The digital city

THE DIGITAL CITY MICHAEL GRANT Technical Director ABACUS Glasgow - United Kingdom www.abacus.ac.uk Introduction The first project to be reported took place in the mid 1980s and was centred on the City of Glasgow. Initially positioned as an academic exercise in the capture, collation and use of geometrical data on a large geographical scale the activity grew into the construction of one of the first computer based urban models. Once established the model attracted widespread interest and through use grew to encompass over 25 square kilometres of the city. The development of the model stimulated many lasting software and modelling techniques for the design and construction of the Digital City. The next project described came at the beginning of the 1990s when the geographical interest switched to the neighbouring city of Edinburgh. This also saw a move with the computing platform due to the introduction of engineering workstations. The Glasgow model had been developed in an era of remote mainframes, vector graphics terminals and wire frame rendering offering a largely static papered based form of output. The evolution of computing capabilities now brought the potential of a dynamic and colourful vision of the city and bit mapped window managers offered the prospect of integrating additional, associated graphical and textual information. This heralded a shift in emphasis from just representing the formal structure of the city towards exploring the prospect of an integrated urban information system wherein the buildings became an interface to further sources of urban data. This then makes the transition to the next era more easily understood. The early 1990s saw the introduction of that set of technologies collectively termed Multi Media. The deployment of this technology offered an easy route into harnessing the potential of mixed media in exploring the city. This lead to a series of targeted investigations each looking to exploit some aspect of the new computing technology. These projects recognised the limitation of geometrical modelling and looked to exploit digital video, interactive panoramas and the generic interactivity of the multimedia platform to represent the city. A final evolution in this research theme saw the initial Glasgow model, now some 15 years old being repurposed as an interactive city database delivered over the internet. This realised some of the initial proposals for the model and through the advance of technology could offer a truly virtual experience of exploring the city and accessing the data that it contained. Development Context By the early 1980s there was growing concern in many industrial nations as to the threat of irreversible change occurring to their traditional townscape and landscape environments through unregulated development. In Britain the Royal Fine Art Commission had attacked architects for not considering how their buildings might integrate with the visual environment and had called for refusal of planning permission unless the new design could be shown in the context of neighbouring buildings and features. Consequently, within the building profession there was a perceived need for a better means of addressing these concerns at an early design stage. (Bridges 1982). In turn this lead to a growing interest in the development of tools and areas of application that fuelled research addressing problems relating to modelling and visualising the built environment. The Glasgow Model Initially, the data set that was destined to become the Glasgow Model consisted of a number of individual buildings on the Strathclyde University Campus. These buildings were being modelled as part of an ongoing, annual, student exercise. The choice of this set of buildings had been made largely due to the availability of data and ease of access to the originals. After the course described above had been running for some time it was apparent that this existing data set now comprised most of the buildings on Campus. Each of the buildings had been modelled in isolation, this being a limitation of time and the processing power of the available computer environment. However with the continual upgrade of the University's central computing resource it became possible to consider amalgamating these individual models into a single data set. As individual buildings these models lacked any sense of place, so while they might be considered worthy representations in their own right, it was also readily apparent that, should it be possible to work with the entire data set, many more evaluations would become possible, such as an appreciation of access, egress and blocking or massing. Potentially this amalgamation would create an entity whose worth would be more than the sum of its parts. The impetus to expand the scope of the Campus Model came from Glasgow's selection as the venue for the National Garden Festival. The Campus Model was held up as an exemplar of what could be achieved if the necessary resources were applied to capturing a larger area of the city. This was enough to persuade Glasgow Action, a local interest group formed by prominent Glasgow businessmen, to fund further development. Given the experience of constructing the Campus Model it was easy to provide a rational and formulate a simple strategy, broken down into three main phases that formed the first and continuing vision for the Digital City. Stage 1: Construct a simple topographical database of the city centre - to include streets, railways, rivers and city blocks. Stage 2: Extend the scope of the first stage and to detail specific buildings of interest. Stage 3: Integration of the topographical model with existing alpha numeric databases of building attributes for example: address; usage; ownership; age; architect; commercial