Project Summary

This solution is a modular, AI-enabled safety and alert system designed to enhance and improve personal security around rail stations, walkways, and car parks, especially for women and vulnerable users. Violent incidents on UK railways increased by more than half from 2021 to 2023, reaching 11,357 cases, according to the British Transport Police. Sexual offences rose by about 10%, and reports of sexual harassment doubled from around 950 to 1,908. Starlight integrates smart-behaviour detection, real-time SOS alerts, and solar-powered modules into compact hardware that can be retrofitted to existing poles or building walls. Unlike CCTV or safety apps, Starlight actively detects incidents such as harassment, loitering, or medical emergencies and triggers alerts using audio or visual cues, routing them to authorities in under 10 seconds. This project will assess Starlight’s AI accuracy, response speed and sensor reliability while gathering feedback from women, young people and passengers with disabilities. Our results will inform national scaling plans and support the UK Government’s Ending Violence Against Women and Girls Project Summary strategy.

Project Achievements

Starlight advanced from TRL 3 to TRL 4 through full system integration and controlled validation. The project successfully unified previously separate subsystems. AI behavioural detection, voice-activated SOS recognition, environmental sensors, solar-powered energy systems, and secure wireless connectivity, into a single operational prototype capable of real-time incident detection and automated alert escalation. Controlled testing demonstrated AI detection accuracy exceeding 90% for predefined distress and harassment scenarios, with alert escalation latency of approximately 10 seconds, a significant improvement over traditional CCTV workflows reliant on manual monitoring. Additional outputs include validated performance datasets, system architecture documentation, and deployment configuration insights relevant to rail-adjacent environments. The project generated foreground know-how, including validated detection algorithms, integration architecture, and benchmarking evidence to support future IP development and live trial preparation.

Conclusions

Starlight demonstrated that AI-enabled modular safety infrastructure can meaningfully improve safety outcomes in rail-adjacent environments without requiring major civil works or grid connectivity. The system’s solar-powered, edge-processing architecture supports low-carbon, privacy-compliant deployment aligned with UK GDPR. Social impact evidence is compelling; prior pilot engagement showed that over 80% of female participants reported feeling safer with visible, responsive safety infrastructure. Automated detection removes the reliance on user-initiated reporting, directly addressing a key barrier for women and vulnerable users in distress. TRIG participation validated the core technical approach, strengthened internal capabilities in AI validation and ethical deployment, and built credibility with transport stakeholders. Key lessons included the value of quantifiable performance metrics, privacy-by-design from the outset, and early operational stakeholder engagement. The work confirms Starlight’s potential as a scalable, infrastructure-based safety solution applicable beyond rail to bus interchanges, rural transport, and active travel corridors.

Next Steps

The immediate priority is progression to TRL 5–6 through live operational trials in a rail-adjacent environment, targeted within 12–18 months, subject to site approvals, ethics clearance, and follow-on funding. Pre-deployment activities include real-world validation of AI detection accuracy under operational conditions, extended false-positive rate testing, and integration with responder workflows and control room systems. Independent cybersecurity and data protection assessments will be completed to ensure full regulatory compliance. On IP, future patent filings are planned around AI alert escalation workflows and system integration architecture validated during TRIG. Commercialisation will follow a staged approach, pilot trials with rail operators and local authorities, followed by technology licensing to infrastructure contractors and security platform providers for scalable rollout. Early interest has been received from regional transport stakeholders. Key barriers, regulatory timelines, procurement processes, and public perception of surveillance will be addressed through privacy-by-design architecture and active stakeholder engagement.