Reconstructing Urban Labyrinths

Reconstructing Urban Labyrinths EURAU’12

ABSTRACT: Whilst in literature and common language the labyrinthine and the maze-like are synonymous, the use of these ancient terms in contemporary urbanism is commonly misunderstood. Based on an appreciation of understandings embedded in the cultural device-symbol-path of the labyrinth, this paper re-examines Space-Syntaxic presumptions about modern cities.

A novel model is outlined and experimental evidence provided to illustrate how distinct spatial movements relate to dialogues between desire-sets in an ecological system. It is suggested that this correct understanding of the ancient relationship between labyrinths and city cultures can be extended to provide re-interpretations of urban conditions.

During the paper, the author provides insights as to how this reconstruction of a collective cultural memory is being utilized in three separate fields of contemporary urbanism. Outlines are given on its influencing of architectural design, ongoing archaeological analysis of the palace-settlements of Minoan Crete and explorations in full-scale interactive experimental installations.

KEYWORDS: labyrinths, mazes, interaction, behaviours, space syntax, Olympics Sam McElhinney Bartlett School of Architecture, University College London Wates House, 22 Gordon Street London, WC1H 0QB

[email protected]

1. Introduction

Getting from 'A' to 'B' often preoccupies the city 'user'. This occurs at many rates and modes, the most embodied of which is walking. Walking offers the commonest urban experience and so how we choose to walk between points is a key concern of many urban designers.

In designing, we mediate a conversation between ourselves as designer and as observer. [Glanville, 1999] A model of 'self' is used as an internal model of a 'user' within a hypothetical design, allowing testing and iteration of ideas [Gage, 2007a]. Essentially designers project what might delight them onto a future observer and trust that this person will construct similar understandings and delight from the completed system. [Gage, 2006]

This results in work that invariably reflects the designers' own tastes; an ego driven process which is often regarded as a problematic. Urban designers in particular have attempted to progress from this by using generic 'user-constructs'. The 'Space Syntax' grouping, headed by Bill Hillier at University College London (UCL), are recognised as being successful at this. As with many Urban designers, they are fascinated by how pedestrians get from 'A to B'; their inter-connectivity.

Key papers from Space Syntax have quantified their approach [Turner et al., 2005] and set out how it informs a species of 'point isovist' agents. [Penn and Turner, 2002] These have been employed to analyse movement at building and urban scales [Turner and Penn, 2007] and also in generative morphological simulations. [Turner and Penn, 2004] This work does hold merit; but a later critique will point out some limitations.

2. The Labyrinth and the Maze

Common literature regularly uses the terms 'labyrinthine' and 'maze' in an interchangeable manner. Their originally separate meanings have long been blurred in prose and lyrics, perhaps because this distinction was a spatial, symbolic one, usually inconsequential to a reader.

A misunderstanding arises when this literal crossover recycles to describe urban networks.

While enjoying the relaxing cruise, you’ll notice how the dark buildings of the city seem to be stacked on top of one another along narrow, labyrinthine streets. [Visit Portugal, 2012]

It is common to find commentary of this ilk; urbanists use it too. Perhaps they do not want to write that they became disorientated by a complex city grain. They apply a suggestive term, but what they really mean is that they found themselves lost in a maze.

A labyrinth is not a maze. Labyrinths offer understandings of intimate, bodily space and the formation, occupancy and inter-connectivities of that around it. [Kern, 2000] They are encountered as complete, coherent symbols in which space is wound around upon itself. This construct is ancient and its meanings obscured, but clear spatial distinctions illuminate these.

A path system can be regarded as multicursal: a network of disorienting interconnecting routes and choices. It may contain multiple branches and dead ends that confuse. This is a maze.

Alternatively, a known path is resolved as a single, monocursal route. Once embarked upon, this may fold, twist and turn, but will trend towards and ultimately reach a destination; this is a labyrinth.

A pedestrian walking these two topologies behaves in very different ways.

2.1. Two Users

The labyrinth-user accepts and internalises a constant, known route, presented symbolically. Whilst following this, they are free to engage in abstract, introvert reflection, ‘wandering’ in their mind. Given a predetermined external bodily path, the labyrinth walker explores internal mental paths.

