[2] Cicerone, S., G. Di Stefano and A. Navarra, Minimum-traveled-distance gathering of oblivious robots
over given meeting points, in: 10th International Symposium on Algorithms and Experiments for Sensor
180 M. D’Emidio et al. / Electronic Notes in Theoretical Computer Science 322 (2016) 169–180

Systems, Wireless Networks and Distributed Robotics (ALGOSENSORS), Lecture Notes in Computer
Science 8847 (2014), pp. 57–72.

[3] Cicerone, S., G. Di Stefano and A. Navarra, Minmax-distance gathering on given meeting points, in: 9th
International Conference on Algorithms and Complexity (CIAC), Lecture Notes in Computer Science
9079 (2015), pp. 127–139.

[4] Czyzowicz, J., A. Kosowski and A. Pelc, How to meet when you forget: log-space rendezvous in arbitrary
graphs, Distributed Computing 25 (2012), pp. 165–178.

[5] D’Angelo, G., G. Di Stefano and A. Navarra, Gathering on rings under the look-compute-move model,
Distributed Computing 27 (2014), pp. 255–285.

[6] Das, S., P. Flocchini, G. Prencipe, N. Santoro and M. Yamashita, The power of lights: Synchronizing
asynchronous robots using visible bits, in: 32nd IEEE International Conference on Distributed
Computing Systems (ICDCS), 2012, pp. 506–515.

[7] Das, S., P. Flocchini, N. Santoro and M. Yamashita, On the computational power of oblivious robots:
Forming a series of geometric patterns, in: 29th ACM SIGACT-SIGOPS Symposium on Principles of
Distributed Computing (PODC) (2010), pp. 267–276.

[8] Degener, B., B. Kempkes, T. Langner, F. Meyer auf der Heide, P. Pietrzyk and R. Wattenhofer, A tight
runtime bound for synchronous gathering of autonomous robots with limited visibility, in: 23rd ACM
Symp. on Parallelism in algorithms and architectures (SPAA), 2011, pp. 139–148.

[9] D’Emidio, M., D. Frigioni and A. Navarra, Exploring and making safe dangerous networks using mobile
entities, in: 12th International Conference on Ad Hoc Networks and Wireless (ADHOC-NOW), Lecture
Notes in Computer Science 7960 (2013), pp. 136–147.

[10] Devismes, S., F. Petit and S. Tixeuil, Optimal probabilistic ring exploration by semi-synchronous
oblivious robots, in: 16th International Colloquium on Structural Information and Communication
Complexity (SIROCCO), Lecture Notes in Computer Science 5869, 2009, pp. 195–208.

[11] Di Stefano, G. and A. Navarra, Gathering of oblivious robots on infinite grids with minimum traveled
distance, Information and Computation. To appear.
[12] Di Stefano, G. and A. Navarra, Optimal gathering of oblivious robots in anonymous graphs, in:
Proceedings of the 20th International Colloquium on Structural Information and Communication
Complexity (SIROCCO), Lecture Notes in Computer Science 8179, 2013, pp. 213–224.

[13] Di Stefano, G. and A. Navarra, Optimal gathering on infinite grids, in: Proc. of the 16th International
Symposium on Stabilization, Safety, and Security of Distributed Systems (SSS), Lecture Notes in
Computer Science 8756, 2014, pp. 211–225.
[14] Dieudonné, Y., A. Pelc and D. Peleg, Gathering despite mischief, ACM Transactions on Algorithms
11 (2014), p. 1.

[15] Dobrev, S., P. Flocchini, R. Královic and N. Santoro, Exploring an unknown dangerous graph using
tokens, Theoretical Computer Science 472 (2013), pp. 28–45.

[16] Flocchini, P., D. Ilcinkas, A. Pelc and N. Santoro, Remembering without memory: Tree exploration by
asynchronous oblivious robots, Theoretical Computer Science 411 (2010), pp. 1583–1598.

[17] Flocchini, P., D. Ilcinkas, A. Pelc and N. Santoro, Computing without communicating: Ring exploration
by asynchronous oblivious robots, Algorithmica 65 (2013), pp. 562–583.

[18] Flocchini, P., D. Ilcinkas, A. Pelc and N. Santoro, Computing without communicating: Ring exploration
by asynchronous oblivious robots, Algorithmica 65 (2013), pp. 562–583.

[19] Flocchini, P., G. Prencipe and N. Santoro, Distributed computing by oblivious mobile robots, Synthesis
Lectures on Distributed Computing Theory 3 (2012), pp. 1–185.

[20] Flocchini, P., G. Prencipe, N. Santoro and P. Widmayer, Gathering of asynchronous robots with limited
visibility, Theoretical Computer Science 337 (2005), pp. 147–168.

[21] Fujinaga, N., Y. Yamauchi, S. Kijima and M. Yamashita, Asynchronous pattern formation by
anonymous oblivious mobile robots, in: Distributed Computing, Lecture Notes in Computer Science
7611, Springer, 2012 pp. 312–325.
[22] Kamei, S., A. Lamani, F. Ooshita and S. Tixeuil, Gathering an even number of robots in an odd ring
without global multiplicity detection, in: Mathematical Foundations of Computer Science (MFCS),
Springer, 2012 pp. 542–553.