Available online at ScienceDirect. Procedia Engineering 161 (2016 )

Transcription

1 Available online at ScienceDirect Procedia Engineering 161 (2016 ) World Multidisciplinary Civil Engineering-Architecture-Urban Planning Symposium 2016, WMCAUS 2016 Use of Ant Algorithms to Optimize Pedestrian Communication Routes with The Application of GIS Tools: A Case Study of Olsztyn Iwona Cieslak a, Karol Szuniewicz a, *, Tomasz Templin b, Szymon Czyza a a Department of Geoinformation Analysis and Cadastre, Faculty of Geodesy, Geospatial and Civil Engineering, University of Warmia and Mazury, Poland b Department of Satellite Geodesy and Navigation, Faculty of Geodesy, Geospatial and Civil Engineering, University of Warmia and Mazury, Poland Abstract The authors describe the rules for the use of qualitative research methods and optimization in conjunction with GIS tools to assess and optimize the pedestrian communication. Research was carried out on a part of the urban space, which is characterized by large and visible population trends. The analysis that was carried out resulted in creating cartographic imaging of the pedestrian communication system state in the urban space, its consistency and to provide directions for its modernization. The simplest example of a system that resembles the behaviour of ants is the system of pedestrian communication. This system has the features of self-organization, a permanent and flexible response to changes in a short period of time. Therefore, it is a perfect example of the system that can be optimized with the use of the so-called ant algorithms. The aim of this paper is to present the applicability of methods based on ant systems, which can be used for searching the optimum agent passage routes according to the imposed selection criteria. The communication function is one of the most important functions of urbanized areas. Communication routes define the city plan and undoubtedly influence spatial perception and usefulness, introduce movement into space and at the same time enable learning about space. Because of the rising urbanization level and accumulation of spatial functions, communication organization is an important task, often determining spatial usefulness. The problem of communication organization results, among others, from the rising number of space users, which requires the adaptation of the communication infrastructure to their needs Published The Authors. by Elsevier Published Ltd. This by Elsevier is an open Ltd. access article under the CC BY-NC-ND license ( Peer-review under responsibility of the organizing committee of WMCAUS Peer-review under responsibility of the organizing committee of WMCAUS 2016 Keywords: pedestrian communication; GIS tools; urban space, ant algorithms. * Corresponding author. Tel.: address: karol.szuniewicz@uwm.edu.pl Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license ( Peer-review under responsibility of the organizing committee of WMCAUS 2016 doi: /j.proeng

2 Iwona Cieslak et al. / Procedia Engineering 161 ( 2016 ) Introduction The economic costs of communication in cities are also steadily rising. The growing congestion is a major danger for modern cities. This problem is the focus of attention for analysts and scientists, who increasingly often separate pedestrian communication from the general problem of communication [1]. This kind of movement, especially in urbanized space, was often considered self-organizing [2] and was rarely the object of conscious planning. This manner of shaping pedestrian communication has been changing with an increasing awareness that the quality of urban space directly depends on the quality of this communication [3,4,5]. The organization of urban space in its multi-dimensional nature and multi-connected conurbations is often compared to an anthill. A human is perceived in this space as a chaotically moving ant seemingly heading in a not very specific direction [6]. Such comparisons are a contribution to the observation of the operation of anthill area systems, which by, as it were, reducing the scale of phenomena enables observation and the establishment of certain permanent and optimizing procedures [7,8]. The case of ant algorithms was discussed for the first time by Dorigo [9]. Dorigo s research shows that collective ant behaviour is determined by certain evolutionary hypotheses: every individual in the population feels and reacts to the stimuli acting on the individual; the rules of collective population behaviour are effective only under specific environmental conditions, changes in the population s living environment cause changes in the effectiveness of population behaviour, the variable factors in the population are: the number of individuals in each caste and the method of communication between individuals. Ant colonies are often called super-organisms, which are characterized by high survival capabilities despite the primitive traits of individuals. The mechanisms which cause the operation of a colony to adapt optimally to its living environment and adapt in a short time to changes occurring in this environment became the subject of examination by many researchers [10, 11, 12]. The universality of this behaviour caused quick adaptation of these mechanisms to fields of science dealing with finding optimum spatial solutions and decision-making processes [13]. 2. Use of ant algorithms to assessment of the pedestrian system network in the subject area The use of an ant algorithm to optimize pedestrian communication routes with the application of GIS software requires the recording of available information on footway assessment in the form of a geobase. The aim of the optimization first assumes the determination of the routes which are the best solutions using ant algorithm mechanisms and then the determination (on their basis) of the best connections in terms of frequency of use. In the next step, indication of the application of these results to the level of the entire network and determination of both weak and best elements form the basis for indicating optimum network solutions. The main operation of the algorithm took place based on the Iterative loop model shown in the diagram. The algorithm, having available the node obtained from the previous model, starts by the detection of all connections originating from the given node. For selected elements, the probability of their selection is determined, computed based on the quality of individual connections represented by the coefficient K, determined based on the assessment of the quality of the connections, and the coefficient f, i.e. the pheromone trail, the characteristic element for ant algorithms, which assumes variable values during the algorithm s operation. The value of the coefficient f of 1 was adopted at the start of the algorithm s operation. Moreover, depending on the number of connections originating from a given node, the sum of their quality was determined for them. The manner of determination of these values is presented by the formulas below. i proi = ij(t) [ij(t)] (1)