Despite apparently being wholly predictable (walking a set path), this occupant may develop innovative internal thoughts. These are expressed in unanticipated behaviours; wilfulness, even divergence from the ‘thinking-route’ of the labyrinth.

In a labyrinth, movement expresses an internalised, abstracted thought process.

Conversely, a maze-user has no solution. They must 'find their way' by direct dialogue with their surroundings [Gibson, 1979 & Brooks, 1999]. A continual assessment of environmental novelty primes simple decision-making mechanisms; response to stimuli or pre-learnt social constructs. 'Dead ends' are avoided, whilst longer vistas are explored. The commonality of these responses makes movement predictable.

Over time, successful movement strategies assemble to form a cogent mapping of the environment. This provides a ‘labyrinthine path’ solution, but prior to this, internal reasoning with a hypothetical model of provides little benefit to a real-time system in a real world ‘maze’.

In a maze, movement expresses an externalised, embodied thought process.

2.2. Extrapolation

We can extend this model of two users. All urbanism could be considered as labyrinth or maze and our 'user-constructs' could be modes of occupancy. This classification is not a simple division; it is not a model of absolutes. The two polar spatial understandings are coincident, overlapping and modulating as the user’s comprehension and the environment change. What have is a novel model for interpreting urban realms.

Unexplored cities and street-scapes appear maze-like. Their occupants seek delight as in a maze; through a gradual construction of understanding. This is a threading together of known familiarities - in architecture these include arcades, stairways or rooms, in urban space, avenues, bridges and squares. These elements are augmented by pre-learnt social constructs; they are understood, labyrinthine configurations which, in recognition, allow reflection.

Around these momentary events, emotive memories conglomerate, allowing a patina of experiences to signpost paths through the city. Orientation and direction become less relevant than accumulative exploration; establishing an intense, immersive relationship between the user and the urban landscape. Knowledge of a network of landmarks is continually recombined both in the mind and in the act of walking to create infinite varieties of labyrinth pathways.

In this way, the route from 'A to B' in the city is rarely an optimised, direct one. It meanders and is circuitous. It perhaps includes 'C', 'D' or 'Z' providing constant opportunity for novelty and delight. Its known, labyrinthine nature allows introspective, wandering moments; but the activity of most urban environments retains a degree of maze. Thus we are often predictable; the unexpected stimuli through which we move causing reversion to maze-navigating behaviours.

The premise is that all space is found, experienced and inhabited in a state of ‘switching’ flux between the diametrically opposed topologies of maze and labyrinth. This offers insights into how urban public space might evoke continual delight in the user.

Perhaps in navigating the complex built environment we constantly perceive different spatialities and seek novelty (and thus delight) by switching between modes of understanding and movement. Successful cities afford this modulating exploration; this flowing process has not yet been examined in urbanism.

2.3. The In-between Space

We should therefore consider what it means for users to engage a path that alternates from a labyrinth into a maze (or vice-versa). Diagrammatically this switch is not between typologies (Nicolson, [2005] already offers excellent taxonomic work in this area), but from route to anti-route, topologically fully or partially collapsing the labyrinth and creating mazes.

Such an actively ambiguous space poses questions. How might it be constructed? What are the experiential ramifications of inhabiting it? Might such an environment in some way 'learn' and react to occupancy? What observational insights could these interactions provide?

3. Maze-Agent Ecologies

3.1. Critiquing Space Syntax

Alasdair Turner’s Space Syntax simulations contain two elements; direct perception agents that assess a visual field to determine movement and an environment that observes and reconfigures. [Turner and Penn, 2004]

In the simulations, spatial plans are pre-assessed and their 'inter-visibility' data (point-isovist information) is stored in a ‘look-up table’. This ‘exosomatic visual architecture’ provides agents with an external ability to infer spatial relationships. [Penn and Turner, 2002] Each agent samples this table, choosing a random point from their field of view. They walk toward this for a number of steps before repeating the choice. This creates ‘wandering’; forward locations tend to be chosen, and so the agent traverses space.

A limitation of this work lies in its pre-computation. Turner's agents only see that which is stored in the lookup table. Environmental change requires a new inter-visibility analysis; without this the agents’ ‘vision’ would be inaccurate. Spatial modulation is thus carried out on a turn-based method and relies on a genetic algorithm; many thousands of generations are required to produce significant results. Inhabiting agents are really passive explorers of a static environment; experiencing none of the activity of the real world.

