-The ambient intelligence scenario depicts electronic environments that are sensitive and responsive to the presence of people. The paper deals with a particular kind of system whose aim is to enhance the everyday experience of people moving inside the related physical environment according to the narrative description of a designer's desiderata. In this kind of situation computer simulation represents a useful way to envision the behaviour of the responsive environments modeled and implemented, without actually bringing them into existence in the real world, in order to evaluate their adherence to the designer's specification. This paper describes the simulation of an adaptive illumination facility, a physical environment endowed with a set of sensors that perceive the presence of humans (or other entities such as dogs, bicycles, cars) and interact with a set of actuators (lights) that coordinate their state to adapt the ambient illumination to the presence and behaviours of its users. The simulation system is used to compare two different selforganization models managing the adaptive illumination system.
The ambient intelligence scenario [
Sometimes the requirements and the specification of this form of adaptation are clear, unambiguous and even already formalized (e.g. in the wintertime keep the internal temperature of each room between 19 C and 22 C); on the other hand, sometimes the idea and specification of the desired adaptation is given in a visual or narrative form by a designer or even by an artist (in case of artistic installations). In this case, while the desired overall effect of adaptation could be clear it may be very complex to fill the gap between this form of high level specification and a computational system.
In this second situation computer simulation can play a
crucial role in supporting the design and realization of
adaptive, self-organizing ambient intelligence systems. In fact,
traditional design and modeling instruments can provide a
suitable support for evaluating static properties of this kind
of environment (e.g. through the construction of 3D models
representing a mock-up, proof of concept of the desired
appearance or also adaptation effect but in a single specific
situation), but they are not designed to provide abstractions
and mechanisms for the definition and simulation of reactive
environments and their behaviours. Through the definition of
specific models and their implementation in simulators it is
possible to obtain an envisioning of the static features of the
ambient intelligence system as well as its dynamic response to
the behaviour of humans and other relevant entities situated
in it. This allows performing a face validation [
This paper describes the application of a modeling and simulation approach to support the design of an adaptive illumination facility that is being designed and realized by the Acconci Studio1 in Indianapolis. In particular, the designed system should be able to locally enhance the overall illumination of a tunnel in order to highlight the position and close surrounding area of pedestrians (as well as other entities such as dogs, bicycles, cars). In this case, the simulation offers both a support to the decisions about the number and positioning of lights and, more important, it encapsulates the self-organization mechanisms guiding the adaptive behaviour of lights reacting the the presence of pedestrians and other relevant entities in the environment. By providing the current state of the environment, in terms of simulated outputs of sensors detecting the presence of pedestrians, as an input to the self-organization model it is possible to obtain its simulated response, and the current state of lights. A schema of the overall simulation system is shown in Figure 1: it must be noted that the self-organization model adopted for the simulator could be effectively used to manage the actual system, simply providing actual inputs from field sensors and employing its outputs to manage actual lights rather that a Design support configuration
Pedestrian simulation
Actual sensors (motion or presence) System management configuration
Simulated
data Field data
Parameters Computational model for adaptive illumination Computational model for adaptive illumination Parameters
Actuators' states
Visualization system Actuators' states
Actual actuators (lights) virtual visualization of the actual environment.
Besides the specific aims of the ambient intelligent
system, there is an increasing interest and number of research
efforts on approaches, models and mechanisms supporting
forms of self-organization and management of the components
(both hardware and software) of such systems. The latter are
growingly viewed in terms of autonomous entities, managing
internal resources and interacting with surrounding ones so
as to obtain the desired overall system behaviour as a result
of local actions and interactions among system components.
Examples of this kind of approach can be found in both in
relatively traditional pervasive computing applications (see,
e.g., [
It must be stressed that in the Indianapoli tunnel renovation
scenario the designer had a precise idea of the desired overall
adaptive illumination effect, combining a functional overall
illumination of the tunnel – allowing pedestrians and drivers
to effectively have a sufficient visibility of the environment
but also emphasizing their presence and passage through
the tunnel – but the choice of the computational model to
achieve this effect was definitely not obvious. For instance, the
rationale of the desired adaptive illumination pattern (that will
be more throughly described in Section II) is to manage lights
as if they were animated and able to follow the movement
of pedestrians. This was tecnically impossible in the specific
scenario and the effect had to be achieved by turning on and
off in a coordinated way a set of lights characterized by a fixed
position in the environment. However, the metaphor adopted
by the designer could lead to consider specific computational
models whose first class concepts are moving entities, like
Boids [
The aims of the paper are thus twofold: on one hand it describes a concrete experience in which computer simulation was adopted to fill the gap between an abstract specification of the desired behaviour of an adaptive self-organizing environment; on the other hand the paper discusses the adequacy and feasibility of the adoption of two alternative computational models to generate the desired adaptation effect and to be effectively deployed in the real infrastructure.
