The Fourteenth International Symposium on Artificial Life and Robotics 2009 (AROB 14th ’09),

B-Con Plaza, Beppu, Oita, Japan, February 5 - 7, 2009

A Design of Algorithms for Real-Time Generation of Linear-Recursive Sequences on

Cellular Automata

Naoki Kamikawa Hiroshi Umeo

Graduate School of Engineering Graduate School of Engineering
Univ. of Osaka Electro-Communication Univ. of Osaka Electro-Communication
Neyagawa-shi, Hatsu-cho 18-8, Neyagawa-shi, Hatsu-cho 18-8,
Osaka, 572-8530, Japan Osaka, 572-8530, Japan

Abstract connected to its left- and right-neighbor cells via a commu-

nication link. Each cell can know state of its left- and right-
A model of cellular automata (CA) is considered to be
neighbor cells via communication link. One distinguished
a non-linear model of complex systems in which an infinite
leftmost cell C1 , the communication cell, is connected to
one-dimensional array of finite state machines (cells) up-
the outside world. A cellular automaton (abbreviated by
dates itself in a synchronous manner according to a uniform
CA) consists of an infinite array of finite state automata
local rule. We study a sequence generation problem on the
A = (Q, δ, F ), where
CA and propose several state-efficient real-time sequence
generation algorithms for non-regular sequences. We show 1. Q is a finite set of internal states.
that Fibonacci sequence can be generated in real-time by 2. δ is a function defining the next state of a cell, such
a CA with 5 states. We also study infinite linear-recursive that δ: Q × Q × Q → Q, where δ(p, q, r) = s, p, q, r,
sequences, such as tribonacci, tetranacci and pell sequences s ∈ Q has the following meaning: We assume that at
generated on the CA. step t the cell Ci is in state q, the left cell Ci−1 is in
state p and the right cell Ci+1 is in state r. Then, at
1 Introduction the next step t+1, Ci assumes state s. The leftmost
cell C1 is connected to the outside world. The outside
A model of cellular automata (CA) was devised orig- world is expressed by *. A quiescent state q ∈ Q has
inally for studying self-reproduction by John von Neu- a property such that δ(q, q, q) = q.
mann. It is now studied in many fields such as complex 3. F ⊆ Q is a special subset of Q. The set F is used to
systems. We study a sequence generation problem on the specify a designated state of C1 in the definition of
CA. Arisawa[1], Fischer[2], Korec[3] and Kamikawa and sequence generation.
Umeo[5], [6] studied generation of a class of natural num-
bers on CA. In this paper, we show that Fibonacci se- We now define the sequence generation problem on
quence can be generated in real-time by a CA with 5 states. CA. Let M be a CA and {tn | n = 1, 2, 3, ...} be an infinite
We also study infinite linear-recursive sequences, such as monotonically increasing positive integer sequence defined
tribonacci sequence, tetranacci sequence and pell sequence natural numbers, such that tn ≥ n for any n ≥ 1. We then
generated on the CA. We show a design of algorithm for have a semi-infinite array of cells, as shown in Figure 1,
real-time generation of linear-recursive sequences on CA. and all cells, except for C1 , are in the quiescent state at
time t = 0. The communication cell C1 assumes a special
state r in Q for initiation of the sequence generator. We
2 Real-time sequence generation prob- say that M generates a sequence {tn | n = 1, 2, 3, ...} in k
lem on CA linear-time if and only if the leftmost end cell of M falls
into a special state in F ⊆ Q at time t = k · tn , where k is
A cellular automaton consists of an infinite array of a positive integer. We call M a real-time generator when
identical finite state automata, each located at a positive k = 1.
integer point (See Figure 1).
C1 C2 C3 C4 Cn-1 Cn 3 Generation Algorithms of Linear-
Recursive Sequences
In this section, we propose generation algorithm of
Figure 1: Cellular automaton. linear-recursive sequences. First, we show a design of algo-
rithm for real-time generation of linear-recursive sequences
Each automaton is referred to as a cell. A cell at point on CA. Next, we show that Fibonacci sequence can be gen-
i is denoted by Ci , where i ≥ 1. Each Ci , except for C1 , is erated on a CA with 5 states.

