Study in three phases
An Adaptive Sound Installation
C l A u d i o p A n A R i e l l o
时间
C
A
右
时间
S
乙
A
Study in three phases is an adaptive site-specific sound installation
that includes 22 solenoids placed on metallic arches that surround
visitors and react to environmental perturbations, creating a self-
regulating soundscape of metallic hits that serves to renew the visitors’
acoustic perspective. Adaptivity is a crucial aspect of the work:
Similar perturbations will not generally cause similar reactions from the
installation based on past interactions, thus allowing evolution over time
to play a key role artistically and technically. This article discusses the
author’s position on adaptivity in music interaction and composition and
reports on the technical and artistic aspects of the installation.
AppRoAChing AdApTive SySTemS
An adaptive system is one capable of modifying its inter-
nal variables (状态) as a function of its inputs in order to
fulfill a task [1,2]. Adaptive systems are increasingly used in
musical interaction because of their adaptability to environ-
mental perturbations (noise, turbulence), thus creating a
connection between audience and system. This connection
is deeper compared with nonadaptive systems: In a generi-
cally interactive installation, the system directly reacts to the
输入; 然而, implicit feedback between output and input
is relegated to a secondary, almost inconsequential, 角色. 在
adaptive systems, environment and system are structurally
coupled [3], influencing each other in a circular way. We then
consider them as living systems, as organisms. Our behavior
toward them therefore differs: We are aware of our effect on
a purely interactive installation. An adaptive system is not
completely predictable in response, and we wonder if it is
building its own knowledge of the environment and how.
所以, adaptive systems especially in the arts can subvert
power relations between humans and systems, allowing new
and unexplored interactions.
Adaptive criteria have been successfully applied in several
Claudio Panariello (作曲家, student), KTH Royal Institute of Technology,
114 28 斯德哥尔摩, 瑞典. 电子邮件: claudiop@kth.se.
ORCID: 0000-0002-1244-881X.
See www.mitpressjournals.org/toc/lmj/30 for supplemental files associated with
this issue.
artistic installations (Agostino Di Scipio’s series of works Au-
dible Ecosystemics [4,5], Michelangelo Lupone’s works [6,7],
Dario Sanfilippo’s series of works LIES [8,9], 除其他外),
highlighting the importance of further investigations into
adaptive systems in music installations.
The most prolific sources of inspiration on these artistic
and technological questions are found in systems theory
[10], cybernetics [11] and the adaptive systemic feedback ap-
proach in music [12]. I have approached adaptivity through
an ecological point of view [13], considering system and en-
vironment coupled via feedback processes. 然后, 为了
satisfy the condition that the adaptive system fulfill a task, 我
asked what “task” means for a musical adaptive system. 这
only way my system interacted with the environment was
through sound (the only sensors for detecting perturbations
were microphones), so I tackled this question by forcing the
system to preserve the coherence of its sonorous output. A
suggestion for achieving this coherence came from the idea
of a dynamical score of interactions [14]—a set of interaction
rules that dynamically change themselves over time, extract-
ing features from external perturbations to drive the system
to a new equilibrium, similar to Di Scipio’s idea of composing
the interactions of the overall work [15].
These observations led me to four interconnected con-
siderations on musical adaptive systems in developing the
installation: emerging behavior, style, system memory and
entropic degeneration.
Emerging Behavior
At a global level, adaptive systems show coherence and some
well-defined macroscopic properties deriving from local in-
teractions, properties barely predictable solely on the basis
of the rules that drive single parts of the system [16]. 这
means that adaptive systems show an emerging behavior. 这
emergence closely connects the system and its surrounding
环境. If the former changes, the latter does too; 这
change is reflected in the system, in a constant process of
adaptation.
44
LEONARDO MUSIC JOURNAL, 卷. 30, PP. 44–49, 2020
从http下载的://direct.mit.edu/lmj/article-pdf/doi/10.1162/lmj_a_01092/2004200/lmj_a_01092.pdf by guest on 07 九月 2023
土井:10.1162/LMJ_a_01092 ©2020 ISAST
根据知识共享署名发布 4.0
国际的 (抄送 4.0) 执照.