The following section will introduce more in details the specific scenario in which this research effort is set, describing the requirements for the adaptive illumination system and the environment adaptation model. Section III introduces the pedestrian modeling approach, while the self-organization models that were experimented to guide the adaptive illumination facility are described in Section IV. A description of the developed environment supporting designers will follow, then conclusions and future works will end the paper.
The Acconci Studio, partner of the described research effort, has recently been involved in a project for the renovation of a tunnel in the Viginia Avenue Garage in Indianapolis. The tunnel is currently mostly devoted to cars, with relatively limited space on the sidewalks and its illumination is strictly functional. The planned renovation for the tunnel includes a set of interventions, and in particular two main effects of Motion Sensor Controlled
area Controller Communication line Light Fig. 2. A visual elaboration of the desired adaptive illumination facility (the image appears courtesy of the Acconci Studio). illumination, also depicted in a graphical elaboration of the desired visual effect shown in Figure 2: an overall effect of uniformly coloring the environment through a background, ambient light that can change through time, but slowly with respect to the movements and immediate perceptions of people passing in the tunnel; a local effect of illumination reacting to the presence of pedestrians, bicycles, cars and other physical entities.
The rationale of this local and dynamic adaptive illumination effect is better explained by the following narrative description of the desired effect:
The passage through the building should be a volume of color, a solid of color. It’s a world of its own, a world in itself, separate from the streets outside at either end. Walking, cycling, through the building should be like walking through a solid, it should be like being fixed in color.
The color might change during the day, according to the time of day: pink in the morning, for example, becomes purple at noon becomes blue, or bluegreen, at night. This world-in-itself keeps its own time, shows its own time in its own way.
The color is there to make a heaviness, a thickness, only so that the thickness can be broken. The thickness is pierced through with something, there’s a sparkle, it’s you that sparkles, walking or cycling though the passage, this tunnel of color. Well no, not really, it’s not you: but it’s you that sets off the sparkle – a sparkle here, sparkle there, then another sparkle in-between – one sparkle affects the other, pulls the other, like a magnet – a point of sparkle is stretched out into a line of sparkles is stretched out into a network of sparkles.
These sparkles are above you, below you, they spread out in front of you, they light your way through the tunnel. The sparkles multiply: it’s you who sets them off, only you, but – when another person comes toward you in the opposite direction, when another person passes you, when a car passes by – some of these sparkles, some of these fire-flies, have found a new attractor, they go off in a different direction.
The first type of effect can be achieved in a relatively simple and centralized way, requiring in fact a uniform type of illumination that has a slow dynamic. The second point requires a different view on the illumination facility. In particular, it must be able to perceive the presence of pedestrians and other physical entities passing in it, in other words it must be endowed with sensors (detecting either the presence or the movement of relatively big objects). Moreover, it must be able to exhibit local changes as a reaction to the outputs of the aforemetioned sensors, providing thus for a non uniform component to the overall illumination. The overall environment must be thus split into parts, cells that represent proper subsystems: Figure 3 shows a schema of the approach we adopted to subdivide the physical environment into autonomous units, provided with motion/presence sensors (able to detect the arrival/presence of relevant entities) and lights (to adapt the ambient illumination, highlighting the presence of pedestrians).
However, the effect of the presence of a pedestrian in a portion of space should extend beyond the borders of the occupied cell. In fact, the illumination effect should “light the way” of a pedestrian through the tunnel. Cells must thus be able to interact, in order to influence neighboring ones whenever a pedestrian is detected, to trigger a (maybe less intense) illumination.