©ISAROB 2009 281

The Fourteenth International Symposium on Artificial Life and Robotics 2009 (AROB 14th ’09),
B-Con Plaza, Beppu, Oita, Japan, February 5 - 7, 2009

3.1 A Design of Algorithm speed 1-cell/3-step in cell space. Therefore, state A1, A2
and A3 which propagate in cell space is called a-wave. A
Let m be any natural number, such that m ≥ 1. Let
sequence generation algorithm is designed geometrical by
k be natural number given, such that k ≥ 1, k < m. Let
using wave which propagates in cell space. The a-wave
b1 , b2 , . . . , bk , c1 , c2 , . . . , ck be natural number given,
propagates in the right direction at 1/3 speed. Figure 3
such that b1 , b2 , · · · , bk ≥ 1, c1 < c2 < · · · < ck . Let am
shows a number of snapshots of the cell configuration at the
be kth order linear-recursive sequences, such that am =
propagation of the a-wave. The a-wave moves to cell C2 at
b1 · am−1 + b2 · am−2 + · · · + bk · am−k , a1 = c1 , a2 = c2 ,
time t = 1. Afterwards, the a-wave moves by one cell every
. . ., ak = ck . We show a design of algorithm for real-time
3 steps. When we assume Pa (t) to be a function whitch
generation of sequence am on CA.
shows the position of the a-wave at time t, it is approved
that Pa (t) = ⌈ 3t ⌉ + 1. At time t = am−1 , cell C1 is in a
3.2 First Order Linear-Recursive Se- state included F and the b-wave is generated on C1 . The
quences b-wave propagates in the right direction at 1/1 speed, and
the b-wave reaches the a-wave. When the a-wave collides
We propose the generation algorithm for k = 1. It is with the b-wave, the a-wave keeps propagating, the b-wave
approved that am = b1 · am−1 , a1 = c1 . However, it is is eliminated, the d-wave is generated and the e-wave is
limited to b1 ≥ 2. Because all terms take c1 for b1 = 1, generated. When am−1 is an even number, the b-wave
and am is not an infinite monotonically increasing positive collides with the second state of 3 states to compose the a-
integer sequence. Figure 2 shows a time-space diagram for wave and the e-wave is generated (See Figure 3). Let r be
generation of the term am , when the term am−1 is an even natural number. When the cell which collides the a-wave
number. with the b-wave is assumed to be cell Cr , it is approved that
a
C1 C am-1 +1 r = Pa (am−1 + r − 1) = m−1 2
+ 1. Therefore, the e-wave
a
2 is generated on cell C am−1 +1 at time t = am−1 + m−1 .
t=0 2
2
The e-wave keeps staying on cell C am−1 +1 . The d-wave
2
࡮࡮࡮࡮࡮

a-wave
1 2 3 4 5 6 7 8 9 propagates in the left direction at 1/1 speed, and the d-
0
1
A3 Q Q Q Q Q Q Q Q
W1A1 Q Q Q Q Q Q Q
wave reaches the leftmost cell C1 at time t = 2 · am−1 .
1/3
2 W2A2 Q Q Q Q Q Q Q When the d-wave collides with the leftmost cell C1 , the
3 W3A3 Q Q Q Q Q Q Q
t = am-1 4 W4 R A1 Q Q Q Q Q Q d-wave is eliminated and the c-wave is generated. The c-
5 W5 R A2 Q Q Q Q Q Q wave propagates in the right direction at 1/1 speed, and
1/1 6 G R A3 Q Q Q Q Q Q
b-wave
7 Q B R A1 Q Q Q Q Q the c-wave reaches the e-wave at time t = 2 . . . am−1 +
8 Q R B A2 Q Q Q Q Q am−1
9 Q R R E1 Q Q Q Q Q 2
. When the c-wave collides with the e-wave, the b-
1/1 10 Q R D E2A1 Q Q Q Q wave is eliminated and the d-wave is generated. The d-
t = 2 am-1 d-wave 11 Q D R E2A2 Q Q Q Q
12 t1 R R E2A3 Q Q Q Q wave propagates in the left direction at 1/1 speed. The
1/1
13
14
t2 C R E2 R A1 Q Q Q
t2 R C E2 R A2 Q Q Q d-wave reaches cell C1 at time t = 3 · am−1 . Therefore,
c-wave 15 t2 R R D R A3 Q Q Q Time where the b-, c- and d-waves reciprocate between
b1-1 16 t2 R D R R R A1 Q Q
e-wave
17 t2 D R R R R A2 Q Q the leftmost cell C1 and the e-wave is am−1 steps. The
18 G R R R R R A3 Q Q
b-, c- and d-waves reciprocate b1 − 1 times between the
t = 3 am-1 19 Q B R R R R R A1 Q
20 Q R B R R R R A2 Q leftmost cell C1 and the e-wave (See Figure 2). When the
d-wave of times b1 − 1 reaches the cell C1 , a state of C1
࡮࡮࡮࡮࡮

Figure 3: A changes to a state included F at time t = b1 · am−1 .

