Abstract
Globally, over a billion people live with disabilities, facing significant challenges in accessing public transport, which impacts their autonomy, social participation, and economic status. Current research indicates that common problems include inadequate service in terms of destination, timing, and travel duration, as well as physical barriers at stops and the behaviour and proficiency of transit staff. These issues are exacerbated for those with visual impairments or mobility challenges, such as wheelchair users, who face even greater obstacles. This research emphasizes the necessity of an “enabling transport” environment that considers all aspects of travel for those with limited mobility. This includes the physical layout of pedestrian routes, the design of buildings, and the functionality of public transportation systems. Practical measures like aligning bus floors with pavements, as mandated in the European Union, and optimizing the deployment of accessibility equipment like pallets are discussed as essential for improving access. The authors propose a research methodology that employs a graph-based approach in combination with recurrent neural networks models to suggest most accessible pathways considering fleet availability, vehicle capacity and road quality of sidewalks. The approach includes a comprehensive case study in Riga, Latvia, utilizing data from local transport operators and crowdsourced information to assess and address physical barriers. This innovative application of deep learning on graphs aims to significantly improve the inclusivity and efficiency of public transport for people with disabilities. The study emphasizes the broader benefits of creating accessible environments that improve usability for all citizens, not just those with disabilities.