Abstract
Current construction worksite layout planning heavily relies on 2D paper media where the worksite planners sketch the future layout adjacent to their real environment. This traditional approach turns out to be ineffective and prone to error because only experienced and well-trained planners are able to generate the effective layout design with paper sketch. Augmented Reality (AR), as a new user interface technology, introduces a completely new perspective for construction worksite planning. This paper disucsses the related AR work and issues in construction and describes the concept and prototype of an AR-based construction planning tool, AR Planner with virtual elements sets and tangible interface. The focus of the paper is to identify and integrate worksite planning rules into the AR planner with the purpose of intelligently preventing potential planning errors and process inefficiency, thus maximizing the overall productivity. Future work includes refining and verifying AR Planner in realistic projects.
Keywords
Introduction
Worksite planning of construction resources includes the design of construction worksite as well as the optimization of resource logistics. Nowadays, construction projects are in need of short duration, which necessitate a fast and flexible strategy for planning construction worksites. Current construction worksite layout planning heavily relies on 2D paper media where the worksite planners sketch the future worksite layout adjacent to their real environment. This traditional approach turns out to be ineffective and prone to error because only experienced and well-trained planners are able to generate effective construction layout design with paper scratch. This also necessitates a long training and learning period to reach an effective level of expertise.
Generally many parties (construction managers, worksite layout planner, foreman, superintendents, construction workers, etc.) from different departments/trades of the construction company are involved in the planning process. Collaboration among parties from different levels involved in the planning process cannot be well supported due to the lack of effective visual rendering inherent in the traditional paper media.
When looking at the paper media, the user could not acquire spatial impression - only a two dimensional sketch. The spatial 3D view of the entire planned scene has to be mentally constructed in human's brain. Also, the method cannot accurately reflect planning logic (e.g. representation of the active areas and safety margins) which would otherwise increases the planning certainty to a higher level. No sense of immersion could be provided in 2-D sketch. So the spatial sense of the users is not well supported and thus the frequency of mistakes during the planning process increases.
A more intuitive cooperation and visualization tool/platform is therefore necessary to accommodate the needs from all these parties involved in the planning process to support effective collaboration and communication. Augmented Reality (AR) technology suggests a new solution to the noted issues. AR technology could create an environment where the additional information generated by a computer is inserted into the user's view of a real world scene. AR technology basically augments human's perception of real world entity by inserting relevant digital information into real environment. AR allows a user to work in a real world environment while visually receiving additional computer-generated or modeled information to support the task at hand. Augmented reality environments have been applied primarily in scientific visualization and gaming entertainment in the past. In recent years, it has been explored for goal-oriented human activities like surgery, training and collaborative work. AR has been investigated in its applications in various areas of manufacturing industry such as assembly (Reiners, D., Stricker, S., Klinker, G. & Müller, S., 1999), quality assurance (Curtis, D., Mizell, D., Gruenbaum, P. & Janin, A., 1999), industrial maintenance (Lipson, H., Shpitalni, M., Kimura, F. & Goncharenko, I., 1998), etc. Regarding its applications in planning, AR was also explored for interior furniture layout planning (Billinghurst, M., Kato, H. & Poupyrev I., 2001) and manufacturing planning (Doil, F., Schreiber, W., Alt, T. & Patron, C., 2003), both of which were developed based on MagicBook from HIT Lab in University of Washington (Billinghurst, M., Kato, H. & Poupyrev I., 2001).
The major advantage of the AR tool proposed in this paper could give construction worksite planners an easy and quick setup where even unskilled novice could easily capture the entire intelligent system. With the help of innovative, light-weight and inexpensive interaction and display devices, the AR interface enables the users to immerse themselves in a new reality which is augmented with computer-generated information. The introduction of AR as a new user interface enables a completely new approach to the development of construction worksite layout.
The proposed AR platform, called AR Planner, contributes to shortening and improving the quality of the construction worksite planning. AR Planner enables the construction worksite planner to place construction materials and machines/equipment, handling devices and the corresponding routing lines in the planned worksite. Therefore 3D-models are used to represent these objects in the virtual world. Users could move and position these items interactively to design and determine the configuration of the construction worksite.