Style
Artistic Concept
The “style” of an adaptive system has been defined by Lupone
in the context of sound installations [17] as the identity of the
work—meaning, the organizational invariants of the acoustic
材料. These invariants answer to compositional criteria,
例如. temporal disposition of acoustic events, data correlation,
congruence of control data coming from installation sensors,
and relations between inputs and activated processes. 这
style is a set of rules (and exceptions) set by the composer
and reflects the composer’s artistic practice [18].
In my artistic approach, I require the musical adaptive sys-
tem to preserve the style during its evolution—I ask the adap-
tive system to be style-invariant. Thus style rules should not
change once the system is running, either through human
intervention or via adaptation; 尽管如此, the rules can
be refined through analyzing coherence of the results both
theoretically and acoustically. Style-invariance is verified at
run-time: The installation needs to live in order to adapt to
环境, thus creating a concatenation of those rules
and eventually showing the invariance.
记忆
The system may have a memory of its previous states. 这是
worth noting that the presence of a memory is not sufficient
for the system to be adaptive. Taking as an example Alvin
Lucier’s famous sound art piece I Am Sitting in a Room, 我们
can identify a memory because in any new iteration of the
process the work preserves its past history; but it is not an
adaptive work, since the system’s internal variables are fixed.
The memory of an adaptive system must be a connection
between the system and its evolutionary history rather than
a buffer where previous data is stored. The history of the
installation’s interactions with the environment will affect
its response to any new perturbation, especially if there are
similar perturbations over time.
Entropic Degeneration
Since during its adaptive process the system can evolve to-
ward disorder and noise, controls to prevent this entropic
degeneration must be provided. There are many reasons why
such a situation might occur, 例如. users introducing an ex-
cessive amount of energy into the system or the presence of
chaotic perturbations within the environment. 任何状况之下,
the system must be able to prune the inputs by interpreting
them as not relevant to its evolution.
The inSTAllATion
Study in three phases [19] is an adaptive sound installation
inspired by the architecture of the Goethe-Institut metallic
portico [20] in Rome. The artistic aim is to let visitors ex-
perience the space where the installation is exposed with a
continuously renewed acoustic perspective, creating a con-
nection between audience, installation and environment.
Eleven couples of solenoids placed on the arches of the por-
tico create a sonorous path that interacts with visitors, 在-
tempting dialogue with the environmental noises modifying
the solenoids’ behavior.
The installation has its deeper roots in my interest in and
fascination with mechanisms and patterns on one hand and
the world of microorganisms on the other. These concepts
found their natural expression in established fields such as
kinetic art [21] and physical computing [22]. These concepts
can also be traced in my instrumental works such as Piccolo
inventario degli insetti (Little inventory of insects) for en-
semble and electronics [23] or Autopsia su una marionetta
(Autopsy on a puppet) for ensemble [24]. In so saying, 这
multiple hits of the solenoids on the arches, driven by adap-
主动过程, create microrhythmical structures that can
resemble a living insect’s soundscape, thus creating another
layer of integration and communication with the cicadas and
natural sounds of the Goethe-Institut garden.
The title of the installation is a nod to the three phases
of the alchemical process, referencing its most general and
symbolic meaning of matter conversion. These phases have
both a technical and an aesthetic meaning: They divide the
work into three parts with three clear, different functions.
The first phase (分解) is symbolized by the sole-
noids’ impulses on the arches of the portico: Single sound
clicks distributed along the path allude to the dissolution
of the metallic matter of which the arches are composed.
This phase includes the conceptual migration from visual
to acoustic space: Visitors walking under the portico, 前
realizing that they affect the installation, are surrounded
by clouds of impulses. The discontinuous and pointillistic
soundscape contrasts with the continuity of the portico’s
材料. 这样, a dialectical articulation is immedi-
ately created between two domains, that of sight and that of
hearing. The second phase (purification) is symbolized by
two microphones hanging between the arches: They capture
all sound events in order to subject them to self-regulation
processes managed by the computer. 最后, the third phase
(recomposition) sends the results of the processing back to
the arches, thus closing a loop and at the same time restart-
理解它. The third phase, more than the other two, strongly
exhibits the composer’s will and the choices made to let a
global behavior emerge from the system.
