This article presents a system to incorporate crash risk into navigation routing
algorithms, enabling safety-aware path optimization for autonomous and
human-driven vehicles alike. Current navigation systems optimize travel time or
distance, while our approach adds crash probability as a routing criterion,
allowing users to balance efficiency with safety. We transform disparate data
sources, including traffic counts, crash reports, and road network data, into
standardized risk metrics. Because traffic volume data only exist for a small
subset of road segments, we develop a solution to project average daily traffic
estimates to an entire road inventory using machine learning, achieving
sufficient coverage for practical implementation. The framework computes
exposure-normalized crash rates weighted by severity and integrates these
metrics into routing cost functions compatible with existing navigation
algorithms. The key strength of our solution is its scalability. In addition to
the mapping data required by the navigation system, it requires only two
additional data sources commonly maintained by transportation authorities:
geolocated crash reports and traffic counts, enabling deployment across diverse
jurisdictions. For connected and automated vehicles, the framework provides
quantitative risk assessment for path planning algorithms. For conventional
vehicles, it enables drivers to make informed routing choices based on safety
preferences. Our empirical validation demonstrates that risk-aware routing
achieves substantial safety improvements while maintaining reasonable travel
times. The methodology also serves transportation agencies by systematically
identifying high-risk corridors and crash patterns across road networks. By
establishing a standardized approach to safety-aware navigation, this work
addresses an important gap in current routing systems and contributes to the
development of safer transportation infrastructure.