If we consider the Space Syntax approach in terms of the proposed bi-polar user construct, Turner's agents are solely 'maze users'. They make simple depth assessments and have no capacity to build labyrinthine understandings. The inter-visibility analysis upon which they rely optimises possible routes; although agents 'wander' this occurs through randomisation of this optimal routing rather than from recognition of familiar spatial elements. Agent paths from 'A to B' are reductionist; they do not aspire to the complex engagements of city user.

This approach emphasises direct perception above individual desires and behaviour traits. Accumulative averages make this effective when considering urban morphologies over extended times, but it cannot approximate the momentary richness apparent in cities. An agent with desires, capable of autonomously resolving space, could react behaviourally in real-time, allowing its environment to be continually active. Such fluctuation may produce intriguing ecologies without requiring genetic iterations and so might better highlight emergent urban orders.

3.2. An Alternative System

An in-silico system was developed using the open source coding language, ‘Processing’. This had the expressed aims of:

- showing that a direct perception agent can autonomously resolve spatial interrelationships. - exploring how an environment might react in real-time to agent actions; and showing that this ecology might not require genetic algorithms to resolve toward a structure that reflects ‘learning’ about its occupancy. - determining how desire interplay might affect agent or maze-user movements. - investigating what movement patterns from such an ecology might depict and if it generates recognisable architectural types.

3.3. Dialogues

The experimental context consisted of numerous 'doors' in a triangular grid forming a 'maze matrix'. Each was capable of being closed, primed or open. Using a cascade algorithm, doors were opened sequentially, establishing routes through the maze. As these twisted and turned differing spatialities were created; some were 'roomlike' whilst others formed 'corridors'.

The matrix had an inherent goal; desiring to exist in a ‘resolved’ state, with complete, stable path structures. It remained like this unless disturbed by the opening of a new door; at which point the cascade restarted.

An agent was designed, utilising 'depth vectors' to autonomously assess surrounding space. These vectors were arrayed to mimic human vision and a comparative system allowed them to impart understanding of the type of spatiality ('room' or 'corridor') occupied. Each agent was given desires which compelled it to seek or attempt to construct particular spatial types by opening doors. This in turn triggered the maze to seek a new stable path configurations.

Whilst successfully occupying or unsuccessfully searching for a particular spatiality, the agent became 'bored' or 'frustrated' and altered its goals respectively. This produced a complex of dialogues between the desires of the agent (to explore, or to accept the current configuration) and of the environment (attempting to form coherent routes) that could be read in the agent trails. Study of these allowed categorisation of distinct movement typologies. (Figs. 1-5)

Fig.1: trans-type movement: sustained and direct path seeking and following.

Each typology reflects a specific modulation of the internal desires of individual agents; expressing these outwardly as distinctive 'dances', patterns that attempt to satisfy spatial goals. Lacking the capacity for internal cognition, the agents display an external 'bodily' reasoning in order to manipulate and navigate the maze.

Fig.2: interrupted trans-type movement: path seeking/following punctuated with room seeking/occupancy.

Fig.4: orbital movement: the two desire sets are equally weighted.

Fig.3: meander movement: room seeking desires outweigh path seeking; an exploratory path evolves.

Fig.5: complex knot movement: allowing continual variation of desires results in hybrid combinations and path linkages. Larger matrices increase scope for this.

Some of these movements are directly equitable to forms long established as labyrinthine elements; the meander and orbit are particularly common. The structure of a labyrinth is being formed from within a maze as it accommodates exploratory desires. As with the urban explorer, we find an intimate linking of the wanderings of the mind, the constraints of the body and the path of the labyrinth. [Nicholson, 2005]

3.4. Ecological Morphologies

A second experiment explored emergence. Each door was given abilities to record agent proximities. This data influenced which doors were opened in the resolving cascade. The simplest choice algorithm was random but later ones chose doors:

- to predict and satisfy agent penetration of the matrix (more popular doors) - to predict and frustrate agent penetration of the matrix (less popular doors)

Depicting doors that were closed for 70% or more of the experiment illustrated which approximated to walls. A second recording marked each agent route. After just thirty agent visits emergent ‘tracks’ (common paths) and ‘screens’ (structures of closed doors) could be observed; comparing favourably to the several thousand generations required by Space Syntax simulations. (Figs. 6-8)

Fig.6 (above): Random door opening produced concentric rings, akin to courtyard cloisters. Agent 'exploration urges' were expressed by crossing rings.