con guration Figure 4 shows a time-space diagram for generation
of generation of the term am , when the term am−1 is an odd num-
t = am of the term am ber. When am−1 is an odd number, the o-wave is gen-
= b1 am-1 (when the term erated by the collision of the a-wave and the b-wave. The
am 1 is an even b-wave collides with the third state of 3 states to com-
Figure 2: Time-space dia- number). pose the a-wave and the o-wave is generated (See Fig-
gram for generation of the ure 5). When the cell which collides the a-wave with
the b-wave is assumed to be cell Cr , it is approved that
term am (when the term a
r = Pa (am−1 +r −1) = ⌊ m−1 2
⌋+1. Because am−1 is an odd
am 1 is an even number). a −1
number, it is approved that r = m−1 2
+1. Therefore, the
o-wave is generated on cell C am−1 −1 . he c-wave propa-
Generation of the term am is described in terms of 6 2
+1
waves: a-wave, b-wave, c-wave, d-wave, e-wave and o-wave. gates in the right direction at 1/1 speed, and the c-wave
a −1
The a-wave is generated on C1 at time t = 0. Figure reaches the o-wave at time t = am−1 + m−1 2
. When the
3 shows a number of snapshots of the cell configuration c-wave collides with the o-wave, the b-wave is eliminated.
at the propagation of the a-wave shown in Figure 2. In The d-wave is generated after 1 step. The d-wave propa-
Figure 3, state A1, A2 and A3 advance toward the right at gates in the left direction at 1/1 speed. The d-wave reaches

©ISAROB 2009 282

The Fourteenth International Symposium on Artificial Life and Robotics 2009 (AROB 14th ’09),
B-Con Plaza, Beppu, Oita, Japan, February 5 - 7, 2009

a −1 a −1
cell C1 at time t = am−1 + m−12
+1+ m−1 2
= 2·am−1 . the e- or o-wave is generated on cell C⌊ am−2 ⌋+1 . At time
2
Therefore, Time where the b-, c- and d-waves reciprocate t = am−1 , cell C1 is in a state included F and the b-wave
between the leftmost cell C1 and the o-wave is am−1 steps. is generated on C1 . The b-wave propagates in the right
The b-, c- and d-waves reciprocate b1 − 1 times between direction at 1/1 speed, and the b-wave reaches the e- or o-
the leftmost cell C1 and the o-wave (See Figure 4). When wave generated on cell C⌊ am−2 ⌋+1 . When the e- or o-wave
the d-wave of times b1 − 1 reaches the cell C1 , a state of 2
collides with the b-wave, the b-wave keeps propagating and
C1 changes to a state included F at time t = b1 · am−1 .
the d-wave is generated. The b-wave propagates, and the
C1 C am-1 -1 +1 b-wave reaches the a-wave. When the a-wave collides with
2
t=0 the b-wave, the b-wave is eliminated and the e- or o-wave
is generated on cell C⌊ am−1 ⌋+1 . The d-wave propagates in
࡮࡮࡮࡮࡮

a-wave 2
1 2 3 4 5 6 7 8 9
the left direction at 1/1 speed, and the d-wave reaches the
0 A3 Q Q Q Q Q Q Q Q
1/3 1 W1A1 Q Q Q Q Q Q Q leftmost cell C1 at time t = am−1 +am−2 . The b-, c- and d-
2 W2A2 Q Q Q Q Q Q Q
3 W3A3 Q Q Q Q Q Q Q
waves reciprocate b2 times between the leftmost cell C1 and
t = am-1 4 W4 R A1 Q Q Q Q Q Q the e- or o-wave generated on cell C⌊ am−2 ⌋+1 . When the e-
5 W5 R A2 Q Q Q Q Q Q 2
1/1 6 W6 R A3 Q Q Q Q Q Q or o-wave generated on cell C⌊ am−2 ⌋+1 collides with the b-
7 G R R A1 Q Q Q Q Q 2
b-wave
1 step 8 Q B R A2 Q Q Q Q Q or c-wave b2 times, the e- or o-wave is eliminated. At the
next, The c- and d-waves reciprocate b1 − 1 times between
9 Q R B A3 Q Q Q Q Q
10 Q R R O1A1 Q Q Q Q
1/1
11
12
Q R R O2A2 Q Q Q Q
Q R D O3A3 Q Q Q Q
the leftmost cell C1 and the e- or o-wave generated on cell
t = 2 am-1 d-wave 13 Q D R O3 R A1 Q Q Q C⌊ am−1 ⌋+1 . When The b-, c- and d-waves reciprocate b2
14 t1 R R O3 R A2 Q Q Q 2
1/1 15 t2 C R O3 R A3 Q Q Q times between cell C1 and the e- or o-wave generated on
c-wave 16
17
t2 R C O3 R R A1 Q Q
t2 R R O4 R R A2 Q Q cell C⌊ am−2 ⌋+1 and reciprocate b1 −1 times between cell C1
1 step
b1-1 2
18 t2 R R D R R A3 Q Q
19 t2 R D R R R R A1 Q
and the e- or o-wave generated on cell C⌊ am−1 ⌋+1 , a state
2
20 t2 D R R R R R A2 Q
21 G R R R R R R A3 Q
of the leftmost cell C1 changes a state included F . The
t = 3 am-1 o-wave 22 Q B R R R R R R A1 first some terms and some e- and o-waves are generated in
an internal state. For example, Figure 8 shows generation
࡮࡮࡮࡮࡮