value. The construction of the model was in part a commercial exercise, in that a sponsor was looking for a return on their investment, and in part an academic exercise in that data capture and modelling on this scale was a technological first. This in turn meant that some of the methods employed were experimental and also that new research had to be conducted to establish the availability of data sources and the most productive means of data capture. On investigation the availability of useful sources data proved to be relatively limited. These being summarised below: Aerial photography. Pre-existing architectural surveys. Field surveys. Physical models. Cartography. What became apparent was that no one source of information could satisfy all requirements. Photogrametric readings from aerial photography provided some of the building heights while existing architectural surveys were useful in obtaining a more complete coverage. Field surveys helped to piece together missing information. Surprisingly a large 1:50 scale wooden model held by the town planning department proved to be a source of easily acquired and accurate data. In essence the Glasgow Model database became a three dimensional plan representation of the city so it is not surprising that accessing existing cartographic information proved to be the most efficient route to capturing basic building information. The survey area was manually digitised from the 1:1250 map base by a group of researchers, each allocated an area of responsibility. Within each of these areas the group would estimate the heights of each individual building from the sources identified above and this data would be transferred to the model database. Within the model the transport networks, including road, rail and river features, were recognised as an integral part of the urban environment and from a visual and aesthetic perspective serve to bind the city fabric together. These features were treated in a number of distinct ways. Individual entities, which had a geometrically distinct character of their own, such as the river Clyde, segments of motorway and railway tracks, were digitised as separate geometry and stored as such. City centre streets were captured in relation to the block that they surrounded. These features were treated as line segments and were assumed to be defined by the pavement edge, where available, and as such were more representative of the land surrounding the buildings rather than the road surfaces themselves. Since Glasgow is not a "flat earth" city and much of the character of the urban fabric comes from the rolling disposition of the buildings on the terrain it was evident that to achieve the correct relative elevation of the buildings then some recognition of the underlying landform was essential. The adopted strategy was to digitise contour stings which were subsequently converted to a digital elevation matrix. The computational procedure then implemented for this process utilised nearest neighbour sampling to build a polynomial function which could then be sampled at a frequency relating to the output parameters of the matrix. This data set had a regular cell dimension of 50 meters which was considered as being able to provide enough detail to represent the texture of the urban landform. The terrain model was primarily seen as an intermediate stage - most useful for its ability to transfer the height data from the source to the city model. This process had previously been developed as a tool for assessing the visual impact of electricity transmission towers within the landscape. (Maver, 1982). Part of this process required that geometry representing pylons and trees should be "floated" down onto the DTM to take up their correct elevational position. This method was adapted to provide the same automatic capabilities for heighting the buildings and roads within the model. As the development of the model progressed a fundamental decision was made as to how to break down the gross area of the city into logical units that could then be assigned to the teams performing the data capture. Originally this division was executed on an arbitrary basis, each team member being assigned the next street block as the previous area was completed. On completion of any section the new data was amalgamated into the whole within a single data file. This soon proved unworkable as the difficulty in manipulating and visualising the model grew in proportion to the size of the data-set. To reduce the monolithic structure of the model a decision was made to superimpose a grid, based on the 1:1250 OS map size, and localise an interlocking structure of city blocks within this layout. Once the city had been subdivided sets of buildings within each region were captured into a single file, numbered sequentially, proceeding in an anti clockwise direction. The intention was to use the hierarchy of the file store on the computer to provide an order for the data. This was achieved by establishing "Glasgow" as a user account on the machine, which then established a "root" for the structure, then each district formed a named directory which in turn contained a number of files containing the building data. Within these files each geometric entity represented a group of buildings captured in a clockwise order from the northwest corner of the block. The hierarchy branched to provide a tree for the transport network in parallel to that of the building data. Figure1. Composite image representing the structure of the Glasgow Model. This structure is represented in Figure 1. The image shows, in the background, the geographical extent of the model divided into a grid at a scale of 1:1250. Superimposed at top left is the graphical representation of how the