Augmented Reality System Requirments
Augmented Reality, which appears in the literature usually in conjunction with the term Virtual Reality (VR), is a technology or an environment where the additional information generated by a computer is inserted into the user's view of real world scene (Milgram P. & Colquhoun, H., 1999; Azuma, R.T., 1997). AR can create an augmented workspace by inserting the virtual space in which we store and interact with digital contents into the physical space where people work. Such augmented workspace is realized by integrating the power and flexibility of computing environments with the comfort and familiarity of the traditional workspace (Wang, X. & Dunston, P.S., 2006). By exploiting people's visual and spatial skills, AR brings virtual information into the user's real world view rather than pushing the user into a completely computer-generated virtual world. This section discusses the characteristics of enabling technologies of AR systems and design issues encountered when building an AR system (Wang, X., Dunston, P.S., Skiniewski, M., 2004).
Media representation
Media representations refer to the format with which digital/virtual information augments the real world view, ranging from abstract to realistic (texts, indicators, tablet and screen, 2D image, 3D wireframe, 3D data, and 3D object). Realistic media representation like 3D virtual object conveys more detailed information than abstract ones. High fidelity representation may indirectly trigger usability problems associated with lag and low frame rate simply because computing power is limited. Virtual information doesn't have to be exclusively three-dimension in geometry. Text-based information such as project data (e.g., project schedules, building codes, etc.) can be made available as 2D windows on the world (Feiner, S., MacIntyre, B., Haupt, M., & Solomon, E., 1993) or on virtual sheets of paper or panels attached to real walls (Klinker, G., Stricker, D., & Reiners, D., 2001). In addition, two or more types of media representation (e.g., texts + 3D object) presented together on a single display, which is termed as hybrid representation. A limited observation of AR systems reveals many of the applications adopted the hybrid representation.
Input device
Virtual information is manipulated via input device. Most of the input devices used in VR environments can also be applied in AR systems. Use of appropriate input metaphor in user interface design is generally considered to be good practice, since they can effectively exploit users' prior knowledge to increase familiarity of action, procedures, and concepts (Neale, D.C. & Carroll, J.M., 1997). A poor input metaphor may create a number of problems for the user. Input device could be as simple as 2D input such as mice and keyboard and could be as sophisticated as embodied input mechanism. More information regarding input devices can be found in (Gabbard, J., 1997). Voice input, a more direct, natural form of interaction, can be achieved as an extra input to increase input capability of the whole system. Also, haptic devices can also be input module through devices like Phantom (Massie, T.H. & Salisbury, J.K., 1994).
Display device
The display can be generally classified as visual display (head-mounted displays), acoustic display (3D localized sound systems), and tactile display (force feedback devices). Visual displays are the most popular one used in AR systems with other types of displays as supplement. Visual displays are used to present augmented scenes and two basic merging technologies (Azuma, R.T., 1997) exist: video-based see through and optical see-through. While video-based see-through merges live video streams with computer-generated graphics and displays the result on the display, optical see-through generates an optical image of computer-generated graphics, which appears within the real environment. Poor lighting conditions could impose difficulties on the use of optical see-through displays because the user sees the real environment directly through the display. Video-based see-through merging typically relies on the use of fiducial markers at known locations, which is not convenient to place in outdoor settings, therefore, video-based see-through displays is not suitable for outdoor applications. Video-based see-through also demands greater amounts of computation power and other resources, which are limited in outdoor applications.
Tracking technology
Tracking, also called position and orientation tracking, is used where the orientation and the position of a real physical object is required. The biggest obstacle to building effective AR systems is the requirement of accurate, long-range sensors and trackers that report the locations of the user and the surrounding objects in the environment (Azuma, R.T., 1997). Applewhite (Applewhite, H., 1993) presented an excellent discussion of tracker evaluation by means of a framework for suitability. A more recent survey on tracking technologies for virtual environments could be found in (Rolland, J., Davis, L., & Baillot, Y., 2001). For detailed comparison of tracking technologies, readers are referred to the surveys in (Applewhite, H., 1993; Ferrin, F.J., 1991; Meyer, K., Applewhite, H.L. & Biocca, F.A., 1992).
Computing device
Computing power is a key requirement in producing a feasible AR system. The capability of available computing power determines the format of media representation that could be used in AR systems. 3D high fidelity objects are computationally-intensive so that devices with limited power cannot bear. More schematic representations such as texts and indicators are less computationally-demanding so that ordinary computers can handle.