1 step
Figure 5: A of pell sequence (am = 2 · am−1 + am−2 , a1 = 1, a2 = 2).
con guration
of generation 1 2 3 4 5 6 7 8 9 10 11 12
C1C am-2 C am-1
0 A3 Q Q Q Q Q Q Q Q Q Q Q +1 +1
t = am of the term am 1 W1A1 Q Q Q Q Q Q Q Q Q Q t=0 2 2

= b1 am-1 2 W2A2 Q Q Q Q Q Q Q Q Q Q a-wave

(when the term 3 G A3 Q Q Q Q Q Q Q Q Q Q ࡮࡮࡮࡮࡮
1/3
am 1 is an odd 4
5
Q O1A1 Q Q Q Q Q Q Q Q Q
Q O2A2 Q Q Q Q Q Q Q Q Q
Figure 4: Time-space dia- number). 6 t1 O3A3 Q Q Q Q Q Q Q Q Q
7 t2 O4 R A1 Q Q Q Q Q Q Q Q
gram for generation of the 8 t2 D R A2 Q Q Q Q Q Q Q Q
t = am-1
9 G R R A3 Q Q Q Q Q Q Q Q
term am (when the term 10 Q B R R A1 Q Q Q Q Q Q Q 1/1
11 Q R B R A2 Q Q Q Q Q Q Q b-wave
am 1 is an odd number). 12 Q R R B A3 Q Q Q Q Q Q Q
13 Q R R R O1A1 Q Q Q Q Q Q t = am-1+am-2
14 Q R R R O2A2 Q Q Q Q Q Q b2
The first term a1 = c1 is generated in an internal state. 15 Q R R D O3A3 Q Q Q Q Q Q
16 Q R D R O3 R A1 Q Q Q Q Q
A state of the leftmost cell C1 is changed a state included 17 Q D R R O3 R A2 Q Q Q Q Q e-wave
or
F at time t = c1 by counting the c1 − 1 step by using an 18 t1 R R R O3 R A3 Q Q Q Q Q o-wave
19 t2 C R R O3 R R A1 Q Q Q Q t = am-1+b2 am-2
internal state. And the b-wave is generated. Therefore, 1st 20 t2 R C R O3 R R A2 Q Q Q Q
1/1
21 t2 R R C O3 R R A3 Q Q Q Q
Order Linear-Recursive Sequences can be generated on CA 22 t2 R R R O4 R R R A1 Q Q Q c-wave
23 t2 R R R D R R R A2 Q Q Q
in real-time. In Figure 6, we show a number of snapshots 24 t2 R R D R R R R A3 Q Q Q 1/1
of the configuration for b1 = 3 and c1 = 3 from t = 0 to 25
26
t2 R D R R R R R R A1 Q Q
t2 D R R R R R R R A2 Q Q t = 2 am-1+b2 am-2 d-wave
b1-1
29. 27 G R R R R R R R R A3 Q Q
28 Q B R R R R R R R R A1 Q
29 Q R B R R R R R R R A2 Q

3.3 Second Order Linear-Recursive Se-

quences Figure 6: A t = am
con guration of = b1 am-1+b2 am-2
Next, we consider the case of k = 2. We propose the real-time genera-
generation algorithm of second order linear-recursive se- tion of a rst order Figure 7: Time-space
quences which enhance the algorithm described in section linear-recursive diagram for generation
3.2. It is approved that am = b1 · am−1 + b2 · am−2 , a1 = c1 , sequence (am = of a second order linear-
a2 = c2 . Figure 7 shows a time-space diagram for gen-
3
am 1 , a1 = 3). recursive sequence.
eration of a second order linear-recursive sequence. The
a-wave is generated on C1 at time t = 0. We assume that

©ISAROB 2009 283

The Fourteenth International Symposium on Artificial Life and Robotics 2009 (AROB 14th ’09),
B-Con Plaza, Beppu, Oita, Japan, February 5 - 7, 2009

1 2 3 4 5 6 7 8 9 10 11 12
C am-k +1
C am-k+1 +1
0 A3 Q Q Q Q Q Q Q Q Q Q Q 2 2
1 G A1 Q Q Q Q Q Q Q Q Q Q
2 G A2 Q Q Q Q Q Q Q Q Q Q
3 W3E3 Q Q Q Q Q Q Q Q Q Q C1 C am-2 +1
C am-1 +1
4 W4E3A1 Q Q Q Q Q Q Q Q Q 2 2
5 G E3A2 Q Q Q Q Q Q Q Q Q t=0

࡮࡮
6 Q E4A3 Q Q Q Q Q Q Q Q Q
7
8
t1 R O1A1 Q Q Q Q Q Q Q Q
t2 C O1A2 Q Q Q Q Q Q Q Q
t = am-1 e-wave
or
9 t2 R O2A3 Q Q Q Q Q Q Q Q t = am-1+am-k o-wave bk

