INFORMATION AND COMMUNICATION TECHNOLOGIES IMPROVING EFFICIENCIES WAYFINDING SWARM CREATURES EXPLORING THE 3D DYNAMIC VIRTUAL WORLDS

Transcription

1 INFORMATION AND COMMUNICATION TECHNOLOGIES IMPROVING EFFICIENCIES Refereed Paper WAYFINDING SWARM CREATURES EXPLORING THE 3D DYNAMIC VIRTUAL WORLDS University of Sydney, Australia Mary Lou Maher University of Sydney, Australia ABSTRACT The use of 3D virtual worlds in the construction industry is not limited to simulation of building models. Whilst in a 3D virtual world, designers can view their work from various viewpoints by moving around. However explorations within these environments are not simple affairs due to the absence of many sensorial stimuli that exist in the physical world. Hence there needs to be a tool aiding designers to explore the world. We present a wayfinding aid based on a swarm model. The wayfinding creatures produce dynamic trails leading to desired destinations. We describe the swarm rules developed to create such behaviour in wayfinding creatures. Keywords: Virtual Environments, Wayfinding, Swarm Intelligence, Construction Industry Clients Driving Innovation: Moving Ideas into Practice (12-14 March 2006) 1

2 1.0 INTRODUCTION The increased use of Information and Communications Technology (ICT) in the construction industry has enabled a more efficient communication between different stake holders. ICT provides an effective communication channel between main contractors and sub-contractors whilst maintaining information between them using automated information management systems. ICT also allows clients and developers to quickly visualise models being built in 3D, and enables them to see the effects of any alteration. Figure 1 Simulated building Figure 2 Virtual environment as a design tool The use of 3D virtual worlds in the construction industry is not limited to a simulation of building models as shown in Figure 1. These worlds enable designers to be in their designs. Whilst in a 3D virtual world, designers can view their work from various viewpoints. They can also design while they are in the worlds as shown in Figure 2. In order to design, designers first have to arrive at a particular location in the world which gives them a better view of the model being considered. However explorations within these environments are not simple affairs (Darken and Sibert 2001) due to the absence of many sensorial stimuli that exist in the physical world (Chittaro and Burigat 2004). Hence there needs to be a tool aiding designers to explore the world (Darken and Paterson 2001). Our wayfinding aid consists of an interface agent and swarm based path forming creatures. The interface agent acts as a mediator between a user and swarm creatures. It identifies the user s requested target and sends out swarm creatures to locate it. Swarm creatures are sent out into the virtual environment to search for the requested target, and, once found, to return back to the user whilst creating a path. This paper focuses on the use of swarm creatures to generate paths in a dynamically changing environment. 2.0 SWARM INTELLIGENCE Bonabeau et. al. (1999) define swarm intelligence as any attempt to design algorithms or distributed problem-solving devices inspired by the collective behaviour of social insect colonies and other animal societies. As an individual, these insects do not possess enough intelligence to survive. However as a colony of insects they find food and shelter to sustain their existence. Social insects have three traits that make them successful: flexibility, robustness, and self-organisation. A colony is flexible in that it can adapt to the changing environment. It is robust in that even when individuals fail, it can continue performing its tasks. Each individual acts autonomously without intervention from a controlling body. There are many swarm models being used for various purposes (Kennedy and Ebergart 2001). For this research, the ant foraging model of swarm is chosen to be a base swarm model for wayfinding in dynamic virtual environments. Clients Driving Innovation: Moving Ideas into Practice (12-14 March 2006) 2

3 2.1 ANT FORAGING MODEL Wayfinding swarm creatures exploring the 3D dynamic virtual worlds Ants perform complex tasks even though each ant is only governed by simple behavioural rules. For example, harvester ants illustrate this complex behaviour in locating food. They find the shortest path to food sources while prioritising food sources depending on the distance and the ease of access. Different species of ants communicate differently when foraging. Some species communicate directly to each other while others use the environment as their communication medium. We focus on the latter form of communication which is termed stigmergy. Stigmergy allows each ant to modify the environment to communicate to others about the location of a food source when foraging for food. Ants use chemical droppings, called pheromones, to indicate trails between the nest and food sources. Pheromones are chemical compositions which evaporate over time. Consequently a trail once laid will dissipate when it is not reinforced by further pheromone markings. When a trail is used frequently, pheromones will accumulate leading to a higher concentration. The following diagrams illustrate the ant foraging model based on stigmergic communication. Figure 3 Ants start looking for food Figure 4 Food located Figure 5 Return while dropping pheromones Figure 6 The shorter path leads to food faster Figure 7 Ants attracted to a stronger scent of pheromones Figure 8 Following the same path to food Initially the ants randomly search for food (Figures 3). An ant following a shorter path arrives before another ant that followed a longer path (Figure 4), and returns sooner to the nest while dropping pheromones along the trail (Figures 5-6). Ants are attracted to a path with a strong scent of pheromones (Figure 7). Pheromones accumulate on a path well travelled by ants while pheromones on a less travelled path will dissipate (Figure 8). The ant foraging model outlined above can be expressed by a simplified set of rules like Resnick s algorithms (Resnick 1995). Rule 1 shows Resnick s ant foraging behaviour algorithm. Clients Driving Innovation: Moving Ideas into Practice (12-14 March 2006) 3