The adopted pedestrian model is based on the Situated
Cellular Agent model, a specific class of Multilayered
MultiAgent Situated System (MMASS) [
A Situated Cellular Agent system is defined by the triple Space; F; A where Space models the environment where the set A of agents is situated, acts autonomously and interacts through the propagation of the set F of fields and through reaction operations. Space consists of a set P of sites arranged in a network (i.e. an undirected graph of sites). The structure of the space can be represented as a neighborhood function, N : P ! 2P so that N (p) P is the set of sites adjacent to p 2 P ; the previously introduced Space element is thus the pair P; N . Focusing instead on the single basic environmental elements, a site p 2 P can contain at most one agent and is defined by the 3–tuple ap; Fp; Pp where: ap 2 A [ f?g is the agent situated in p (ap = ? when no agent is situated in p that is, p is empty); Fp F is the set of fields active in p (Fp = ; when no field is active in p);
Pp P is the set of sites adjacent to p (i.e. N (p)). A SCA agent is defined by the 3–tuple < s; p; > where is the agent type, s 2 denotes the agent state and can assume one of the values specified by its type (see below for definition), and p 2 P is the site of the Space where the agent is situated. As previously stated, agent type is a specification of agent state, perceptive capabilities and behaviour. In fact an agent type is defined by the 3–tuple ; P erception ; Action . defines the set of states that agents of type can assume. P erception : ! [N Wf1 ] : : : [N WfjF j ] is a function associating to each agent state a vector of pairs representing the receptiveness coefficient and sensitivity thresholds for that kind of field. Action represents instead the behavioural specification for agents of type . Agent behaviour can be specified using a language that defines the following primitives: emit(s; f; p): the emit primitive allows an agent to start the diffusion of field f on p, that is the site it is placed on; react(s; ap1 ; ap2 ; : : : ; apn ; s0): this kind of primitive allows the specification of a coordinated change of state among adjacent agents. In order to preserve agents’ autonomy, a compatible primitive must be included in the behavioural specification of all the involved agents; moreover when this coordination process takes place, every involved agents may dynamically decide to effectively agree to perform this operation; transport(p; q): the transport primitive allows to define agent movement from site p to site q (that must be adjacent and vacant); trigger(s; s0): this primitive specifies that an agent must change its state when it senses a particular condition in its local context (i.e. its own site and the adjacent ones); this operation has the same effect of a reaction, but does not require a coordination with other agents.
For every primitive included in the behavioural specification of an agent type specific preconditions must be specified; moreover specific parameters must also be given (e.g. the specific field to be emitted in an emit primitive, or the conditions to identify the destination site in a transport) to precisely define the effect of the action, which was previously briefly described in general terms.
Each SCA agent is thus provided with a set of sensors that allows its interaction with the environment and other agents. At the same time, agents can constitute the source of given fields acting within a SCA space (e.g. noise emitted by a talking agent). Formally, a field type t is defined by Wt; Di usiont; Comparet; Composet where Wt denotes the set of values that fields of type t can assume; Di usiont : P Wf P ! (Wt)+ is the diffusion function of the field computing the value of a field on a given space site taking into account in which site (P is the set of sites that constitutes the SCA space) and with which value it has been generated. It must be noted that fields diffuse along the spatial structure of the environment, and more precisely a field diffuses from a source site to the ones that can be reached through arcs as long as its intensity is not voided by the diffusion function. Composet : (Wt)+ ! Wt expresses how fields of the same type have to be combined (for instance, in order to obtain the unique value of field type t at a site), and Comparet : Wt Wt ! fT rue; F alseg is the function that compares values of the same field type. This function is used in order to verify whether an agent can perceive a field value by comparing it with the sensitivity threshold after it has been modulated by the receptiveness coefficient.
The above introduced SCA model has beed applied to represent a very simple tunnel with two ends and some columns in it; pedestrians enter the tunnel from one end and they move towards the other end, avoiding obstacles either immobile (i.e. columns), and mobile (i.e. other pedestrians moving in the opposite direction).
The SCA Space is the same cellular space defined for the D-MAN described in Section IV. To support agent navigation in this space, in each end of the tunnel we positioned an additional site in which a “beacon” agent (a static agent emitting a simple presence field) is situated. In the environment, thus, only two types of field are present.
To exploit this environmental specification in order to obtain the above overall system behaviour, we defined two types of agent, respectively interpreting the one type of field as attractive and ignoring the other one. This can be achieve through a simple transport primitive, specifying that the agent should move towards the free adjacent site in which the intensity of the field considered attractive is maximum. The behavioural specification of these agents is completed by an obstacle avoidance rule (another transport that moves the agent towards a random different lane whenever the best possible destination is occupied by an obstacle). Finally, agents reaching their destination, that is, one of the tunnel ends, are removed from the environment and they are positioned at the other end, so they start over their crossing of the tunnel.