࡮࡮
10 t2 R O3 R A1 Q Q Q Q Q Q Q ࡮࡮࡮
11 t2 D O4 R A2 Q Q Q Q Q Q Q t = am-1+bk am-k
12 G R O4 R A3 Q Q Q Q Q Q Q
13 Q C O4 R R A1 Q Q Q Q Q Q a-wave
14 Q R O5 R R A2 Q Q Q Q Q Q
15 Q R D B R A3 Q Q Q Q Q Q b-wave
16 Q D R R B R A1 Q Q Q Q Q t = am-1+bk am-k+am-k+1 1/3 bk

࡮࡮࡮࡮࡮
17 t1 R R R R B A2 Q Q Q Q Q 1/1
18 t2 C R R R R E1 Q Q Q Q Q
19 t2 R C R R R E1A1 Q Q Q Q
20 t2 R R C R R E1A2 Q Q Q Q
21 t2 R R R C R E1A3 Q Q Q Q t = am-1+bk am-k+bk-1 am-k+1

࡮࡮
࡮࡮࡮
22 t2 R R R R C E1 R A1 Q Q Q
23 t2 R R R R R E2 R A2 Q Q Q t = am-1+bk am-k+bk-1 am-k+1+
24 t2 R R R R D E3 R A3 Q Q Q
25 t2 R R R D R E3 R R A1 Q Q
+b3 am-
26 t2 R R D R R E3 R R A2 Q Q
27 t2 R D R R R E3 R R A3 Q Q
28 t2 D R R R R E3 R R R A1 Q
29 G R R R R R E3 R R R A2 Q
30 Q C R R R R E3 R R R A3 Q

Figure 8: A con guration of real-time generation of t = am-1+bk am-k+bk-1 am-k+1+ b2

pell sequence (am = 2
am 1 + am 2 , a1 = 1, a2 = 2). +b3 am- +am-2
࡮࡮࡮࡮࡮

3.4 kth Order Linear-Recursive Se-

quences
t = am-1+bk am-k+bk-1 am-k+1+
Next, we generalize the generation algorithm described +b3 am- +b2 am-2
in section 3.3, and propose the generation algorithm of
kth Order Linear-Recursive Sequences. Figure 9 shows a
c-wave
time-space diagram for generation of a kth order linear-
1/1
recursive sequence. The a-wave is generated on C1 at
time t = 0. We assume that the e- or o-waves are gen-
erated on cell C⌊ am−k ⌋+1 , C⌊ am−k+1 ⌋+1 , · · ·, C⌊ am−3 ⌋+1
2 2 2
and C⌊ am−2 ⌋+1 . At time t = am−1 , cell C1 is in a state
2 d-wave
included F and the b-wave is generated on C1 . The b-wave
1/1
propagates in the right direction at 1/1 speed, and the b- t = 2 am-1+bk am-k+bk-1 am-k+1+ b1-1
wave reaches the e- or o-wave generated on cell C⌊ am−k ⌋+1 . +b3 am- +b2 am-2
2
When the e- or o-wave collides with the b-wave, the b-wave
keeps propagating and the d-wave is generated. The b-
࡮࡮࡮࡮࡮

wave propagates by passing the e- or o-waves generated on

cell C⌊ am−k+1 ⌋+1 , · · ·, C⌊ am−3 ⌋+1 , and the b-wave reaches
2 2
the a-wave. When the a-wave collides with the b-wave,
the b-wave is eliminated and the e- or o-wave is generated
on cell C⌊ am−1 ⌋+1 . The d-wave propagates by passing the
2
e- or o-waves in the left direction at 1/1 speed, and the d-
wave reaches the leftmost cell C1 at time t = am−1 +am−k .
t = am= b1 am-1+b2 am-2+
The b-, c- and d-waves reciprocate bk times between the bk-1 am-k+1+bk am-k
leftmost cell C1 and the e- or o-wave generated on cell
C⌊ am−k ⌋+1 . When the e- or o-wave generated on cell
2 Figure 9: Time-space diagram for generation of a kth
C⌊ am−k ⌋+1 collides with the b- or c-wave bk times, the
2 order linear-recursive sequence.
e- or o-wave is eliminated. At the next, The c- and d-