General Description of the Installation
The 28.8-m-long portico of the Goethe-Institut is made up of
11 iron arches 2.6 m high and 2.7 m distant from each other.
Visitors enter through the upper garden, which leads to the
beginning or the end of the portico, or through stairs on
the main garden, leading to its center (如图. 1). Twenty-two
linear solenoids [25] were placed in couples on the arches,
one couple per arch, all controlled by two microcontrollers
(Arduino Uno) and specifically designed circuits. Two om-
nidirectional microphones were collocated under the arches
to monitor the environment. Owing to their position, 这
microphones also captured the solenoids’ hits on the arches;
因此, the whole installation, after an initial noise burst, 骗局-
tinued to resonate in a feedback loop (如图. 2).
从http下载的://direct.mit.edu/lmj/article-pdf/doi/10.1162/lmj_a_01092/2004200/lmj_a_01092.pdf by guest on 07 九月 2023
Panariello, Study in three phases 45
如图. 1. The metallic arches of the Goethe-Institut portico. (© Claudio Panariello)
如图. 2. Sketch of the general structure of Study in three phases.
(© Claudio Panariello)
Challenges of Installation
Given preexistent architecture, I considered how to integrate
the installation concept therein. I aimed to be noninvasive
and address aesthetic and practical aspects without neglect-
ing visitors’ safety. Practically, I needed to decide where and
how to place the solenoids. Aft er empirical testing, I posi-
tioned two on each arch (one solenoid per column) as shown
in Fig. 3, having found that point to be most resonant. To fi x
the solenoids to the arches, I glued them into transparent
plastic boxes into which two holes were drilled to let the axis
come out (如图. 4A). Th e boxes were fi xed to the arches with
a tape designed to resist mechanical tearing and changing
temperatures (如图. 4乙). Th is tape provided stability for the
6-day duration of the installation and meant that no other
invasive (or anti-aesthetic) support structures apart from the
transparent boxes were needed. 此外, the transpar-
ent boxes allowed the audience to see the solenoids in ac-
的, enhancing the fi rst phase (the decomposition concept)
and adding a sense of baroque beauty in the mechanisms’
placement.
Sonic Interaction
Th e portico of the Goethe-Institut is a passageway in the
middle of the garden, accessed in multiple ways. 康塞-
经常地, it was crucial to consider how visitors would enter it.
Th e garden is a small, peaceful oasis in the chaotic cen-
ter of Rome, and the portico is an almost protected, 秘密
space in that already confi ned area. 同时, 存在
made up of open arches, the portico does not exclude the rest
of the garden from sight. It is a space that naturally invites
people to walk more slowly and pay attention to the fountain
and the trees. People entering the garden from the chaotic
outside have time to release tension and slow their pace be-
fore arriving at the portico. I therefore imagined an adaptive
soundscape adjusting itself to visitors’ pace, responding to
their perturbations or trying to catch their attention upon
their approach. Visitors would then experience that confi ned
space from unusual acoustic perspectives, having to interact
with a massive yet gentle living organism. Th e installation
adapted to environmental noises through a combination of
two strategies, one based on generative algorithms and the
other on negative feedbacks [26].
Overall microphonic input was divided into 11 频率
bands through 11 bandpass fi lters; a temporal shift was added
on each band. Th e fi nal result was a sort of spectral delay.
Th e idea was to connect the bands to the 11 arches so that the
amplitude on a frequency band crossing a certain threshold
(determined while testing the installation in the portico)
would activate the corresponding arch’s solenoids. 在这个
sense there was a correspondence between arch position in
the portico and the sound events occurring under arches:
46
Panariello, Study in three phases
从http下载的://direct.mit.edu/lmj/article-pdf/doi/10.1162/lmj_a_01092/2004200/lmj_a_01092.pdf by guest on 07 九月 2023
Low frequencies activated one end of the portico and high
frequencies the other, whereas medium frequencies activated
the center. The temporal shift on each band was added to
avoid triggering of the installation by any environmental per-
turbation that, in the absence of the delay, would simultane-
ously activate a couple of adjacent arches. This setup would
activate them in a more complex way, also spontaneously
creating rhythmic structures. When the first activation of
an arch occurred, a sequence (一代) of solenoids’ hits
was created through simple L-system generative rules, 一
for each couple of solenoids. After the sixth generation, each
arch started to propagate itself to its two adjacent ones. 我
imagined the portico as a circular structure; the arches ad-
jacent to the first one were the second and the last. To avoid
overly long sequences that might result from unrelated exter-
nal events the length of the sequence was cut at a limit length.