The requirements for computing power are also determined by the environment where AR systems are used. A major difference between indoor and outdoor AR systems is the amount of infrastructure and resources available. An extensive computing infrastructure can be available indoors, with a wider range of computing devices that may not be practical outdoors. Outdoor settings impose constraints on the portability of devices and associated power supply. Development in laptops, wearable computers, PDA and other upcoming portable devices may provide improvements of which outdoor AR systems can leverage off.
Related Work in Augmented Reality
AR technology is envisioned to improve the current state-of-the-art of architecture visualization, design process, building construction processes and engineering management systems. Design and construction industries involve the generation of great amounts of data and information that must be accessed by numerous parties, in numerous locations, and under varied conditions. Specifically, the need for good operator training and the access to large amounts of engineering and management information in construction industry creates conditions making use of AR techniques most promising by involving construction personnel into the augmented workspace (Wang, X., Dunston, P.S., Skiniewski, M., 2004). Human-computer interfaces that blend a view of an existing workspace with relevant design or field information should be an attractive class of technology for design and construction industries.
Recent advances in computer interface and hardware power have fostered AR prototypes for various design and planning applications. However, most of these lab-based prototypes were investigated by computer science/engineering societies which picked design and construction as testing areas for proof-of-concept. Due to the lack of in-depth understanding for design and construction practices, these efforts could hardly progress beyond the lab-phase to eventually become a usable system for field operations. Although the idea to use AR for construction, maintenance, and repair dates back to the early 1996's (Webster, A., Feiner, S., MacIntyre, B., Massie, W. & Krueger, T., 1996) and AR has actually matured from a pure research field into certain practical industrial applications, until now it has not been implemented as a real product in architectural design and construction. On the contrary, design and construction communities apparently have the knowledge of the operations/tasks that AR can potentially enhance as well as the motivation to bring in this new technology for improving the current practices. Regardless the noted efforts, both communities have been moving very slowly toward the potential applications of this cutting-edge technology.
Azuma (Azuma, R.T., 1997; Azuma, R.T., Baillot, Y., Behringer, R., Feiner, S., Julier, S. & MacIntyre, B., 2001) implemented two consecutive reviews on comprehensive topics for the AR applications in many areas. Neither of the reviews extensively addressed the AR applications in design and construction. The review presented in this section does not duplicate the content of Azuma's reviews and tend to specifically focus on up-to-date AR work related to design and construction arenas (Wang, X., 2007). This review consolidates existing information from an extensive bibliography of papers in the field of AR applications in architecture, design, construction, maintenance, assembly, etc. This review does not claim to have exhaustive coverage, but include the representative works. This review provides a good beginning point for guiding and encouraging the researchers who are interested in starting research in this exciting multidisciplinary area. This review also discusses the issues and challenges encountered when building AR systems for design and construction.
This review organizes the state-of-the-art AR lab-based development in design and planning areas. Lab-based applications include recognized prototype-based efforts in various applications in design and construction domains which develop AR systems to improve their current practice.
Most of noted AR applications in design and planning embraced the ARToolkit (2007) as their platform because it is simple to implement. Since ARToolkit is mostly suitable for small-scale working space (e.g., tabletop), the resulting AR systems can be regarded as Tabletop AR systems. One of the applications of AR in the early phases of a design is for sketch which is a rapid and fuzzy embodiment of the architectural discussion. The sketchand+ system (Seichter, H., 2003) is an experimental prototype to make a first attempt to use AR for the early architectural design stages, which could have an impact for the quality of a design-process. This AR prototype utilized a scribbling interface through the metaphor of a digitizer tablet and the virtual response is a 3D sketch. Sketchand+ demonstrated that collaborative tangible interaction with models could be more appropriate in early design investigation. Another example is an AR prototype implemented by Aliakseyeu et al. (Aliakseyeu, D., Martens, J. & Rauterberg, M., 2006) for architectural design based on a thorough analysis of the characteristics and requirements of the early architectural design stage. The system has three basic interaction elements: the brick elements, a digitizer tablet with a digital pen, and the enhanced paper prop. The prototype system can assist in creating and editing sketches, which preserves the naturalness of the traditional way of sketching.