3 2008 Iwona Cieslak et al. / Procedia Engineering 161 ( 2016 ) pro k i p (t)= i= 1 (2) ij proi n where, proi: choice factor for i-th connection; di: quality i-th connection based on feature values and their weight; q: pheromone trail for i-th connection; pi: probability value for i-th connection. The connection with the highest probability value was selected based on the obtained values. According to the selection made, the pheromone trail values are recalculated according to the formulas below: - for the best selected connection q(t+ 1)= 0,5 q(t)+( 1d) (3) - for other connection q(t+ 1 )= 0,5 q(t) (4) where, q(t): pheromone trail for i-th connection in t time; q(t+1): pheromone trail for i-th connection in t+1 time. The best connection is recorded in the object class route_node_number together with all computed parameters and the final node of the selected connection is then determined, which becomes the new initial node, for which the actions from the above model are performed again. The described analysis is carried out until the nodes are exhausted, which can occur in two ways: basic, when all nodes existing in the network are used, secondary, when the algorithm encounters during its operation a node at which there is only one connection with the node used earlier as the initial node. 3. Results and Discussions It was necessary to select a representative area for the purposes of conducting the study. Since the main subject of the study is linear technical infrastructure, urbanized space was examined, densely developed and growth-oriented. Mainly because of these features, the study was conducted within the city of Olsztyn. As a result of the conducted field study and inventory of the subject of the study, a communication system network was created based on the determined pedestrian communication routes. The created network became the object of optimization by the proposed method. For the purposes of optimization, the created model was defined as a closed network. The said network is built of 2,001 connections linked by 1,100 nodes. It should be remembered, however, that many of edges set out in the field do not run in a straight line. Their real length is often much greater. The real length of all edges is m. The total route extension is 1.06, which is equal to the max route extension. This may indicate a quite detailed reflection of the network routes with the prepared model. However, the real length of edges was adopted for further analysis because it affects the final quality of these edges. Six features were analyzed more thoroughly during the conducted study. The examined features of pedestrian space are: surface, width, location relative to the road, length, inclination and the manner of footway administration (formal and informal paths) [14, 15]. Concrete quality characteristics and discriminant values corresponding to this quality ranging from 0 to 1 were assigned to each of them using qualimetric methods [16, 17]. The spatial relationship of the objects was the starting point for optimization analysis within the subject networks [18]. The focus was then on obtaining information about the features of

4 Iwona Cieslak et al. / Procedia Engineering 161 ( 2016 ) connections present in the network and determination of their quality. A packet of such data was obtained by analysing the available cartographic documents, topographic maps, orthophotomaps supported by site visits or for some features directly with analytical tools from the used software. This allowed the quality of individual network elements to be determined. The Summary Statistics tool was applied to determine the frequency values of individual connections. With such a data set available, it was possible to join the input objects of the class connect with this table based on OBJECTID, which allowed the frequency of selection to be assigned under the genetic algorithm s operation to individual objects. Cartographic documents (general map Figure 1 and identifying the best connections in the attached networks) were created on this basis. 4. Conclusions Fig. 1. Map of the best connection selected. The developed an algorithm and analysis of its operation results allowed important conclusions to be considered regarding the structure and principles of the design of pedestrian communication networks. The area of the subject space can be divided into two parts based on the study results. In the northern part of this area, where the building density is much lower, the network seems far more coherent. The path selection level in this part is evenly distributed, which may mean that both edge density and quality is appropriate for building density. Colloquially speaking, all paths are needed there and their routes and quality are optimal to serve the needs of this area. In the southern part, the density of routes selected by successive algorithm steps is clearly visible, which may suggest that the most important lines of the network lie in the central part of the area and their exclusion from the network would cause considerable difficulties in movement in the communication system model. Regardless of building density and, thus, communication network density, around fifteen edges with specific quality features prove the most important and guarantee optimum