The overall trend of the impulses sent to the solenoids
self-regulated following negative feedback rules: If high in-
put amplitude was detected, the impulses became more rar-
efied and the installation approached a more silent stage; 如果
input amplitude was low, the solenoids were activated more
迅速地. 然而, if the input tended to zero, i.e. no new
external stimuli arrived, the installation tended to be quiet,
也. 因此, an excessively rich or a stimulus-free en-
vironment could lead the installation to die. I avoided this
risk by programming an ordered and recognizable sequence
of solenoids’ impulses, disconnected from the environment,
when needed to reactivate the installation. 这样, 如果
global amplitude was detected to be below (or above) a given
threshold for a defined amount of time, the ordered sequence
was activated. While this sequence completely bypassed the
generative rules and the negative feedback loop, it did not
disable the microphones. The entire portico resonated, cre-
ating a new series of events that, recorded by microphones,
如图. 3. Ideal resonating point on the vertical arch’s pillar for
the solenoid placement. (© Claudio Panariello)
乙
A
如图. 4. (A) Box with solenoid. (乙) Box fixed to the arch. (© Claudio Panariello)
从http下载的://direct.mit.edu/lmj/article-pdf/doi/10.1162/lmj_a_01092/2004200/lmj_a_01092.pdf by guest on 07 九月 2023
Panariello, Study in three phases 47
During testing in the garden just aft er setup, the in-
stallation reached an acoustic balance with the aft er-
noon nature sounds of the garden. 尤其, 当。。。的时候
cicadas in the garden stopped singing, introducing a
sudden silence, the installation responded by increasing
its activity. Th en it found a new balance with this qui-
eter environment. During the exposure period, 存在
personally present, I informally collected observations
of visitors’ personal communications, noticing that they
found the interaction with the installation itself intrigu-
英: As they arrived in the portico, they stopped and
started to listen more carefully to the impact sounds
generated on the arches, trying to understand how and
why the installation was answering them in one way in-
stead of another. Some looked carefully around to fi nd
light or proximity sensors that could explain the activation.
When they realized that the interaction was due to nothing
other than the acoustic signal (i.e. the microphones hang-
ing from the portico), they were positively surprised. 什么时候
a noisy audience led the installation to a quiet mode, 因此
forcing the ordered security sequence to start, the installation
was briefl y inhibited in its acoustic behavior. Th e sequence
succeeded both in catching visitors’ attention and in rebal-
ancing the overall noise.
ConCluSionS
Th e installation Study in three phases is a work that adapts
its behavior to external perturbations in a constant musical
dialogue with the environment. Th e self-regulating processes
created a memory of past events, while the pruning opera-
tions managed to avoid entropic degeneration; the installa-
tion disclosed its emergent behavior showing coherence in
terms of style. Beyond the mere presentation of adaptivity
used with artistic intent, I wish also to raise questions about
which characteristics the output of a musical adaptive system
should have in terms of musical form and style. Related to
这, how does such a system behave in a musical context?
How do we behave in its presence (implicating perceived au-
tonomy and power relations)? 有趣的是, the installation’s
如图. 5. Overall structure of the algorithm conceived for Study in three phases.
(© Claudio Panariello)
reactivated the installation once the ordered sequence was
fi nished and control passed back to the previously explained
规则. Th e overall schema of the algorithm is shown in Fig. 5.