cartographic data relates to the city structure. This representation was used as an interface to the model as the inset image at bottom right shows. This is the geometrical data resulting from picking the shaded area in the interface. The final inset image at the centre is representative of the applied detail and the number of buildings held within a single data file. The Glasgow Model was built as a natural progression in the pursuit of a technological means of representing the built environment on a large scale. The preceding description does not do justice to the number of iterations in its format or variations in the method of construction. The developments reported here spanned a period of some four to five years, many people contributed to the project, much of the model was continually reworked – some areas several times – and within this period the model grew and metamorphosed from an academic exercise into a design tool used by a wide range of Glasgow professionals. One key aspect of its success was its transformation from a map to a model. As reported earlier, the first version of the model was largely a 2.5-D extrusion of the cartographic data. As such it was devoid of detail or formal articulation and many users reported difficulty in orientating themselves within the city. As the construction progressed it became apparent that the model needed to account for the way in which people form a mental image of the real world. This image is formed by the conjunction of a set of physical forms which are comprised of five urban elements; paths, edges, districts, nodes and landmarks. This concept is explored in much greater detail in "The Image of the City" (Lynch, 1960). It is still interesting to note that the same theories are applicable to the Digital City. As the model evolved, becoming a more faithful and graphically interesting representation of the city, interest in exploiting its potential grew proportionally. While there were some technical uses for the model – such as solar availability studies – the most common application was to regard the model as an infinite source of hard copy with designers requesting views onto which they could sketch their proposals. (Grant, 1993a). The Edinburgh Model The original Glasgow Model was pioneered in order to study the methods of data capture, storage and representation required by large scale urban geometries. Over this period the capacity and capability of available computer hardware proved equal to providing a platform for these activities. However, whereas the hardware and software environment had matured in line with the scope of the project the single most limiting factor remained unchanged in that the collection, description and implementation of geometry had remained the one most difficult, time consuming and error prone activity. Unless this restriction could be removed the implication would be that the scale and scope of constructing large scale urban models would always be constrained. (Grant, 1993b). Within the UK, the Ordnance Survey had always been the single body charged with acquiring and maintaining cartographic products at a national level of coverage. Given that this resource is held to be the definitive data set from which all other similar mapping products are either derived or referenced within, then it would appear prudent to adopt this data as the prime source of two dimensional scale and location. A collaboration between ABACUS and the Ordnance Survey delivered access to a new generation of cartographic information in a digital format and it was decided to trial the development of software intended to bridge the transformation between the 2-D and 3-D domains. This project was targeted on Edinburgh with the intention of modelling a portion of the Old Town and delivering a model that would incorporate aspects of good practice learned from the construction of the preceding model of Glasgow. Software was written to generate the basic building blocks of the model directly from the cartographic data. This was performed by structuring the 2D line segments from the map base into individual polygons and combining individual elements of the general line detail into roads, paths and similar components of the transport network. A digital terrain model with a 10m grid dimension was generated in a similar manner to that employed within the Glasgow model and the geometrical components of the buildings and the transport networks were then aligned with the height of the terrain. This procedure proved to be an accurate and automatic means of generating contextual areas of the city. The individual building heights still remained as the missing component but by substituting an assumed value the data set could provide a good approximation of the urban fabric. The basic building block for a more detailed model was obviously the 3-D volume obtained from the original data source. However this was purely an extrusion of the 2-D plan and might not bear much relationship to the true 3-D form. Typical problems occur when the 2-D boundary contains spurious representative detail and also when the outlines of secondary structures are incorporated into the outline of the main building. These secondary structures may be extensions, lean-tos or utility sheds and may not be representative of the main structure of the building. These problems and other artifacts reflecting the cartographic origin of the data meant that the process of transforming the data into a useful representation of the city was still a long and labour intensive process. Figure 2. Image of the Western portion of the Edinburgh Model. The method adopted while constructing the Edinburgh Model was to maintain the automatically generated data set as the main context model and to selectively remove