InclusivNet represents an important step forward in the integration of inclusive design and transport engineering for personalized travel recommendation. This toolbox aims to enable individuals with diverse needs to navigate the streets with ease by prioritizing accessibility, customization and safety, while capturing information prone to change over time. At its core, the technology is based on a temporal knowledge graph (tKG), a topological structure that takes into account several layers of information:
- Open data from various sources, such as live traffic updates, road infrastructure details, weather conditions, and more. This real-time information allows the system to stay up-to-date with the dynamic nature of urban environments.
- Personal user data including past trips recorded as GPS trajectories, favorite points of interest, and ratings as a “feedback” for proposed paths. By analyzing this data, the toolbox gains insights into each user’s preferences, historical travel patterns, and specific accessibility requirements;
- Quality measures to assess the perceived ability to enter specific locations based on four key criteria, such as attractiveness, comfort, safety, and accessibility. By considering walkability scores, it can suggest paths and areas that offer higher levels of accessibility and ensure a more seamless navigation.
This mathematical model plays a crucial role in InclusivNet’s decision-making process. In fact, it leverages the understanding of human mobility patterns, it allows to forecast future locations using graph-based link prediction techniques and generate recommendations that consider the temporal dynamics of the underlying data. This intrinsic property of the tKG allows our model to evolve and adapt to changing circumstances, ensuring that recommendations are sensitive to real-time conditions.
Our prototype also incorporates a variation of the popular Dijkstra’s algorithm, a commonly used method for calculating the shortest path between an origin and destination. By integrating walkability scores as independent variables (weights), the system optimizes path recommendations to favor a more convenient, inclusive and efficient travel experience, in accordance with United Nations’s Sustainable Development Goals (SDGs). Therefore, such combined data-driven “expertise” should allow us to propose a human-centric approach fulfilling each individual’s mobility needs and requirements, with a special care for privacy.
To conclude, thanks to state-of-the-art deep learning methods, more specifically graph neural networks (GNNs), and high-dimensional multimodal data, our toolbox strives to create an environment where people of all abilities can navigate their cities with confidence and ease, following inclusive design principles and fostering a more sustainable digital future.