We employed a Cellular Automata model to realize the
local effect of illumination as a self-organized reaction to
the presence of pedestrians. CA cells, related to a portion
of the physical environment, comprise sensors and
actuators, as schematized in Figure 3. The former can trigger
the behaviours of the latter, both through the interaction of
elements enclosed in the same cell and by means of the local
interaction among adjacent cells. The transition rule models
mechanisms of reaction and diffusion, and it was derived by
previous applications to reproduce natural phenomena such as
percolation processes of pesticides in the soil, in percolation
beds for the coffee industry and for the experimentation of
elasticity properties of batches for tires [
Multilayered Automata Network have been defined as a
generalization of Automata Networks [
The irregular nature of the cellular space is not the only
difference between the adopted approach and the traditional
CA models. In fact, CAs are in general closed and synchronous
systems, in which cells update their state in parallel triggered
by a global clock. Dissipative Cellular Automata (DCA) [
The model we defined and adopted, Dissipative Multilayered Automata Network (D-MAN), takes thus the advantages of both the Multilayered Automata Network and the Dissipative Cellular Automata. An informal definition this model describes D-MAN as Multilayered Automata Network in which the cells update their state in an asynchronous way and they are open to influences by the external environment.
The multilayered cellular structure of the D-MAN is composed of three layers: the first layer is related to the basic discretization of the physical environment into cells, corresponding to a local controller. Each of these cells effectively comprises the two additional layers;
Level 0 Sensors Layer
Level 0 Actuators Layer
Level 1 Intra-controller communication
Level 2 Inter-controller communication the perception and actuation layers, respectively comprising the sensors and actuators (lights).
This structure is schematized in Figure 4. The rationale of keeping separated these cells is to be able to specify and configure specific functions describing (i) how to compute the overall internal activation state of a cell given the status of the internal sensor(s) and the current state of activation of neighbours and (ii) how to translate this state of activation into a state of actuation for that specific layer (in other words, how to translate into a lighting effect the state of the cell).
Specific transition rules must thus be defined to manage
different interactions and influences that take place in this
structure, and mainly (i) the direct influence of a sensor that
detected a pedestrian to the actuators in the same cell, and
(ii) the influence of a high level cell to the neighboring ones
(given the internal structure of each cell, due to the presence of
a specific level of actuators inside it, this interaction effectively
affects a part of a neighboring cell). Moreover, the effect of
external stimuli must gradually vanish, and lights must fade in
absence of pedestrians: while an active state of the sensor and
high activation states of neighbours cause an increase of the
cell activation state, it decreases in absence of these triggering
conditions. More details about the formal definition of the
model can be found in [
The adaptive illumination model is thus characterized by
several features that make it difficult to predict how it will
react to particular stimuli (i.e. patterns of pedestrian movement
in the related environment), from the number and positioning
of sensors and actuators, to the parameters of the transition
rule. The transition rule per se is characterized by several
parameters whose configuration can actually deeply alter the
achieved illumination effect, to the point that we developed an
ad hoc UI to show its behaviour when triggered trough mouse
clicks [
The narrative description of the desired adaptive illumination effect by the designer explicitly mentioned “fire-flies”, insects forming a swarm. This description effectively leads to consider the possibility to adopt for the self-organization of the illumination facility computational models developed to generate collective behaviours of insects (but also flocks, herds and schools) such as the boids model. This particular model is very effective in generating coordinated animal motion, with a relatively simple computation essentially based on the mutual local perception of individuals situated in a physical environment. In our system, lights cannot actually move, but boids can represent the fact that a light is active in a predefined environment structure representing a discrete and finite set of allowed positions. The movement of boids represents thus the dynamic update of lights’ statuses in the illumination facility.
The basic model comprises three simple steering behaviors, describing how an individual boid maneuvers basing on the positions and velocities its nearby flockmates. These behaviours, also schematized in Figure 5, are: cohesion: steer to move toward the average position (the centroid) of local flockmates; separation: steer to avoid crowding (i.e. in the opposite direction of the the centroid of local flockmates); alignment: steer towards the average heading of local flockmates.
The notion of locality mentioned in these rules is essentially related to the possibility of boids to detect others of their kind situated within a certain range from their current position. These three steering behaviours produce vector representing a contribution to the overall action of the single boid, that is thus obtained as the vectorial sum of these contributions multiplied for specific scalar constants. These constants must be properly calibrated to avoid excessive dispersion and cohesion of the boids. Moreover, in this specific application, boids must also be attracted by people situated in the environment: a fourth contribution to the overall boid behaviour must thus be introduced, otherwise the boids wander in a realistic way (from the point of view of simulating a collective behaviour) but completely ignoring the presence of humans in the environment.
Before introducing an additional contribution to the behavioural specification of a boid to tackle this issue, we first considered that the basic boid model is conceived for a continuous environmental spatial representation, that is not suited to this situation. The adopted approach was to translate the cohesion and separation contributions of the model into elements of the aforementioned SCA model, mainly for two reasons: (i) this model is by definition discrete and (ii) it was already adopted to model pedestrians and the environment they are situated in. Alignment was not considered since SCA agents are not characterized by a direction or heading in space, but just by their position (i.e. a node in a graph structure).