each building that required more detailed modelling. The process of removing buildings required the identification and selection of all polygon outlines that comprised a particular building. This action allowed the user to select a local origin and normal axis and then copy the selected data to a new file, the data being deleted from the context model at the same time. This new data file was saved complete with the reverse transformation matrix that allows it to be returned to its original world co-ordinates. In this manner the model could be progressively refined while still retaining its global structure at a wider scale. (Grant, 1994). This also provided a means through which the model could be structured at the level of the individual building and in turn make it easier to generate an interface to city information using building addresses as an index. One stated aim of the Edinburgh Modelling project was to construct the model in such a fashion that it would be possible to attribute the geometrical model with other, alpha numeric, properties relating to the building stock. This was proposed in order to address shortcomings discovered to be inherent in the construction of the prior model, that of Glasgow. Previously, attempts to link geometry and information had been thwarted by a mismatch in the granularity of the two sets of data. This was due to the capture of Glasgow's buildings in a block format as opposed to structuring the data at the level of individual buildings. Retrospective action to correct these problems in the Glasgow model were considered to be too close to the effort and manpower required to remodel the whole city and unfortunately any corrective action was never implemented. The methodology adopted in the creation of the Edinburgh Model was to develop an interface that allowed the user to move around the database and graphically browse the data using 2-D maps to select buildings and their associated records. Unlike a conventional database where each record might point to a building, this implementation takes each building as a graphical pointer to a record containing further data. Also provided is a query interface that enables a limited "search" functionality. This enabled a user to search for a record on the basis of context combined with a selection of Boolean, arithmetical or lexical operators. In turn this then allowed for a search pattern of, say, a mixture of use category, ownership, date, condition and location. The added value inherent in this approach is the bi-directional flow of information between the graphical content and the information content. Although never implemented on the wider geographic scale this research did highlight the potential for associating data with geometry and demonstrated the additionality that comes with correlating information attributes and the building stock. The worth of the concept was embodied in the conjunction of the geometrical description, its allied information content and the ability to interact with, and manipulate, the data. The disadvantages were mainly found in the particular hardware and software platform that only provided a limited development solution. This wasn't to say that the potential was not there but that the lack of any high level interface to the required functionality meant that the technical skills and attendant time scale required for the development of authoring and presentation tools was a negative factor. Equally, it had been proven that the most immediate audience for such a system would be mainly comprised of essentially non technical users and that the high cost graphical workstation with it's complex UNIX™ operating system would prove unsuitable Multimedia At the close of the previous section it could be seen that the evolution of the Edinburgh Model and an attendant desire to develop a close coupled system of graphics and associated information content could potentially yield a powerful and genuinely useful tool with wide ranging applications. However it was equally obvious that there were a number of impediments to the realisation of this ambition and it was fortunate that at this stage of development there was a parallel evolution in the capabilities of ubiquitous, low cost machines at the bottom end of the computational hardware scale. This step formed the basis of a new generation of IT products that encompass all aspects of diverse media types and hence fall into the generic category of Multimedia. This was a major advance in that for the first time developers had access to a platform that was essentially “good at everything” and that would encourage not just progress but also diversity. Since the Digital City is, by its very nature, a Multimedia conglomeration the conceptual fit between the desires of the concept and the capabilities of the technology potentially provides a direct mapping. Previously the vision of the Digital City had resided in the incorporation of diverse media types describing the city within an environment that had facilitated the interactive manipulation of geometry. The major disadvantage of this approach was with the computing infrastructure within which it had been conceived. The lack of any inbuilt high level functionality meant that every feature of the system had to be laboriously constructed from primitive elements. In turn this made for slow development cycles and a limited audience restricted to a similar computing system. The arrival of that strand of technology broadly labelled multimedia had the potential to provide a very different view of the problem and proved fundamental to the accelerated pace of development in terms of consumer hardware and software products. (Grant, 1995). The innate abilities of this platform meant that photography could replace