©ISAROB 2009 284

The Fourteenth International Symposium on Artificial Life and Robotics 2009 (AROB 14th ’09),
B-Con Plaza, Beppu, Oita, Japan, February 5 - 7, 2009

waves reciprocate bk−1 times between the leftmost cell C1 algorithm of Fibonacci sequence on CA is not exist. A
and the e- or o-wave generated on cell C⌊ am−k+1 ⌋+1 , re- consists of an infinite array of finite state automata A =
(Q, δ, F ), where Q = {Q, A, B, C, D}, F = {A}. We show
2
ciprocate bk−2 times between the leftmost cell C1 and the
e- or o-wave generated on cell C⌊ am−k+2 ⌋+1 , · · ·, recip- that Fibonacci sequence can be generated in real-time by
2 a CA with 5 states that is given in Table 1. In Figure 11,
rocate b2 times between the leftmost cell C1 and the e-
we show a time-space diagram for real-time generation of
or o-wave generated on cell C⌊ am−2 ⌋+1 and reciprocate
2 Fibonacci sequence.
b1 − 1 times between the leftmost cell C1 and the e- or
o-wave generated on cell C⌊ am−1 ⌋+1 . At time t = am =
2
b1 · am−1 + b2 · am−2 + · · · + bk · am−k , a state of the leftmost Table 1: Transition rules for real-time generation of
cell C1 changes a state included F . The first some terms Fibonacci sequence.
and some e- and o-waves are generated in an internal state.
For example, Figure 10 shows generation of tetranacci se- Right State Right State Right State

quence (am = am−1 + am−2 + am−3 + am−4 , a1 = 1, a2 =

Q Q A B C D
A Q A B C D
B Q A B C D
2, a3 = 4, a4 = 8). Q Q Q Q C Q Q Q Q Q Q B D
A B Q D A A C A C C C D

Left State

Left State

Left State
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 B Q B A B C
0 A3 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
1 G A1 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q C Q D C D C C B
2 G A2 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
3 Q t1 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q D B D A B A A A D C B B B B
4 G t1 A1 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
5 Q t2 A2 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q A Q Q A A Q
6 Q t3 E1 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
7 Q t4 E1A1 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
8 G t2 E1A2 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Right State Right State
9
10
C2 t1 E1A3 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
t2 E11E2 R A1 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
C Q A B C D
D Q A B C D
11 C3 E7 E1C1A2 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
12 Q E8 E1 R E1 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q D D Q D Q C C
13 Q E7 E3 R E1A1 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
14 Q E9 E4 R E1A2 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q A D Q C C A C A A A
15 G E7 E4 R E1A3 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
Left State