Th e negative feedback rules provided the installation an
eff ective way to avoid environmental acoustic overload. A
typical installation behavior is shown in Fig. 6A: Th e aver-
age number of activated solenoids is plotted as a function of
the input amplitude: A perturbation occurring at second 10
causes the spike between 10 和 30 where up to 9 solenoids
are activated. In that temporal window the negative feedback
tends to have a high number of solenoid activations. 什么时候
no other perturbations occur and the amplitude is low, 这
installation tends to a quiet state with low solenoid activity:
之间 40 和 100 seconds, few solenoids are activated. Th e
second environmental perturbation at second 100 causes a
new increase in the number of activated solenoids, followed
by a decrease since it is longer compared to the fi rst one
和, 最后, leads the installation to react in a negative
方式, avoiding acoustic overload. Figure 6b shows activation
of the ordered sequence involving all solenoids (seconds
220–245). Th e amplitude detection is not inhibited, 和
microphones are still monitoring the environment. 什么时候
the sequence ends, the solenoids adapt again to the input
振幅 (from second 245 在).
A
乙
如图. 6. Average number of activated solenoids (lighter line) as function of the detected overall input amplitude (thicker line). (A) A typical installation behavior.
(乙) Behavior after activation of the ordered solenoids’ sequence. (© Claudio Panariello)
48
Panariello, Study in three phases
从http下载的://direct.mit.edu/lmj/article-pdf/doi/10.1162/lmj_a_01092/2004200/lmj_a_01092.pdf by guest on 07 九月 2023
output showed coherence when presented in other venues
and environments [27]. This outcome, although acceptable
on the practical level, can be interpreted as a consequence of
too-restrictive rules that did not permit a deeper dive into
adaptivity. This leads to the open question of whether the
system should show the style of the author who designed it
or it should evolve its own style. This is almost to ask whether
the adaptive system should escape the control of its author
或不. 再次, this is a matter of hierarchies and power, 不是
only in relation to audience–installation interaction but also
to author–opera interaction. Possibly, implementation of a
musical adaptive system based on an artificial neural network
(神经网络) could indeed increase the potentiality of adaptation:
The system would be able to organize its internal structure
使用, 例如, perturbations as a training set and even-
tually showing its own style. The emergent behavior of such
a system could take unexpected interactive directions.
致谢
I thank Edoardo Bellucci, Leonardo Mammozzetti and Giuseppe Silvi,
who helped with setup of the first exhibition at the Goethe-Institut.
Study in three phases (Studio in tre fasi) was commissioned by Centro
Ricerche Musicali-CRM (罗马, 意大利) for ArteScienza 2017 与
support of SIAE Sillumina.
参考文献和注释
1 Dario Sanfilippo and Andrea Valle, “Feedback Systems: An Analyti-
cal Framework,” Computer Music Journal 37, 不. 2, 12–27 (2013).
13 Agostino Di Scipio, “Sound Is the Interface: From Interactive to
Ecosystemic Signal Processing,” Organised Sound 8, 不. 3, 269–277
(2003); Alex McLean and Roger T. 院长, The Oxford Handbook of
Algorithmic Music (Oxford Univ. 按, 2018).
14 Alessio Gabriele et al., “AD-OPERA: Music-Inspired Self-Adaptive
系统,” in Proceedings of the FSE/SDP Workshop on the Future of
Software Engineering Research, FoSER 2010 (一月 2010); Lupone
and Seno [6].
15 Di Scipio [13].
16 Sanfilippo and Valle [1].
17 The definition can be found in Lupone’s documentation of his instal-
2 Alessio Gabriele, “LAdOp1: Definizione e computazione degli adat-
tamenti in opere d’arte intermediali adattive,” in Proceedings of the
XX Colloquio di Informatica Musicale (2014).
lation Gioco Delle Risonanze.
18 Gabriele [2,14].
3 Humberto Maturana and Francisco Varela, Autopoiesis: The Realiza-
tion of the Living (多德雷赫特, 荷兰: Reidel, 1980).
4 Agostino Di Scipio, Audible Ecosystemics n.2 (Feedback Study), 分数
(2003).
5 Agostino Di Scipio, Polveri Sonore. Una prospettiva ecosistemica della
composizione (La Camera Verde, 2014).
6 Michelangelo Lupone and Lorenzo Seno, “Gran Cassa and the Adap-
tive Instrument Feed-Drum,” in International Symposium on Com-
puter Music Modeling and Retrieval (柏林: 施普林格, 2005).
