Browsing by Author "Sawandi, H"

Now showing 1 - 2 of 2

item: Conference-Abstract
Real-time tracking data and machine learning approach for assessing pedestrian walking patterns – impact of spatial dynamics at the University of Moratuwa
(Transportation Engineering Group, Department of Civil Engineering, University of Moratuwa, 2024) Sawandi, H; Jayasinghe, A
Understanding the key elements that affect pedestrian walking patterns is vital for enhancing transportation systems in urban environments. This experimental study tackles important gaps in current literature, especially the insufficient examination of walking behavior concerning land use characteristics and spatiotemporal factors. By focusing on walking patterns in connection to these aspects, this study intends to provide insights that can help improve urban transportation planning and policy. This study employs real-time tracking data and machine learning to evaluate the intricate dynamics of walking behavior using GPS-enabled location-based services and mobile accelerometer data. The investigation utilizes k-means clustering to identify walking patterns and the eXtreme Gradient Boosting (XGBoost) machine learning algorithm to evaluate the influence of spatial factors on pedestrian movement. The primary objective is to assess the influence of land use attributes on walking patterns, offering quantifiable information regarding how spatial design could promote or impede walking behavior. The study is predicated on a case analysis performed at the University of Moratuwa, providing a concentrated examination within a campus setting that embodies educational and mixed-use attributes. Identified key components encompass tree views, mean depth, and choice—elements demonstrated to influence land use planning in a manner that improves walkability. Tree views enhance visual appeal and offer shade, promoting walking by fostering a more pleasant pedestrian experience. Mean depth, denoting the spatial connection and accessibility of various areas, defines the navigability and simplicity of a space for pedestrians. Furthermore, the presence of "choice," or the availability of alternative routes, enhances navigational flexibility, motivating individuals to traverse the area on foot. These findings underscore the necessity of integrating these spatial dynamics into urban transportation strategy. By comprehending the influence of land use features on pedestrian behavior, transportation planners and urban designers can formulate specific methods to enhance walkability. These solutions may include incorporating shaded walkways, optimizing spatial layouts to minimize walking effort, and providing several route options to cater to pedestrian preferences. Moreover, these elements facilitate the alleviation of automotive congestion, encourage active transportation, and enhance sustainable urban mobility objectives. This study employs machine learning and real-time data collecting to analyze pedestrian movement patterns, offering insights that surpass conventional survey-based methodologies. The findings highlight the necessity for evidence-based, data-driven urban design and planning methodologies to develop pedestrian-friendly environments that enhance health, accessibility, and sustainable mobility. Policymakers and urban planners are urged to utilize these findings to develop strategies that emphasize pedestrian requirements in transportation systems, guaranteeing that metropolitan areas are both walkable and supportive of efficient and resilient urban transport networks.
item: Conference-Full-text
Walking behavior mapping and spatiotemporal analysis using mobile phone and GEOAI
(Sri Lanka Society of Transport and Logistics, 2024) Sawandi, H; Jayasinghe, A; Retscher, G; Gunaruwan, T. L.
Currently, there is an ongoing discussion regarding the role of urban planning and transport planning in the development of walkable cities. It argues for rethinking the technology-centric approach that combines urban/transport planning and technological domains, such as developing field called Geospatial Artificial Intelligence (GEOAI). This study addressed theoretical and practical challenges in walking behavior analysis. First, map pedestrian walking behavior. Second, quantifying spatiotemporal element’s impact on walking behavior is challenging. The utilization of GEOAI in this field is still deficient. The methodology of this study employs GPS-enabled location-based services to capture walking behavior and street view, isovist factors, and space syntax to quantify the environment. This method maps walking behavior using GIS and k-means clustering, an unsupervised machine-learning model used for splitting data. Additionally, Extreme Gradient Boosting (XGBoost), a supervised machine learning, is employed to analyze how spatiotemporal factors influence walking behavior. The findings highlight a significant relationship between tree view, mean depth, choice, and walking behavior. This research provides transport and urban planners with crucial insights and a novel methodological framework to develop more walkable cities, optimize urban design, transport planning strategies, and enhance urban livability and sustainability.