Slime-mold algorithm boosts on-demand transit by 101% in simulations

A team of researchers led by Simon Garnier and Petras Swissler at the New Jersey Institute of Technology has published a paper proposing a new approach to demand‑responsive transit (DRT) inspired by the foraging behavior of slime molds. The algorithm treats each bus as an agent in a distributed swarm that dynamically routes itself using a cooperative bidding process. Buses compete and collaborate to pick up passengers and optimize transfers, eliminating the need for a central scheduler. The method also introduces dynamic transfers — passengers can switch between buses mid‑trip to further improve efficiency.

Simulations were run on real road networks from OpenStreetMap for three scenarios: suburban, urban, and semi‑rural. Compared to a fixed‑route baseline, the swarm approach increased passenger delivery rates by 28% (suburban), 49% (urban), and 101% (semi‑rural). Walking time was reduced by more than 75% in every scenario, meaning passengers spent dramatically less time walking to stops or transfers. The results suggest that bio‑inspired swarm intelligence could make on‑demand transit viable not just in low‑density areas but also in denser urban environments, potentially lowering operating costs while improving reliability.