7 Michelangelo Lupone et al., “Struttura, creazione, interazione, evo-
luzione: la ricerca al CRM per la realizzazione di Forme Immate-
riali di Michelangelo Lupone,” in Proceedings of the XXI Colloquio
di Informatica Musicale (2016).
8 Dario Sanfilippo, “LIES (distance/incidence) 1.0: A Human-Machine
Interaction Performance,” in Proceedings of the XIX Colloquio di
Informatica Musicale (2012).
9 Dario Sanfilippo, “Tuning Perturbation into Emergent Sound, 和
Sound into Perturbation,” in Interference Journal of Audio Culture
3 (2013): www.interferencejournal.org/turning-perturbation-into
-emergent-sound (访问过 3 可能 2020).
10 Maturana and Varela [3]; Heinz Von Foerster, Understanding Under-
常设: Essays on Cybernetics and Cognition (纽约: 施普林格,
2003); Edgar Morin, “Restricted Complexity, General Complexity,”
in J.-L. Le Moigne and E. Morin, 编辑。, Intelligence de la Complexité:
Épistémologie et Pragmatique (Cerisy-La-Salle, 法国, 26 六月 2005)
(Carlos Gershenson, trans., 2006).
11 William Ross Ashby, An Introduction to Cybernetics (伦敦: Chap-
男人 & 大厅, 1971); Norbert Wiener, Introduzione alla Cibernetica.
L’uso umano degli esseri umani (Turin: Bollati Boringhieri, 1966).
12 Agostino Di Scipio, “Émergence du Son, Son d’Emergence: Essai
d’épistémologie expérimentale par un compositeur,” Intellectica 48–
49 (2008) PP. 221–249; Dario Sanfilippo, “Time-Variant Infrastruc-
tures and Dynamical Adaptivity for Higher Degrees of Complexity
in Autonomous Music Feedback Systems: The Order from Noise
(2017) 项目,” Musica/Tecnologia 12, 不. 1, 119–129 (2017).
19 At the first exhibition, the installation was presented with the Italian
title Studio in tre fasi. It premiered at ArteScienza 2017, “Il Giardino
dei Suoni Segreti” (罗马, 17–22 July 2017). See video at www.vimeo
.com/327037884 (访问过 3 可能 2020).
20 A portico is an arcade consisting of a colonnade and a roof structure
above it. In this article I refer to it using Italian terminology.
21 See Zimoun for a recent example: www.zimoun.net (访问过 3 可能
2020).
22 Dan O’Sullivan and Tom Igoe, Physical Computing: Sensing and
Controlling the Physical World with Computers (Course Technology
按, 2004); Andrea Valle, “Making Acoustic Computer Music: 这
Rumentarium Project,” Organised Sound 18, 不. 3, 242–254 (2013).
23 Claudio Panariello, “Piccolo Inventario degli Insetti,” d.a.t. [divulga-
zioneaudiotestuale] 1 (2017).
24 Both Autopsia su una marionetta and Piccolo inventario degli insetti
can be heard at www.soundcloud.com/claudiopanariello (访问过
3 可能 2020).
25 A linear solenoid is a common solenoid with a metallic axis through
its coil, providing a linear actuation. When an electric current flows
through the coil, the axis is retrieved into it; an optional spring helps
the axis return to resting position when no current is running.
26 The code is written in the SuperCollider programming environment:
https://supercollider.github.io. See www.github.com/claudiopanari
ello/studio-in-tre-fasi (访问过 3 可能 2020).
27 例如, at the Festival “Stanze di Musica,” at which the instal-
lation adapted to the spaces of Casa della Cultura e delle Arti in
Camerota (萨勒诺, 意大利, 十二月 2019).
稿件收到 4 可能 2020.
Claudio Panariello is a composer and sound researcher
currently pursuing a Ph.D. in Sound and Music Computing at
the KTH Royal Institute of Technology in Stockholm, 瑞典.
Working with electronics, machines and renowned ensembles
of contemporary music, he explores algorithmic and adaptive
processes applied to music.
从http下载的://direct.mit.edu/lmj/article-pdf/doi/10.1162/lmj_a_01092/2004200/lmj_a_01092.pdf by guest on 07 九月 2023
Panariello, Study in three phases 49