Fig.7 (above): Satisfaction of agent door choice produced diagonal structures, promoting access into the matrix. Initially this benefited exploration; but in later generations these screens became permanent, dictating movement.

Fig.8 (right): Frustration formed lateral structures; forcing agents to travel sideways. High activity near the entrance made this marked here, creating 'vestibule' spaces. In later generations the screens behind here dissolved into single panels; creating central areas of exploration.

Each ecology went through several generational phases. This accumulation and change depicts the evolution of a dialogic accord between the two systems. The maze developed understandings of how it was being understood; this ‘learning’ was expressed in the ‘screen’ constructions. Gradually the environment became effective at influencing occupancy to suit its desires.

Starting from a maze, this system has demonstrated the folding of space into labyrinthine (or recognisable) orders. It hints how common architectural and urban features and compositions may have accumulated from interactions of emotive and desire based occupancy. This provides an experimental basis for a novel understanding of urban order.

4. Current Applications

4.1. The Cretan Palace Complex

Since the initial digital experiments the agent genus has been entirely re-coded. It now achieves a greater spatial resolution and can be implemented in high numbers. Further optimisations allow it to operate successfully within more complex active architectural plans. A collaboration between the author and Quentin Letesson [Letesson, 2009] is deploying these agents within reconstructed plans of the ancient Cretan Palaces of Malia, Phaistos and Zakros.

To date this has demonstrated that the approach can provide similar results to Space Syntax including agent tracks and an inter-visibility plan outputs. It can also produce isovist pathway mappings that record spatial comprehension at key behavioural switches. (Fig. 9)

Fig.9 (left): Agent analysis of Zakros Palace

Further behavioural layers are now being added to the agent genus. Future experiments will seek common behaviour patterns within the palace complexes for comparison to established archaeological analysis. A degree of correlation is expected to identify labyrinthine orders within these proto-urban constructions. Particular focus is being given to areas of 'polythyra', pier and door partitions which were opened or closed depending on spatial occupation; a device which strongly emulates the switching spaces experimentally identified previously.

4.2. Experimental Architectures

The themes outlined in this paper were first implemented in built form in the author's own 'Switching Labyrinth' installation. This was constructed in a warehouse near Euston Station, London, in June 2009, providing a framework for two weeks of experimental occupation.

Consisting of 250 metres of curtains hung from a suspended space-frame, the installation provided pathways around two small 'room' spaces. As occupants explored these, an algorithm observed their progress and periodically shifted curtains. This collapsed the extant labyrinth structure and produced fragmentary, 'maze-like' arrangements; these in turn coalesced into novel labyrinthine routes based upon user movement.

The author now tutors graduate students at UCL's Bartlett School of Architecture who are developing full scale interactive installations. All of these explore notions of switching space and behaviour and how they might apply to create novel, delightful architectures.

4.3. Olympic Wayfinding

2012's Olympic Games centred around Stratford Park, in East London. In the four years prior to the Games, this was 'regenerated' from industrial wastelands to create a new urban public space with extensive gardens, bridges and plazas. With four major venues within the park, creation of a coherent pedestrian wayfinding strategy was a paramount design issue.

Fig.10: Olympic Wayfinding, Zone Beacon, rotational views.

Fig.11: Olympic Wayfinding, Main Beacon, rotational views.

Within professional practice, the author acted as Architect for the design and delivery of a family of thirteen wayfinding structures across the Olympic Park; six 7m high 'zone beacons', five 14m high 'main beacons' and two 12m high, 32m wide entrance gantries. (Figs 10-12) Rather than simply highlighting optimal routes, these high profile structures were detailed to echo ancient and familiar architectural features (such as masonry quoining); facilitating them to be recognised and operate more fully as a chain of landmarks for the passerby's mental re-construction of labyrinth paths across the landscape.

Fig.12: Olympic Wayfinding, Entrance Gantry.