Reviewing urban design proposals are different to that of architectural design. Generally the issues are not of detailed design but rather understanding space and spatial features. Exploring the relationship of human and the city is of major influence for urban design. BenchWorks (Seichter, H., 2004), a next generation of sketchand+, was developed as an AR prototype for analyzing representational design in an urban design scale, which focused on techniques and devices needed to create 3D models for urban design. The system was designed as a workbench and combined optical tracking (use of ARToolkit) with magnetic tracking. A 1.5×1.0m2 digital whiteboard was also mounted horizontally as working surface, providing a much larger interactive shared space allowing new interaction techniques. The traditional pen, paper and eraser metaphor creates the link between the traditional and virtual tools. The way a user can design in BenchWorks is by representation of void and non-void space of the city, adding volumes and making notes in a tangible manner. The user can be aided in understanding the provided 3D information more easily and therefore can speed up the time for decision-making.
Another example of tabletop AR systems for urban planning is ARTHUR project (Broll, W., Lindt, I., Ohlenburg, J., Wittkämper, M., Yuan, C., Novotny, T., Fatah gen Schieck, A., Mottram, C. & Strothmann, A., 2004) where optical see-through AR displays were used together with decision support software for architectural and urban design. ARTHUR involved a critical characteristic of a first-person perspective in which the scale of rendition of the design model follows many of the conventions that designers are used to. Simulating pedestrian movement inside urban design allowed easy scale recognition as well as smooth transition from a first-person to a third-person understanding of the design.
Dunston and Wang (Dunston, P.S., Wang, X., Billinghurst, M., & Hampson, B., 2002; Dunston P.S. & Wang, X., 2005) developed an AR system called Augmented Reality computer-aided-drawing (AR CAD) for individual mechanical design detailing. This AR system allows the user to visualize the virtual piping skid design that looks floating on a physical real tracking marker via head-mounted display (HMD). Detailed design is modeled in AutoCAD® and then sent to the customized AR program for rendering and visualizing in a real time manner. The entire virtual model can be manipulated (observed from different perspectives) through physical manipulation of tracking marker. Provided benefits include enhanced spatial cognition and perception of piping designs.
MIXDesign (Dias, J.M.S., Santos, P. & Diniz, N., 2002) provides a Mixed Reality system oriented towards tasks in architectural design. The system explores tangible interfaces using ARToolkit patterns on a paddle and gestures. Yabuki et al. (Yabuki, N., Machinaka, H. & Li, Z., 2006) applied the AR technology to erection planning of steel girder bridges based on ARToolkit. They developed a simple prototype for checking interference among design objects such as steel girders and surrounding objects such as buildings in order to support erection planning.
An increasingly common use of computers is to support communication and collaboration. Many of the applications proposed for AR are naturally collaborative activities and these systems are collaborative Augmented Reality (Billinghurst, M. & Kato, H., 1999). Other collaborative activities, especially those involving design and visualization of 3D structures/buildings, can benefit from having multiple people simultaneously view, discuss and interact with the virtual 3D models. As computer-based collaborative tools become more common, the human-computer interface is giving way to the notion of human-human interfacing mediated by computers. This emphasis on the mediation role of computers adds new technical challenges to the design of human-computer interfaces. AR offers new potentials for mediating human-human interactions/communications for the entire life cycle of the engineered facility because the proper blending of the real and virtual and the attendant interaction metaphors can be tailored to enhance group-oriented decision-making for specific life cycle activities. There are several noted efforts towards collaborative AR systems in design and planning. For instance, Wang et al. (Wang, X., Shin, D. & Dunston, P.S., 2003) developed an intuitive mixed environment called Mixed Reality-based collaborative virtual environment (MRCVE) to support the collaboration and design spatial comprehension in collaborative design review sessions for mechanical contracting. The environment could be face-to-face manner or distributed over network.
BUILD-IT (Rauterberg, M., Fjeld, M., Krueger, H., Bichsel, M., Leonhardt, U. & Meier, M., 1997) is a multi-user planning tool using a 2D projection on a table top. A video camera is used to track manipulations of a small, specialized brick that can be used as a “universal interaction handler”. A second, vertical projection screen behind the table provides a 3D view of the virtual scene.
Issue and Challenges
To design and implement AR systems for industrial problems in design and construction arenas, researchers and developers face challenges in extraction of industrial domain knowledge, preparation of reality model, and technological limitations. Besides these technological issues, social issues such as technology transfer need to be addressed as well.
Preparation of reality model
In order to accurately register virtual information into real environment, AR systems need to obtain a precise description of the real environment: a reality model (Klinker, G., Stricker, D. & Reiners, D., 2001). A carefully controlled environment could define a relatively accurate reality model. One of the important issues in developing AR systems for design and construction is to apply a systematic method to create such reality model.