Left State
16 t1 E10E4 R E1 R A1 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q B D C C B D C
17 C2 R E5 R E1 R A2 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
18 t2 C2 E4C1 E1 R A3 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q C D C C C C C D D
19 t2 R E6 R E2 R R A1 Q Q Q Q Q Q Q Q Q Q Q Q Q Q
20 t2 D2 E7 R E1 B R A2 Q Q Q Q Q Q Q Q Q Q Q Q Q Q D A D D D D B D D D D
21 C3 R E7 R E1 R B A3 Q Q Q Q Q Q Q Q Q Q Q Q Q Q
22 Q C3 E7 R E1 R R O1A1 Q Q Q Q Q Q Q Q Q Q Q Q Q A
23 Q R E8 R E1 R R O1A2 Q Q Q Q Q Q Q Q Q Q Q Q Q
24 Q R E7C3 E1 R R O1A3 Q Q Q Q Q Q Q Q Q Q Q Q Q
25 Q R E7 R E3 R R O1 R A1 Q Q Q Q Q Q Q Q Q Q Q Q
26 Q R E7D3 E4 R R O1 R A2 Q Q Q Q Q Q Q Q Q Q Q Q
27 Q R E9 R E4 R R O1 R A3 Q Q Q Q Q Q Q Q Q Q Q Q
28 Q D3 E7 R E4 R R O1 R R A1 Q Q Q Q Q Q Q Q Q Q Q
29
30
G R E7 R E4 R R O1 R R A2 Q Q Q Q Q Q Q Q Q Q Q
t1 C1 E7 R E4 R R O1 R R A3 Q Q Q Q Q Q Q Q Q Q Q
Real-time generation of Fibonacci sequence is described
31
32
t1 RE10 R E4 R R O1 R R R A1 Q Q Q Q Q Q Q Q Q Q
t1 D1 R C1 E4 R R O1 R R R A2 Q Q Q Q Q Q Q Q Q Q in terms of 5 waves: a-wave, b-wave, d-wave, we-wave, wo-
33 C2 R R R E5 R R O1 R R R A3 Q Q Q Q Q Q Q Q Q Q
34 t2 C2 R R E4C1 R O1 R R R R A1 Q Q Q Q Q Q Q Q Q wave. Fibonacci sequence is 2nd linear-recursive sequence.
35 t2 R C2 R E4 R C1O1 R R R R A2 Q Q Q Q Q Q Q Q Q
36 t2 R R C2 E4 R R O2 R R R R A3 Q Q Q Q Q Q Q Q Q Therefore, When each the we- or wo-wave collide with the
37 t2 R R R E6 R R O1 B R R R R A1 Q Q Q Q Q Q Q Q
38
39
t2 R R D2 E7 R R O1 R B R R R A2 Q Q Q Q Q Q Q Q
t2 R D2 R E7 R R O1 R R B R R A3 Q Q Q Q Q Q Q Q
b- or c-wave 1 time, the we- or wo-wave is eliminated. The
40
41
t2 D2 R R E7 R R O1 R R R B R R A1 Q Q Q Q Q Q Q
C3 R R R E7 R R O1 R R R R B R A2 Q Q Q Q Q Q Q
initial configuration is the leftmost cell C1 takes state D
42
43
Q C3 R R E7 R R O1 R R R R R B A3 Q Q Q Q Q Q Q
Q R C3 R E7 R R O1 R R R R R R O1A1 Q Q Q Q Q Q
and other cells take a quiescent state Q. At time t = 0, the
44 Q R R C3 E7 R R O1 R R R R R R O1A2 Q Q Q Q Q Q
45 Q R R R E8 R R O1 R R R R R R O1A3 Q Q Q Q Q Q a-wave is generated on the leftmost cell C1 . The a-wave
46 Q R R R E7C3 R O1 R R R R R R O1 R A1 Q Q Q Q Q
47 Q R R R E7 R C3O1 R R R R R R O1 R A2 Q Q Q Q Q propagates in the right direction at 1/3 speed. State B,
48 Q R R R E7 R R O3 R R R R R R O1 R A3 Q Q Q Q Q
49
50
Q R R R E7 R R O4 R R R R R R O1 R R A1 Q Q Q Q
Q R R R E7 R D3O5 R R R R R R O1 R R A2 Q Q Q Q
State C and State D are used for the propagation of the
51
52
Q R R R E7D3 R O5 R R R R R R O1 R R A3 Q Q Q Q
Q R R R E9 R R O5 R R R R R R O1 R R R A1 Q Q Q
a-wave. At time t = 0, δ(C, Q, Q) = B are applied in cell C2 .
53
54
Q R R D3 E7 R R O5 R R R R R R O1 R R R A2 Q Q Q
Q R D3 R E7 R R O5 R R R R R R O1 R R R A3 Q Q Q
At the next step, a state of C2 changes to B. At time t = 1,
55 Q D3 R R E7 R R O5 R R R R R R O1 R R R R A1 Q Q
56 G R R R E7 R R O5 R R R R R R O1 R R R R A2 Q Q δ(A, B, Q) = C are applied in cell C2 . At the next step, a
57 t1 C1 R R E7 R R O5 R R R R R R O1 R R R R A3 Q Q
58 t1 R C1 R E7 R R O5 R R R R R R O1 R R R R R A1 Q state of C2 changes to C. At time t = 2, δ(A, C, Q) = D are
59 t1 R R C1 E7 R R O5 R R R R R R O1 R R R R R A2 Q
60 t1 R R RE10 R R O5 R R R R R R O1 R R R R R A3 Q applied in cell C2 . At the next step, a state of C2 changes
to D. At time t = 3, δ(D, Q, Q) = B are applied in cell C3 .
Figure 10: A con guration of real-time generation of At the next step, a state of C3 changes to B. The a-wave
tetranacci sequence (am = am 1 + am 2 + am 3 + propagates by repeating the application of these transition
am 4 , a1 = 1, a2 = 2, a3 = 4, a4 = 8). rules. State A and State D are used for the propagation
of the b-wave. State C is used for the propagation of the
d-wave. The first 5 terms, the we-wave generated on cell
C5 and the d-wave to generate the 6th term are generated
with an internal state. At time t = 11, the d-wave is gener-
3.5 Fibonacci Sequence ated on cell C3 . The d-wave propagates in the left direction
In this section, we show real-time generation algorithm at 1/1 speed, and reaches the leftmost cell C1 . When the
of Fibonacci sequence. In a past research, Arisawa showed d-wave reaches the leftmost cell C1 , δ(*, Q, C) = A are ap-
that Fibonacci sequence can be generated in 2 linear-time plied in cell C1 . At time t = 13, a state of C1 changes to
by a CA with 9 states. However, real-time generation A, and the b-wave is generated. The b-wave propagates in

©ISAROB 2009 285

The Fourteenth International Symposium on Artificial Life and Robotics 2009 (AROB 14th ’09),
B-Con Plaza, Beppu, Oita, Japan, February 5 - 7, 2009

