“SafeCircle AI: A transparent rule-based route safety assessment and emergency support web application for women travellers in Tamil Nadu” (2026), by R Jenifer et al. Engineering research paper. Original abstract below, unchanged. A free PDF is available.

SafeCircle AI: A transparent rule-based route safety assessment and emergency support web application for women travellers in Tamil Nadu

R Jenifer, Sriram D, Srinithi B et al. · October 29, 2026 · Engineering

Abstract

Concern for personal safety shapes how, when and where many women travel, yet mainstream navigation services rank routes almost entirely by travel time or distance. This paper presents SafeCircle AI, a full-stack web application restricted to Tamil Nadu, India, that adds a transparent safety assessment and emergency assistance to conventional routing. A React and Leaflet client and a Node.js/Express server obtain alternative routes from Open Source Routing Machine (OSRM) services and score sampled points along each route with deterministic rules that combine OpenStreetMap features from the Overpass API (emergency facilities, unlit roads and roads under construction), current Open-Meteo weather, community reports and a night-time factor. The scores are rule-based, not learned; the AI component is a multilingual assistant that answers routine questions from templates built on the computed context and forwards other questions to a Google Gemini model, with speech input and output in six Indian languages. Emergency support combines a hold-to-activate SOS with a cancellable countdown, alerts carrying the user’s location to trusted contacts, one-tap calls to contacts, the national emergency number 112 and district police control rooms, and time-limited links for sharing a live journey. The platform also provides authentication, administrator moderation of community reports and offline support. A code-level review, a production build, the bundled unit tests and deterministic re-execution of the scoring functions confirm that the routing and scoring pipeline works and identify its main limitations, notably a heuristic traffic estimate and a rank-based adjustment that overrides computed route scores. No user study or field validation has yet been conducted. The paper documents the design and implementation and sets out the corrections required before an empirical evaluation.

Free PDF DOI: 10.55248/gengpi.07.1026.34080