Extraction of industrial domain knowledge
As a facility is constructed, there creates large amounts of design and as-built information in wide range of specialty services such as HVAC, mechanical, electrical, structural, etc. Unfortunately much information is currently represented as 2D plots rather than 3D models in design and construction industries. There apparently exists a lack of well-organized integrated 3D database supporting information source extraction that could be readily used by AR technology due to the fact that design and construction parties are not committed in this regard. For example, the architectural application of “seeing into the walls” assumes that the AR system has a database for the locations of all the pipes, wires and other hidden objects within the inspected facility area. Construction contractors and subcontractors tend to maintain their own 2D systems of information relevant to their specialty services rather than an all-inclusive large-sized 3D project model. Therefore, significant modeling (2D to 3D) and integration efforts are required to construct an AR-compatible database with appropriate formats. Furthermore, a specialized AR system geared towards one or multiple specialty services only need to display the most relevant information to the operation in hand instead of overloading the user's attention by inserting too much information. Therefore, it is also necessary to develop an automatic and intelligent domain information extraction mechanisms (e.g., relevance, level of details) to always provide most critical and relevant information with respect to the task at hand, which maximizes the essential advantage of AR concept. Thus, modeling, integration, and extracting efforts may be required and should be taken into consideration when building an AR application.
Technological limitations
Technological limitations remain the major obstacle for AR systems. For instance, AR requires highly accurate trackers because even tiny tracker errors can cause noticeable mis-registrations between real and virtual objects. The biggest obstacle to building effective AR systems is the requirement of accurate, long-range sensors and trackers that report the locations of the user and the surrounding objects in the environment. As an sucessful example for large-scale tracking, Abderrahim, M., Garcia, E., Diez, R. & Balaguer, C. (2005) developed a tracking infrastructure which embeds the miniature positioning and communication instruments into the compulsory safety helmet required for all workers in construction sites. According to this, workers and machines' positions are known in each instant and sent to a monitoring station. Future tracking systems that can meet the stringent requirements of AR will probably be hybrid systems, such as a combination of inertial and optical technologies (Azuma, R.T., 1997). The advancement of tracking technology heavily relies on the both industrial and academic efforts in hardware domain. Besides accurate and long-range tracking, high-quality and real-time rendering (Reiners, D., 1994; Rohlf, J. & Helman, J., 1994) is essential to an AR system. Achieving such rendering effects could come from two methods. Simple sensors-based tracking could be run on a wearable computing device and high-quality rendering of a complex virtual model could be achieved on a stationary graphics supercomputer (Starner, T., Mann, S., Rhodes, B., Levine, J., Healey, J., Kirsch, D., Picard, R.W. & Pentland, A., 1997). Likewise, a resource-intensive tracking algorithm based on a simple reality model could run on a stationary supercomputer and simple virtual information such as texts could then be rendered by a wearable computing device.
With little or no prior knowledge about the surrounding real environment, occlusion detection becomes a critical issue in AR system because the digital representations of the same structure hidden into and outstanding out of a real object may present the same combined view to users. Such occlusion errors and confusions could easily ruin the feeling of presence the user might otherwise experience. This issue is now well explored by research communities in computer science and cognitive psychology. However, some researchers in design and construction also started to explore potential solutions. For example, Behzadan and Kamat (Behzadan, H. & Kamat, V.R., 2005; Behzadan, H. & Kamat, V.R., 2006). proposed a solution which uses a combination of rapid geometric modeling of the surrounding environment or other depth sensing techniques (e.g. stereo cameras) and utilizing the graphics processor's z-buffer to draw the appropriate set of pixels in each composite AR frame. In other words, if this depth of real objects is greater than the depth of virtual object(s) for a given view, the real object does not occlude any virtual objects. In the opposite set of circumstances, appropriate corrections should be made to user's view to take into account the existence of an occluding real object.
Social concerns
There is a lack of motivation for AR technology transfer. It is well-known that construction people are conservative and reluctant to change much especially in the aspect of moving toward new technology. Whether AR is truly a cost-effective solution in its proposed applications has yet to be determined. Much research can be done to prove to the design and construction practitioners about the feasibility and profitability of applying AR system.
Social concerns should not be ignored during attempts to move AR out of the research lab and into the hands of real users (Azuma, R.T., 1997). For example, if workers perceive lasers to be a health risk, they may refuse to use an AR system with lasers as the tracker, even if those lasers are eye safe. Another important factor is whether or not the technology is perceived as a threat to jobs, as a replacement for workers. This will not be a big problem for AR because it is intended as a tool to make the user's job easier, rather than something that completely replaces the human worker.
