Contact us
Contact us

 

 

Vehicle Accident Rate Forecasting: Leveraging AI for Road Safety Improvement

Managing 3,000 km of toll motorways across Italy, our client is a key player in European mobility. Beyond traditional infrastructure management, they are at the forefront of smart mobility innovation, constantly evolving to enhance safety and efficiency. Beyond motorways, they pioneer innovative services, transforming engineering into 'smart' infrastructure to ensure a seamless experience for the traveler.

The Challenge

In the first decade of the 2000s, the company implemented road safety improvements by enhancing technical and regulatory standards, utilizing Automotive technology and the Tutor system, and introducing robust and consolidated internal procedures. However, in the last 10 years, a plateau has been reached. To take the next step toward "Vision Zero" - eliminating road fatalities - the company needed to transition from reactive safety measures to proactive, predictive strategies.

Despite having access to massive volumes of data, including 750 million rows spanning more than a decade, covering over 5,000 km of highways and 130,000 recorded accidents, the challenge was transforming this raw information into precise, actionable insights. The complexity of analyzing accident probabilities - while accounting for human behavior, infrastructure conditions, and unpredictable traffic dynamics - posed a significant barrier to progress. The company required a strategic partner capable of bridging this gap and unlocking the potential of its data.

italian highway

The Approach

Aiming to provide the company with a fully data-driven model, we leveraged AI and Machine Learning on AWS cloud infrastructure to create a solution that guides the transition from a reactive to a proactive approach in road safety.

Our team integrated academic research, historical accident data, and real-time traffic patterns into a hybrid analytical model. Using advanced metrics such as Crash Modification Factors, we identified key risk contributors, while scenario analysis tools optimized the planning of preventive safety interventions. To ensure a seamless transition, we deployed two Minimum Viable Products (MVPs), testing and refining the model before full-scale implementation. This ensured scalability and maximized the accuracy of predictive insights.

By combining statistical formulations with AI-powered models, Avvale provided a solution that allows the client to precisely assess accident probabilities for every highway segment, enabling a level of foresight previously unattainable. Our agile methodology ensured that the model would continuously evolve, adapting to new data and reinforcing the client’s long-term commitment to smart infrastructure and sustainable mobility.

highway technology

The Impact

Today, our client has a cutting-edge application active on the entire motorway network. This innovative solution efficiently processes vast amounts of data to deliver actionable insights on the phenomenon of accidents, promoting a proactive approach to road safety and a circular approach to infrastructure management.

The model proves superior in identifying relationships between road conditions and observed accident rates compared to neural net models and time series analysis models, achieving an accuracy rate of 99.6% in predicting the annual number of accidents.

Through the integration of Machine Learning algorithms, customers can now analyze extensive datasets with ease, receiving timely and pertinent information tailored to their specific needs from the entire managed motorway network. The tailor-made solution addresses the escalating demand for comprehensive variable monitoring, facilitating informed decision-making and improving operation and maintenance service.

highway lights