To realize a model featuring the main characteristics of
boids using SCAs we decided to provide each boid with a
form of presence field, diffusing a sign indicating its position
in nearby sites. Presence fields’ diffusion function decreases
by a constant value the intensity of a field for each node
the site crosses in the course of the diffusion operation until
the value is void, while the composition function simply
adds up the value of all presence fields in a given site. In
this way, presence field in a given site represents a measure
of its crowdedness and it can be used by boids to select
sites that represent a good compromise between cohesion
and separation. The behavioural specification of SCA boids
comprises thus essentially two basic actions, a field emission
and a trasnport action interpreting the value of the presence
field as sort of social force [
The above introduced models were adopted and tested in the Indianapolis tunnel scenario; both of them proved their adequacy to effectively represent a formal, computational, non ambiguous and executable specification of the designer’s narration. They undergone a successful face validation, that followed several iterations to define a good value for the models’ parameters to achieve the desired results. However, the models have specific features that can have an impact on some of the non-functional properties they exhibit.
It must be stressed that even if the designer used the term “fire-flies” in the narrative description of the desired effect, the idea to follow the metaphor and to employ one of the most commonly adopted model for this kind of collective behaviour does not necessarily imply a smooth and simple definition of a model achieving the desired effect. First of all the boids model is based on a continuous spatial representation and thus it must be adapted to the discrete spatial structure of the illumination facility. Then a modification to the basic model must be introduced to achieve a “goal driven” behaviour (i.e. the tendency of boids to move towards people, preserving the swarm behaviour). Finally it must be noted that this modeling approach does not take into account the effective infrastructure that will be effectively employed to realize the illumination facility.
The CA based approach is based on the idea of viewing the environment itself as an assembly of autonomous units able to interact with their neighbours. The adaptive illumination effect is achieved as a reaction of these units to an external stimulus generated by sensors and as a results of their interaction. As a result it is much simpler to conceive a direct effective implementation of this approach in a concrete system made up of a set of micro-controllers responsible for the monitoring of a certain part of the environment and for the control of the lights it includes. On the other hand, there is no simple way to distribute the boids model in a distributed control system: the simple fact that boids can perceive the presence of other individuals of their kind in a potentially distant position, according to their range of perception, leads to consider that in case of distribution of this form of computation to different autonomous units would lead to higher costs in terms of network communication, that is instead essentially constant in the CA based case.
The paper introduced a simulation approach to supporting
the design of an ambient intelligence infrastructure aimed at
improving the everyday experience of pedestrians and people
passing through the related environment. A specific scenario
related to the definition and development of an adaptive
illumination facility was introduced, and two different
computational models model specifying its dynamic behaviour was
defined. An agent-based pedestrian model simulating inputs
and stimuli to the adaptation module was also introduced.
The models described in the previous Sections are part of a
prototype supporting the design and definition of the overall
illumination facility through the simulation and envisioning
of its dynamic behaviour according to specific values for the
relevant parameters (e.g. parameters of the transition rule of
the CA, but also the number of lights and sensors, and so on);
a screenshot of the system is shown in Figure 6: the central
window shows a three-dimensional view of the tunnel, while
the top panel shows a bi-dimensional view highlighting the
position of pedestrians and the state of lights. The
simulation takes place in the actual 3D model of the tunnel, that
was adopted to achieve in a semi-automatic way a discrete
representation of the environment that was adopted both for
enabling the pedestrian simulation and for the positioning
of lights. Part of the simulator, based on an platform for
agent–based simulation [
The renovation project is currently under development on the architectural and engineering side, whereas the introduced models have shown their adequacy to the problem specification, both in order to provide a formal specification of the behaviour for the system components and possibly as a control mechanism. The realized prototype explored the possibility of realizing an ad hoc tool that can integrate the traditional CAD systems for supporting designers in simulating and envisioning the dynamic behaviour of complex, self-organizating installations. It has been used to understand the adequacy of the modeling approach in reproducing the desired self-organized adaptive behaviour of the environment to the presence of pedestrians. We are currently improving the prototype, on one hand, to provide a better support for the Indianapolis project and, on the other, to realize a more general framework for supporting designers of dynamic self-organizing environments.
The authors wish to thank some collaborators at the Acconci Studio and in particular Nathan DeGraaf, Jono Podborseck and James Clar for their collaboration to the present work and research effort.