Prorototype System Architecture
The application of the AR system is to optimize construction site layouts. The traditional way for the on-site planning crew is to discuss and improve site layouts using paper-based plans. This mehtod can be enriched by superimposing 3D objects onto an existing paper-plan via this AR system. With the visual combination of virtual planning results and real environment, planners can validate the generated layout solutions/proposals by easily comparing different geometries. This enables the planners to validate the planning results fast and to change the plans according to the results of the comparison. This process improves the planning quality and therefore can avoid replanning activities. An example that shows the benefit of the described system, is the overlay of an existing, virtual model of heavy construction equipment (e.g., backhoe, loaders, etc.) with the corresponding earth-moving construction scenarios. Possible collisions of construction equipment and overall equipment fleet productivity can eas-ily be identified and compared.
The described AR system for planning tasks incorporates three main components: rendering and visualization, rules engine, and simulation, as shown in Fig. 1.
System setup of AR planner
The prototypical realization of the AR system supports the following features: The basic task of the system is the layout design of construction worksite. With a tangible user interface the user has the possibility to place and move the objects in a virtual model of a worksite. It permits even non-experts to work with this tool without prior training. Collaboration functionality allows collaborative work among various parties. In a collaborative planning session, every user wears a head mounted display and gets a tangible interface. For each head mounted display, a PC is used to render the scene from the corresponding viewpoint. Due to the usage of different interfaces the activities of each user during the planning process could be stored.
AR planner allows animations aside from static objects to show the behavior of the dynamic components. For example it is possible to show the movement of a truck in order to visualize its active space or to detect possible crashes with its environment. If a crane cannot reach to a desired area within its range capability, such shortfall could be automatically identified, visualized, and highlighted.
Users could plan a virtual construction worksite straightforward with planning rules built in the system to ensure the correctness of the planning alternatives. These may then be imported into an AR scene and displayed as seen in Fig. 2 by AR Planner.
Virtual construction site planning on top of a real ARTag tracking marker
The exact touching of the virtual element with the paddle is difficult to achieve. Accurate placements, movements and alignments of the virtual elements in the AR scene are therefore very difficult to realize. If the planned worksite consists of a high number of 3D models which are densely placed, it is not easy to select and move a single virtual element without interfering other adjacent virtual elements.
The computing unit used is a Pentium 4 PCs with a 1.6 GHz processor with 512 MB RAM and a GeForce II Ultra graphics card. Display device used is a head mounted display with a color video camera (real environment sensor) attached. The input mechanism uses the software Magic Paddle from HIT Lab, which is based on the ARToolKit software (2007). A high-performance, lightweight tablet PC was used as the mobile computing unit that consists of Local Database and AR program. Existing imagery and data about equipment can be collected in order to compile a comprehensive, equipment-specific database of information. The way that the system adds labels is that the fiducial beside an object or component identify the component. As the component is identified, the system sends a query to the Local Database, which returns any labels matching the object's features. The AR program is the core technical component for rendering pipeline for the whole system, just like the brains of the system. Visual C++, OpenGL, and VRML are used as the development environment.
Display system
The display system (see Fig. 3) used in the prototype is a lightweight head-mounted display (HMD). The HMD is the V hi-Res800™ (PC) with the InterTrax2™ head tracker device produced by CYBERMIND Interactive Nederland Inc. which is a professional full-color SVGA immersive HMD for advanced research. The cameras act as the eyes of users and their position and orientation are tracked by special tracker.
The head-mounted display (image is from CYBERMIND Interactive Nederland Inc.)
The tracking approach adopted by the system makes use of the ARTag system (Fiala, M., 2004), a 2D fiducial marker and computer vision system for Augmented Reality. Fiducial marker systems consist of patterns that are mounted in the environment and automatically detected in digital images using an accompanying detection algorithm. ARTag uses digital coding theory to get a very low false positive and inter-marker confusion rate with a small required marker size, employing an edge linking method to give robust immunity to lighting effects and occlusion.