cartography and any lack of traditional 3-D functionality was offset by the ability to replace geometry rendering with digital video and 2-D animation techniques. Far from proving restrictive this offered diversity in the character and format of the applications and provided new ways of capturing the context of the city. The inclusion of different graphical formats and the ease with which a homogeneous package could be built allowed greater insights through the conjunction and cross referencing of information which would have been difficult if not impossible within a conventional computing system. Figure 3. MultiMedia applications, replacing cartography with photography. This new found freedom makes it easier for the developer to concentrate on building applications for the Digital City; for example way finding to specific destinations, routes between sites of historic interest and many other arrangements of databases of buildings of interest. The major difficulty is still apparent in that the developer must bring all the required information to the application and that the worth of the enterprise is still measured in the quality and quantity of this data. The Digital City on the Internet An intriguing facet of the development of the Digital City is the manner in which applicable technologies became re-discovered or re-invented. The growth of the Internet has provided many opportunities not just as a medium but also in terms of access to knowledge, techniques and software products. The ready availability of these tools has allowed the rapid development of applications that would otherwise have taken many man-years to realise. This economy has made it possible to re-visit the Glasgow model, the dissemination of knowledge has improved the tools used to construct the data-set and the availability of a common denominator in terms of the delivery medium has made the product available to all. The final factor that has enabled the uptake of 3-D applications has been the increased availability of consumer grade graphics systems bringing affordable high level performance to almost every machine specification. It is interesting to observe that in the space of ten years what was once the pinnacle of graphical development has become the lowest common denominator in an era of new technology. With the introduction of VRML1 in November 1995 the sharing of 3-D models over the Internet became a reality. The arrival of VRML as a standard made the issue of application specific hardware and software a thing of the past. The entire philosophy behind the WWW is directed at platform independence by positioning browsers to form an abstraction layer over the native hardware. As the ubiquity of the internet grew and the quality and competence of the available standards and tools also developed then this medium offered an ideal mechanism for promoting the Digital City. If the ultimate goal had been to align the geometrical model with urban information sources then this concept can be envisioned as a “database of databases” within which the city model acts as the interface to further layers of information. This had previously proved difficult to realise due to the complications of collecting, collating and accessing such records but as the internet became populated with more and more city specific data then these sources became available just with the inclusion of the relevant URL. (Lindsay, 1999) Figure 4, The Glasgow Directory, the Digital City on the Internet Geometry processing was also facilitated as the VRML standard supported similar constructs as were employed in the original Glasgow model as well as providing collision detection, picking in 3D and the ability to in-line data to give a dynamic response to user interaction. The detail provided by the model geometry was no more complete than it had been in the original model but this was addressed by a process of data-mapping. This referred to the ability to choose within a list of categories which resulted in a series of markers being displayed with the model. Querying these markers then lead to the display of further relevant information in a graphical and textual format. This upper layer of information need not reside on the home site and may refer to hotel bookings, train timetables or even the local current weather situation. (ABACUS, 2001) Summary The development of the Glasgow Model was not driven by any one predetermined set of circumstances or strategy. Progress was marked by the availability of new hardware, newly introduced features in the suite of software used in its construction or even the addition of new data. This has meant that the structure and format of this area of the model was continually in a fluid state that suited this particular form of evolution. This proved to be a distinct advantage in that the model could exist in different forms, from the most basic of building to more detailed and realistic representations as shown above. With this facet of the project being revisited on an annual basis this iterative design approach gave increased insights into the requirements and potential of a larger scale model. The experienced gained from developing the Glasgow model made the construction of the Edinburgh model more detailed, more user friendly and more closely aligned with the concept of an urban information system. However despite the ease of initial data capture the detail required by the vision for the model merely curtailed the geographical scope of the model by concentrating effort at the level of individual buildings. The concept of populating the model with allied graphical and textual information was proven but in the realisation was difficult and time consuming through having to develop the attendant tool sets and interfaces within the model.