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

t=0
C1 C5 C7 C11 0 D Q Q Q Q Q Q Q Q Q Q Q Q Q Q
1 A B Q Q Q Q Q Q Q Q Q Q Q Q Q
2 A C Q Q Q Q Q Q Q Q Q Q Q Q Q
3 A D Q Q Q Q Q Q Q Q Q Q Q Q Q
4 Q C B Q Q Q Q Q Q Q Q Q Q Q Q
5 A Q C Q Q Q Q Q Q Q Q Q Q Q Q
6 Q D D Q Q Q Q Q Q Q Q Q Q Q Q
7 Q C B B Q Q Q Q Q Q Q Q Q Q Q
8 A Q B C Q Q Q Q Q Q Q Q Q Q Q
9 Q Q B D Q Q Q Q Q Q Q Q Q Q Q
10 Q Q D D B Q Q Q Q Q Q Q Q Q Q
11 Q Q C D C Q Q Q Q Q Q Q Q Q Q
12 Q C D D A Q Q Q Q Q Q Q Q Q Q
t = 13 13 A D D D A B Q Q Q Q Q Q Q Q Q
we-wave 14 Q A D D A C Q Q Q Q Q Q Q Q Q
15 Q Q A D A D Q Q Q Q Q Q Q Q Q
16 Q Q Q A A C B Q Q Q Q Q Q Q Q
17 Q Q Q Q C Q C Q Q Q Q Q Q Q Q
a-wave 18 Q Q Q C D D D Q Q Q Q Q Q Q Q
b-wave 19 Q Q C D D D B B Q Q Q Q Q Q Q
1/1 20 Q C D D D D B C Q Q Q Q Q Q Q
t = 21 1/3 21 A D D D D D B D Q Q Q Q Q Q Q
22 Q A D D D D B D B Q Q Q Q Q Q
wo-wave 23 Q Q A D D D B C C Q Q Q Q Q Q
24 Q Q Q A D D B C D Q Q Q Q Q Q
b-wave 25 Q Q Q Q A D B C C B Q Q Q Q Q
26 Q Q Q Q Q A B C C C Q Q Q Q Q
1/1
27 Q Q Q Q Q Q D C C D Q Q Q Q Q
28 Q Q Q Q Q Q C D C C B Q Q Q Q
29 Q Q Q Q Q C D D D C C Q Q Q Q
30 Q Q Q Q C D D D D D D Q Q Q Q
d-wave
31 Q Q Q C D D D D D D B B Q Q Q
1/1 32 Q Q C D D D D D D D B C Q Q Q
t = 34 33
34
Q
A
C
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
B
B
D
D
Q
B
Q
Q
Q
Q
35 Q A D D D D D D D D B C C Q Q
36 Q Q A D D D D D D D B C D Q Q

Figure 11: Time-space diagram for real-time genera-

tion of Fibonacci sequence. Figure 12: A con guration of real-time generation of
Fibonacci sequence.

the left direction at 1/1 speed, and reaches the we-wave References
generated on cell C5 at time t = 17. When the we-wave
collides with the b-wave, the we-wave is eliminated, the d- [1] M. Arisawa; On the generation of integer series by the
wave is generated and the b-wave keeps propagating. The one-dimensional iterative arrays of finite state machines (in
Japanese), The Trans. of IECE, 71/8 Vol. 54-C, No.8, pp.759-
d-wave reaches the a-wave. When the a-wave collides with 766, (1971).
the b-wave, the b-wave is eliminated, the wo-wave is gener- [2] P. C. Fischer; Generation of primes by a one-dimensional real-
ated. The d-wave generated on cell C5 reaches the leftmost time iterative array. J. of ACM, Vol.12, No.3, pp.388-394,
cell C1 at time t = 21. When the d-wave reaches the left- (1965).
most cell C1 , a state of C1 changes to A, and the b-wave [3] I. Korec; Real-time generation of primes by a one-dimensional
cellular automaton with 9 states. Proc. of MCU ’98, pp.101-
is generated. Therefore, Fibonacci sequence can be gen- 116, (1998).
erated by repeating the propagation of 5 waves. We have
[4] N. Kamikawa and H. Umeo; Some Algorithms for Real-Time
implemented the algorithm on a computer. We have tested Generation of Non-Regular Sequences on One-Bit Inter-Cell-
the validity of the rule set from t = 0 to t = 20000 steps. Communication Cellular Automata, SICE Annual Conference
2007, pp.953-958, (2007).
We obtain the following theorem. In Figure 12, we show a
number of snapshots of the configuration from t = 0 to 36. [5] N. Kamikawa and H. Umeo : Some state-efficient algorithms
for real-time generation of non-regular sequences on cellular au-
tomata. The Thirteenth International Symposium on Artificial
Life and Robotics (AROB 13th ’08), pp.47-50, (2008).
[6] N. Kamikawa and H. Umeo : A Note on Sequence Generation
Power of Two-States Cellular Automata. SICE Annual Confer-
4 Conclusions ence 2008, pp.3315-3320, (2008).

We have studied a sequence generation problem on CA.

A design of algorithm for real-time generation of linear-
recursive sequences on CA has been given. We have shown
that Fibonacci sequence can be generated in real-time by
a CA with 5 states. A future study in sequence generation
problem on CA is to compare sequence generation power
of CA and sequence generation power of 1-bit inter-cell-
communication CA.

©ISAROB 2009 286