The fiducial markers used in the ARTag system are bi-tonal planar patterns that consist of a square border and an interior region filled with a 6×6 grid (a digital 36 bit word) of black or white cells. Fig. 4a shows some examples of ARTag markers. Quadrilateral contours are located in the image which may belong to the outside border of a marker. They are found in ARTag with an edge-based method, and edge pixels are threshold and linked into segments, which are in turn grouped into “squads.” The four corners of the squad boundary are used to create a homography mapping to sample the marker interior. Then the 36-bit word can be extracted from a camera image of the marker once the boundary is determined. The digital word contains a unique ID number.
(a) Bi-tonal planar ARTag marker patterns consisting of a square border and a 6×6 interior Grid of cells; (b) ARTag marker detection under a Large occlusion (Wang, X. & Dunston, P.S., 2006)
A major advantage to ARTag's edge-based approach over similar fiducial approaches is the ability to still detect marker outlines in the presence of an occlusion. In Fig. 4b, one marker has a corner covered but is still detected by heuristics of line segments that almost meet.
Different geometries are used to represent different kinds of components. The virtual objects can also automatically change in size — even in their proportion to the fiducial marker — based on the distance between the tag and the user. While the user views the marker from far away, the tag appears large so that the information can be seen clearly. When the user leans in close to the tag, the tag appears smaller, so as not to obstruct the user's view of real object details. Fig. 5 is the large-scale marker where virtual constructon site is displayed.
ARTag tracking marker
The virtual element sets actually provide the virtual elements that users could manipulate and interact with in the virtual construction site and they are also the elements that need to be planned. The components of the virtual element sets are 3D models of construction materials, machines, devices, equipment, and transport systems, which are created by conventional 3D CAD modeling tools. The user (equipped with a head mounted display) finds himself/herself in a scene such as in Fig. 6. The components such as machines and devices of the virtual element sets are stored in a sort of book (see Fig. 6). Users can select, place, and manipulate each virtual element of the virtual element sets in an easy and intuitive way using a tangible paddle. Using a paddle, modules from the virtual element sets can be brought into the existing virtual construction worksite. To be used in construction worksite planning, virtual element sets must be flexibly designed in order to enable easy adaptation to different construction worksite scenarios.
Virtual element set: virtual 3D models of resources overlaid on real tracking marker
A database is to store and manage the 3D models in virtual element sets, the datasheet for the relevant description of each element and the planning rules. During the planning process the database also manages the updates of the scene. Each time a new virtual element is modeled, it is stored in the database with a datasheet. This datasheet contains information about the properties of the virtual element such as active areas, safety margins, etc. Furthermore, attributes for the planning rules could also be set in the database such as capacity ranges, connection points, etc.
Planning rules
¶The integration of planning rules allows a computer-supported planning process with a high degree of faultlessness. Worksite planning rules are identified and integrated into AR Planner to intelligently assist planners in preventing possible planning errors and process inefficiency and thus maximize the overall productivity. The following major categories of planning rules are established to be integrated into AR Planner:
Spatial constraints
Static or dynamic site layout planning (Choo, H.Y. & Tommelein, I., 1999; Hegazy, T. & Elbeltagi, E., 1999): The site layout planning algorithms automate the allocation of macro-level spaces, which are the coarse spaces located at the construction worksite, based on user-defined qualitative adjacency constraints such as close or far between defined spaces. The major disadvantage of the mechanisms implemented in site layout planning is that it cannot generate micro-level work spaces. Examples of considerations in spatial constraints are listed below:
Active areas and safety margins: for example, the capability range of one crane could be visualized based on the pre-determined value from the equipment specification. Volume of workspaces: the minimal amount of workspace which the worker needs has to be considered in advance. Composition and setup rules for machines: for example, certain on-site machines such as rebar bender need to be setup at a location where the path of materials flow would be optimized. The minimum and maximum distances among machines and construction equipment should also be adhered to.
Path planning (Morad, A., Cleveland, A., Beliveau, Y., Francisco, V. & Dixit, S., 1992)
Path-planning approach involves materials flow rules, level of access to materials, work sequence, etc. This approach would be integrated into AR Planner for material flow paths planning. The path-planning approach usually focus on providing a problem-solving strategy in a reasonable amount of time by generating a collision-free path to move a specified object from one position and orientation in the 3D computer model to another position and orientation using available manipulation mechanisms. For example, in a detailed representation of construction site, path planning will generate the specific material flow paths required to perform critical operations.
Environmental constraints
Environmental constraints such as temperature, humidity, wetland, etc. control the planning of construction worksite to a significant extent. An interdisciplinary team composed of engineers, ecologists, aquatic and wildlife biologists, and architectural and landscape consultants need to evaluate every aspect of the construction regard to its environmental impact.
