Research Paper

Core Innovation

This paper introduces MotoSafety, an edge-AI architecture leveraging Learned Temporal Importance to improve collision risk prediction accuracy under time pressure. It outperforms existing baselines with fewer parameters and lower latency, enabling deployment on low-cost CPUs. The model also incorporates cognitive stress as an inductive bias and demonstrates transferability to other domains.

Why It Matters

Powered two-wheeler riders face high collision risks, especially under cognitive stress like time pressure, with limited existing solutions. MotoSafety improves safety by providing accurate, low-latency risk predictions deployable on affordable hardware, enabling scalable adoption in resource-constrained regions. This supports safer road systems and reduces accident rates in vulnerable populations.

Potential Customers & Pain Points

• Transportation safety agencies – Need scalable collision risk tools
• Two-wheeler manufacturers – Need integrated safety solutions
• Insurance companies – Need accurate risk assessment
• Fleet operators – Need real-time rider safety monitoring
• Governments in low- and middle-income countries – Need cost-effective road safety interventions

Business Model

Licensing the MotoSafety edge-AI software to vehicle manufacturers, fleet operators, and safety agencies; offering customization and ongoing support services.

Competitive Landscape

• TimesNet
• LLM4TS
• Time-LLM
• iTransformer

Implementation Challenges

• Integration with existing vehicle and infrastructure systems
• Data privacy and regulatory compliance in diverse regions
• Adoption resistance due to cost or technology unfamiliarity

Validation Strategy

• Pilot deployments with two-wheeler fleets in target markets
• Partnerships with transportation safety authorities for field testing
• Performance benchmarking against existing collision risk systems