5. Reconstructing the Labyrinth

Theseus, the hero of Greek legend, is known for navigating the Minoan Labyrinth and slaying the Minotaur. Less well known is how, on his return to Greece, he became a prototypic urban designer, recognised for the first synoecism (dwelling together or civil union) of the polis of Athens. A conurbation of small towns was threaded into a monocursal entity; at the very beginnings of public urban culture lies a hidden labyrinth order.

We no longer recognise this order in the contextualisation of civil spaces of our cities. If we did, it would provide an innovative emotive and behavioural re-alignment of the syntaxes that we currently apply in readings of urban complexes. The works described in this paper are small interventions which have attempted to establish evidence for and design using this approach.

Reconstruction of an understanding of continuous labyrinthine systems within contemporary city complexes offers potential alternative approaches to their shaping and occupation. Doing this would break down current reductionist readings of navigation ‘from A to B’ within the city suggest innovative urban orders which have at their heart multi-various, rich, active interpretations of public realm.

6. Acknowledgements

All images are copyright the Author.

Thanks go to my colleagues at Surface Architects, as well as the London Organising Committee for the Olympic Games. Professor Stephen Gage of UCL has, as ever, been supportive.

Final thanks go to Elizabeth Upham, for her patience.

7. Bibliography

[Brooks, 1999] Rodney A. Brooks, Cambrian Intelligence: the early history of the new AI, (Cambridge, MA: MIT Press, 1999)

[Gage, 2006] Stephen Gage, ‘The Wonder of Trivial Machines’, Systems Research and Behavioural Science, vol.23 (2006) pp.771-8

[Gage, 2007a] Stephen Gage, ‘How to design a black and white box’, Kybernetes, vol.36 no.9/10 (2007) pp.1329-39

[Gage, 2007b] Stephen Gage, ‘Constructing the User’, Systems Research and Behavioural Science, vol.24 (2007) pp.313-22

[Gibson, 1979] James J. Gibson, The Ecological Approach to Visual Perception, (Boston: Houghton Mifflin, 1979)

[Glanville, 1999] Ranulph Glanville, ‘Re-searching Design and Designing Research', MIT Press, Cambridge, MA (1999)

[Kern, 2000] Hermann Kern, Through the Labyrinth: Designs and Meanings over 5,000 Years, (New York: Prestel, 2000)

[Letesson, 2009] Quentin Letesson, Du phénotype au génotype: analyse de la syntaxe spatiale en architecture minoenne, (Louvain-la-Neuve, 2009)

[Nicholson, 2005] Ben Nicholson, ‘Chapter V: Labyrinths of Form and Matter’, Horror Vacuii, (unpublished, personal correspondence, dated 9 March 2005)

[Penn and Turner, 2002] Alan Penn and Alasdair Turner, ‘Space Syntax Based Agent Simulation’, in Michael Schreckenberg and Som Deo Sharma (eds.) Pedestrian and Evacuation Dynamics (Berlin: Springer, 2002) pp.99-114

[Turner et al., 2005] Alasdair Turner, Alan Penn and Bill Hillier, ‘An algorithmic definition of the axial map’ Environment and Planning B: Planning and Design, vol.32 (2005) pp.425-444

[Turner and Penn, 2007] Alasdair Turner and Alan Penn ‘Evolving direct perception models of human behavior in building systems’ in N. Waldau, P. Gattermann, H. Knoflacher, and M. Schrekenberg, (eds.) Pedestrian and Evacuation Dynamics (Berlin: Springer, 2007) pp.411-22

[Turner and Penn, 2004] Alasdair Turner and Alan Penn, ‘An Ecological Approach to Generative Design’, Design Computing and Design, (2004) pp.259-274

[Visit Portugal, 2012] www.visitportugal.com/BoatTripsinPorto (last accessed 04/06/2012)

8. Biography

Sam McElhinney is an Architect who studied at Cambridge University and the Bartlett, University College London. On graduation from UCL he won the Ambrose Poynter Prize for his thesis 'Labyrinths, Mazes and the Spaces Inbetween' and in 2010 he won ‘Best Paper’ at the European Meeting of Cybernetics and Systems Research, Vienna.

Since 2005 Sam has been a key member of Surface, an award winning design practice. He is currently lead Architect for Surface's most high profile built project yet, a series of Wayfinding Structures for the 2012 Olympic Games.

Sam tutors Architecture at Canterbury and UCL universities and has lectured at Cambridge University and at the MAXXI Centre in Rome.