5 2010 Iwona Cieslak et al. / Procedia Engineering 161 ( 2016 ) movement in all directions. The edges selected most frequently in the algorithm work process are often informal edges, which also indicates a mismatch between the system and spatial conditions. This is also the indication of the edges which should be included in the pedestrian communication network in the basic system. Therefore, they require work on improving their quality to the level guaranteeing comfort of movement. Many of the identified edges in the southern part of the city were not selected even once. This means that their routes, as well as their quality, are random and insufficient for the purposes of network optimization. The analysis clearly shows the principles of design of a pedestrian communication system with a simple, multidirectional and not very dense structure. Especially in the space which is strongly urbanized or designated for dense development, when designing a pedestrian communication system, the basic system should be built to guarantee long maintenance of the direction of passage and perfect quality of width of the surface and comfort of location relative to the road. Only on the basis of this system should the auxiliary system be designed, to ensure meeting the needs of individual built-up blocks and even single building References [1] Hutabarat LR, Walkability: what is it?. Journal of Urbanism: International Research on Placemaking and Urban Sustainability 2(2): pp ; 2009 [2] Helbing D, Molnar P, Farkas I, Bolay K, Self-organizing pedestrian movement. Environment and Planning B: Planning and Design, 28(3): ; 2001 [3] Turner GE, Simulation of pedestrian movement in means of egress. Journal of Architectural and Planning Research. 1 (3): pp ; 1984 [4] Olszewski PS, Walking as a mode of transport a planning and policy perspective. Warszawa: Oficyna Wydawnicza Politechniki Warszawskiej.; 2007 [5] Oreskovic NM, Roth P, Charles SL, Tsigaridi D, Shepherd K, Nelson KP, Bar M, Attributes of form in the built environment that influence perceived walkability. Journal of Architectural and Planning Research. 31(3):pp ; 2014 [6] Cielak I, Jako ukadu komunikacji pieszej w przestrzeni zurbanizowanej. In: Z. Rusnak and B. Zmylona (eds.) Jako ycia a zrównowaony rozwój. Proceedings of the Fifth International Scientific Conference Quality of Life and Sustainable Development; September 20-21, Wrocaw: pp.45-56; 2013 [7] Cielak I, Wspóczesna waloryzacja przestrzeni zurbanizowanej. Olsztyn: Wydawnictwo UWM; 2012 [8] Cielak I, Szuniewicz K, The quality of pedestrian space in the city: a case study of Olsztyn. Bulletin of Geography. Socio-economic Series 30, : pp.31-41; 2015 [9] Dorigo M, Optimization, Learning and Natural Algorithms (in Italian). PhD thesis, DEI, Politecnico di Milano, Italy: pp.140; 1992 [10] Dorigo M, Gambardella LM, Ant colonies for the travelling salesman problem. BioSystems 43: pp [11] Fenet S, Solnon C, Searching for Maximum Cliques with Ant Colony Optimization. Applications of evolutionary computing (EvoCOP 2003), Lecture Notes in Computer Science 2611, TN. Springer Verlag:, pp ; 2003 [12] Dorigo M, Stützle T, Ant Colony Optimizations. Massachusetts London: Institute of Technology, A Bradford Book, The MIT Press Cambridge; 2004 [13] Grzymkowski R, Kaczmarek K, Kietyka S, and Nowak I, Wybrane algorytmy optymalizacji. Algorytmy genetyczne. Algorytmy mrówkowe. Gliwice: Wydawnictwo Pracowni Komputerowej Jacka Skalmierskiego; 2008 [14] Land Transport NZ, The principles of pedestrian network planning [15] NCHRP REPORT 616, Multimodal Level of Service Analysis for Urban Streets. Transportation Research Board. Executive Committee; [16] Nazarov NG, Krushnyak NT, What is measured in qualimetry? Measurement Techniques 49(3): pp ; 2006 [17]Martínez P, Qualitology. Unlocking the Secrets of Qualitative Research. ESIC; 2008 [18] Longley PA, Batty M, Advanced Spatial analysis: The CASE Book of GIS. Redland, CA: ESRI Press; 2003