Resource allocation optimization strategy
Indicators such as throughput, manpower requirements, inventory levels, routing behavior effects could be determined dynamically in a real time manner in an AR environment. The interactive, true 3D animation and statistical results could be used to allocate resources appropriately, which allows justifying costs associated with work-in-process inventory, labor scheduling, equipment failure, and capacity planning.
Space-scheduling rules can also be enabled (Choo, H.Y. & Tommelein, I., 1999)
The space-scheduling approach focuses on modeling the different types of work spaces required by construction activities. However, space-scheduling assumes that the user specifies the geometric attributes of the project-specific activity space requirements. The algorithms developed in space scheduling mainly focus on creating a schedule to eliminate spatial conflicts once the user defines all of the micro-level spaces (Akinci, B., Fischer, M. & Kunz, J., 2002). The consideration of these planning rules is implemented using different types of attributes with a set of corresponding parameters. The attributes can be assigned to each object/machine which is available only for a specific space and/or scheduel during the planning process. Only those objects which have an accordant attribute can be placed to a single area. For instance, all the materials, tools, and equipment related to construct the foundation 1 should be located in the area of X.
Different graphical representations are used in the system. The representation of active areas, safety margins and workspaces is realized using bounding boxes. AR Planner uses grids to divide the virtual construction ground/site for different level of details: coarse grids for rough planning and fine grids for detailed planning. The entire model containing the construction worksite generated from AR Planner can be exported to VRML. Then the whole scene can be imported in 3D-realtime rendering systems (e.g., CAVE, workbench, head-mounted-display, etc.) for a higher degree of immersion. In the real-time environment additional interaction between user and 3D-scene can be implemented like interactive walkthroughs or the visualisation of information flows.
Simulation
Apart from static layout design, the design and the optimization of materials (resources) flow path/access and logistics is extremely important. In this case the process simulation is used for designing material flows and logistics. The analytical results could be displayed in customizable numerical tables, bar graphs, pie charts, histograms, and time series graphs. Certainly ergonomic layout of the manual workplaces is also crucial for the construction crew. Elements such as tools and material can be positioned within an optimized range for the construction worker. There are two types of simulation that could be implemented on top of the rendered planned virtual construction worksite : (1) Simulation within the AR-System, this system allows to use animations aside from static objects to show the behaviour of the machines. For example it is possible to show the movement of an excavator in his later environment in order to visualise its active space or to detect possible crashes with its environment; and (2) Simulation using external softwares, the results of the layout design are stored in the database. They contain information about the position of the objects and their transmission behaviour. This data can be used by discrete construction simulation tools like Microcyclone or Vitascope to simulate the construction site operations. These analytical tools could assist in identifying and quantifying the impact of bottlenecks on planned worksite layout as well as predicting expected producivity.
Conclusions and Future Work
This paper introduced a new concept of an AR-based construction worksite planning system, AR Planner and presented the preliminary work associated with AR Planner. The prototype described in this paper shows that the use of AR-based element sets in the construction worksite planning is beneficial. The efficiency of the system mainly depends on the number and quality of construction elements available in the virtual element sets. The implementation of planning rules assists the user and prevents possible errors which normally occur during the planning phases.
Future work includes refining the AR prototype and verifying the AR system in a realistic project. This prototype will be improve/refined further to realize all the itemized functions described in this paper. Furthermore, experimental validation will be implemented in future to demonstrate the effectiveness of the AR system, especially the benefits for novice planners. The benchmark method compared will be paper media sketch. The next step of the project is to invite human subjects into an experiment where they will be asked to use the AR system, and paper-media specifications to perform a routine constructio site playout planning task. Their performance measured by various indicators will be recorded and compared. After the configuration is designed with the AR-based element sets, it can be verified in the real environment using an outdoor wearable AR-system. This system would blend planned layout of all relevant virtual objects/components into the real construction worksite. In general the tracking of the users is the main problem.
Another powerful application of the developed system is the visualization of virtual materials or components in the real construction site. After the configuration is designed with the AR-based construction set, it can be verified in the real environment using a wearable outdoor AR-system. This system blends the before designed site layout into the real construction site.»
Footnotes
7. Acknowledgement
The described prototype is implemented using the new version of the ARTag system by Dr. Mark Fiala in National Research Council Canada, Institute for Information Technology.