4 Rule 1 Resnick s ant (Resnick 1995) 1. Looking for food * If pheromone trail is weak then wander * Else move towards increasing concentration 2. Acquiring food * If at food then a. Pick it up b. Turn around c. Start laying pheromone trail 3. Returning to nest * Deposit pheromone * Decrease amount of food available 4. Depositing food * If at nest then a. Deposit food b. Stop laying pheromone trail c. Turn around 5. Repeat forever Resnick's ant, as shown in Rule 1, exhibits a similar pattern of behaviour as real ants. The ants wander around randomly until either food or a pheromone trail is found. These ants will continue to travel until they find food. When they return to the nest with food, they drop it and then go back to where the food was found. 2.2 ADAPTIVE BEHAVIOUR OF ANT FORAGING MODEL The ant foraging model allows the ants to adapt to the changing environment. A path is established (Figure 9) as shown in Figures 3 to 8. On this path, an object is introduced causing the ants to stop (Figure 10). Initially the ants try to move around the object randomly (Figure 11). Once again with the use of pheromones, a shorter path around the object will be more concentrated encouraging more ants to follow that path (Figure 12). Figure 7 Following the shortest path to food Figure 8 Introducing an obstacle on the path Figure 9 Moving around the obstacle Figure 10 A new path formed 3.0 SWARM BASED WAYFINDING AID The ant foraging behaviour model has characteristics which are ideal for a wayfinding application (Yoon and Maher 2005). It not only allows a path to be generated but also allows Clients Driving Innovation: Moving Ideas into Practice (12-14 March 2006) 4

5 the established path to adapt to subsequent changes made in the environment. This is vital as when virtual environments are used to design buildings, objects will be created, deleted, and moved to visually assess the design outcomes. The wayfinding swarm rules are presented in Rules 2-4. These rules define how each individual creature makes the decision about a local move. The rules also define what each creature senses and how it acts. These rules have been successfully implemented and simulated in a 2D environment. The implementation demonstrates that the swarm creatures are able to locate a particular target in a virtual environment. A path is formed between the target and home once the target is located. The generated path then can be adapted to subsequent changes made in the environment. 3.1 WAYFINDING SWARM RULES Rule 2 Overall behaviour rule Rule wayfinding_creature_behavior repeat Explore_World until Target_located Return Home The overall rule for the swarm creatures is shown in Rule 2. Rules 3-4 mention attractants and repellents. Both the attractants and the repellents are electronic pheromones dropped by the swarm creatures as they move about in the world. The attractants are dropped by creatures returning home once they find the target. The wayfinding creatures also drop repellents to mark visited spaces while exploring. Hence when they sense repellents in adjacent locations, they are encouraged to move to unexplored spaces. Because the repellents evaporate, the creatures are encouraged to explore the spaces previously visited in due time. Rule 3 describes how a wayfinding creature explores the world while looking for the target. Until the target is located, every time the creature moves, it checks to see if the required target is located in the adjacent locations. If the target is found, the creature simply moves to the target and creates a teleport gate. Otherwise the creature senses the adjacent locations for traces of pheromones. If attractants are found, it moves to the location with the highest concentration of attractants. If not, it drops a repellent in its current location prior to moving onto an adjacent location that is not occupied by objects, other creatures, and, if repellents exist, repellents of concentration above a certain threshold. Rule 3 Explore world rule Rule Explore_World if Target found in adjacent locations Move to Target Create teleport gate else if attractant found in adjacent locations Move to location with highest concentration of attractant else Drop repellent in current Clients Driving Innovation: Moving Ideas into Practice (12-14 March 2006) 5

