Thicket data repository for the EEG
anil.recoil.org
1{
2 "id": "https://gabrielmahler.org/walkability/compsci/2025/06/01/introduction",
3 "title": "Walkability Chapter 1: Introduction",
4 "link": "https://gabrielmahler.org/walkability/compsci/2025/06/01/introduction.html",
5 "updated": "2025-06-01T10:40:11",
6 "published": "2025-06-01T10:40:11",
7 "summary": "Introduction",
8 "content": "<h1>Introduction</h1>\n\n<p><em>Walkability</em> is an urbanist concept referring to how easy and desirable\nit is to walk in a given place, with considerations of the physical\nenvironment and the human individual. Typically, to estimate\n“walkability indices”, theoretical urbanist frameworks extend beyond the\nfactors related to pedestrian temporal efficiency and leverage physical\nelements such as greenery, public amenities, or other common geospatial\ninformation. Despite that, pedestrian path-finding frameworks, which\nhave been around for several decades and are relied upon by millions of\nusers every day, generally ignore any such notions described in the\nurbanist literature. Instead, these frameworks typically aim to maximize\nsimplistic objectives, most commonly the estimated duration to undertake\na path, or even only the path’s overall length. From the urbanist\nstandpoint, however, these metrics represent only a subset of the\nfactors that determine whether someone chooses to walk or selects an\nalternative mode of transportation. This problem is further amplified by\nthe fact that existing routing frameworks either entirely preclude\nuser-defined preferences or allow them only through highly complicated\nand constrained configuration files.</p>\n\n\n\n\n\n<img src=\"https://gabrielmahler.org/assets/images/thesis/new images/intro/intro valhalla Medium.jpeg\">\n\n\n\n<img src=\"https://gabrielmahler.org/assets/images/thesis/new images/intro/intro heatmap Medium.jpeg\">\n\n\n\n<img src=\"https://gabrielmahler.org/assets/images/thesis/new images/intro/intro general Medium.jpeg\">\n\n\nHigh-level illustration of our approach. Top:\nLow-walkability path generated by a popular routing framework\n(Valhalla). Middle: Our walkability scores. Bottom: Walkability-optimized\npath.\n\n\n<p>In this work, we study and answer three questions essential to\naccurately addressing the issues imposed by the methodologies used in\npopular path-finding frameworks:</p>\n\n<ol>\n <li>\n <p><strong>Shortcomings of existing path-finding:</strong> <em>What are the\nimplications of using preference inflexible, time efficiency-focused\npath-finding algorithms, particularly through the lens of\nwalkability? Why are the path-finding frameworks so inflexible to\nspecific user needs?</em></p>\n\n <p>To answer these questions, we assemble a corpus of five realistic\nrouting scenarios within the boundaries of the city of Cambridge,\nUK, discuss the unifying nature of solutions generated by three\npopular open-source frameworks, and identify potential improvements\nand missed routing opportunities. Furthermore, we discuss the\ndefinitions of preferences used to generate these outputs, and\nhighlight their complexity and poor accessibility.</p>\n </li>\n <li>\n <p><strong>Improving the quality of path-finding solutions:</strong> <em>How can urban\npath-finding be reoriented towards the concepts of walkability?\nFurthermore, how can path-finding frameworks respond more\nreceptively and comprehensively to specific user requirements and\npreferences?</em></p>\n\n <p>We provide solutions to these issues with two novel contributions.\nFirst, we present a computationally efficient tool for automated\nassessment of walkability in urban areas. We\nleverage modern natural language processing models (particularly our\ncustom fine-tuned transformer-based sentence encoders) and a\nknowledge base aggregated from public geospatial datasets, primarily\nthe OpenStreetMap. By utilizing\nrich semantic embeddings, our method significantly improves upon\nstate-of-the-art (generally computer vision-based) walkability\nassessment methods. Second, building on the acquired assessment\ntool, we present a new pedestrian path-finding framework based on\nthe A* search for the generation of pedestrian routes according to\nthe urbanist walkability principles.</p>\n\n <p>Finally, we leverage our semantically-based pipeline to develop an\napproach for embedding nuanced pedestrian objectives reflective of\nreal-world scenarios into path-finding solutions.</p>\n </li>\n <li>\n <p><strong>Simplifying user inputs:</strong> <em>What alternative approach to routing\nconfiguration files can be leveraged to simplify the process of\ninputting specific preferences?</em></p>\n\n <p>Lastly, to address this problem and provide a simplified way to\ndefine user-specific pedestrian preferences, we leverage sentence\nencoders’ ability to extract semantic associations. As our\nwalkability assessment component is based on the use of sentence\nanchors (which are utilized as points of reference for specific\nqualities and “levels” of walkability), our pipeline is also able to\nreflect user-specific preferences projected into these anchors. This\napproach allows not only for very loosely constrained preference\ndefinitions, but also their straightforward representation (as they\ncan be defined with natural language).</p>\n </li>\n</ol>\n\n<p>To evaluate our approach, we follow the order of the above questions and\nanalyze the problem and our solution on the aforementioned corpus of\nrouting scenarios. For this purpose, we also conceive four realistic\nsets of pedestrian preferences (in addition to the general walkability\npreference) that aim to maximize the presence of historical, green,\nshopping, and public safety-oriented elements in their respective\npath-finding solutions. We employ these preferences in the assessment\ncomponent to compile unique walkability maps, and then use\nthese maps in our path-finding algorithm to generate highly walkable and\nspecific-objective maximizing paths.</p>",
9 "content_type": "html",
10 "author": {
11 "name": "",
12 "email": null,
13 "uri": null
14 },
15 "categories": [
16 "walkability",
17 "compsci"
18 ],
19 "source": "https://gabrielmahler.org/feed.xml"
20}