A novel Machine-Learning, Multi-Criteria, Centralized Bicycle Routing Algorithm

In this paper, we present a novel efficient multicriteria route model for bicycles. Efficiency is particularly relevant for handling large-scale routing requests, as is essential for centralized routing services and traffic assignment problems. The multicriteria aspect is crucial in the context of bicycles, since their route choice is heavily influenced by infrastructural and topological factors. We leverage high-density GPS-based route data to optimize the efficient weighted shortest path algorithm. Depending on whether the networks of the GPS data and the route requests are identical, we propose two approaches: OptiRoute - using optimized edge weights to generate routes with the weighted shortest path algorithm, or PrediRoute - employing machine learning to learn to predict near-optimal edge weights based on edge attributes using a neural network and then generating routes. We validate the proposed method using bicycle data from the year-long mobility-tracking study Mobilität.Leben. OptiRoute shows substantial improvement across all metrics, including a 34 % increase in how much of each observed route is included in the corresponding generated route (59 % compared to 44 % for the distance-based shortest path). This is also evident for e-bike data. PrediRoute consistently outperforms the distance-based shortest path, even when testing cross-regional transferability. Our findings demonstrate that the model’s performance is robust even when GPS data does not align with route request areas, although optimal results are achieved when they do. These advancements underscore the model’s potential to enhance route planning in various contexts, paving the way for more efficient and user-centric transportation systems.     «

In this paper, we present a novel efficient multicriteria route model for bicycles. Efficiency is particularly relevant for handling large-scale routing requests, as is essential for centralized routing services and traffic assignment problems. The multicriteria aspect is crucial in the context of bicycles, since their route choice is heavily influenced by infrastructural and topological factors. We leverage high-density GPS-based route data to optimize the efficient weighted shortest path algor...     »