6 The wayfinding creatures follow Rule 4 when returning home after locating the target. Prior to relocation, the creature drops an attractant in the current location. It then senses whether home is found in the adjacent locations. If located, the creature moves to it then turns back again to explore the world. Otherwise, the creature moves to an adjacent location closer to home than the current location. This adjacent location must not be occupied by obstacles. If the location is occupied by an obstacle, the creature moves around the obstacle by choosing some other adjacent location taking it closer to home either horizontally or vertically compared to the current location. Rule 4 Return home rule Rule Return_Home Drop attractant in current location Repeat if Home found in adjacent locations Move to Home else Move to empty adjacent location closer to Home until Home 3.2 DYNAMIC 2D ENVIRONMENT SIMULATION The swarm rules have been implemented and simulated in a 2D world to test the validity of the algorithms. The size of the world used in the simulations does not properly reflect the size of a large scale virtual environment for which the creatures are being developed. However, the initial result indicates that the creatures are able to create a trail establishing a path between Target and Home. The simulation also shows that the creatures can adapt as the world changes. This is shown as the trail is changed according to the changes made in the world. The various symbols used in the simulation are identified in Figure 11. Figure 11 Symbol Representation Home and Target are located in the world. The creatures have the knowledge of the location of Home but not of the location of Target. Hence the creatures begin exploring first the immediate vicinity. When the creatures find Target, they return while dropping attractants, shown as grey squares in Figure 12. These attractants emerge as a trail to which the creatures in adjacent locations are attracted. They follow the same trail till they too locate Target. They return Home also depositing attractants, thus strengthening the trail. Other creatures which are not in adjacent locations to the trail are unaffected by it. Figure 12 Formation of a trail Figure 13 Finding a new path Clients Driving Innovation: Moving Ideas into Practice (12-14 March 2006) 6

7 When Target and Home change their locations, initially the creatures continue to follow the old path due to the high concentration of the attractants that have accumulated on it. At the end of the old path, the wayfinding creatures randomly wander around due to the relocation of Target. This is shown in Figure 13 where a large number of creatures are gathered around the old Target location (bottom right hand corner of Figure 13). Unable to locate Target in the vicinity of the old Target location, they eventually move away. The creatures reexplore the world trying to re-locate Target. When it is found, a new trail is created. The old path evaporates in time as new attractants are not deposited on it. The simulation shows that it is possible to develop a wayfinding aid based on a swarm model that can readily adapt to the environment. By having simple creatures exploring the world, the swarm tool generates wayfinding aids which adjust to dynamically changing environments without a controlling body. 4.0 FUTURE WORK Currently, the wayfinding swarm rules are being transferred from a simple 2D virtual environment (Figures 12-13) to a large-scale 3D virtual environment (Figures 1-2). Given the disparity between the environments, a feasibility study needs to be carried out to ascertain whether wayfinding aids in micro-worlds can be scaled up to be of use in macro-worlds of considerable size and complexity. The most challenging part would be enabling swarm creatures to effectively sense adjacent locations in 3D. References Bonabeau, E., Dorigo, M., and Theraulaz, G. Swarm Intelligence: From Natural to Artificial Systems, Oxford University Press, Chittaro L. and Burigat S. 3D Location-pointing as a Navigation Aid for Virtual Environments, Proceedings of AVI 2004: 7th International Conference on Advanced Visual Interfaces, ACM Press, New York, May 2004, Darken, R. P. and Peterson, B. Spatial Orientation, Wayfinding, and Representation, In K. Stanney (Ed.), Handbook of Virtual Environment Technology, Lawrence Erlbaum Associates, New Jersey, Darken, R. P. and Sibert, J. L. Wayfinding Strategies and Behaviors in Large Virtual Worlds, In Proceedings of CHI 96, ACM Press, New York, 2001, Kennedy, J., and Ebergart, R. C. Swarm Intelligence. Academic Press, Resnick, M. Turtles, Termites, and Traffic Jams: Explorations in Massively Parallel Microworlds. Cambridge MA, MIT Press, Yoon, J.S. and Maher, M.L. A swarm algorithm for wayfinding in dynamic virtual worlds, Proceedings of ACM Symposium on Virtual Reality Software and Technology 2005 (VRST 05), Naval Postgraduate School, Monterey, USA, 2005, to appear. Clients Driving Innovation: Moving Ideas into Practice (12-14 March 2006) 7