New pre-study: Distributed learning systems for real-time accident risk assessment
In a pivotal advancement for the future of road safety and traffic management, the AI Enhanced Mobility program has approved funding for an innovative pre-study titled "Predictive Capabilities in the Mobility Environment (LARA)".
Central to the project is "Distributed Learning Systems for Accident Risk Assessment (LARA)," a system that merges vehicle sensors and data to assess and predict real-time accident risks. LARA employs a federated deep neural learning model, amalgamating data from various sources for a comprehensive risk estimation that factors in the driver, vehicle, and immediate environment.
LARA's edge-computing-based distributed learning system is highly scalable and bypasses the need for a central data hub, ensuring swift data processing.
The risk assessment includes vehicle risk, segment risk, and context risk. Vehicle risk depends on driver behavior, time of day, and traffic conditions. Segment risk factors in weather, road quality, and proprietary fleet data. Context risk considers a range of factors linked to the driving environment, adjusting for regional and national distinctions.
The primary goal of the pre-study is to develop an application for an overarching project, led by a Ph.D. student. This larger project aims to explore possibilities for establishing a robust infrastructure capable of supporting scalable distributed learning models, with a strong emphasis on prioritising end-user experience and ensure AI-driven risk assessments are trustworthy.
- Project Title: Predictive Capabilities in the Mobility Environment.
- Core Concept: Distributed Learning Systems for Accident Risk Assessment (LARA).
- Innovation: Leveraging federated deep neural learning models and edge computing for real-time accident risk assessment.
- Partners: University of Skövde, Smart Eye, and Volvo Cars.
- Funding: 100 kSEK in funding from the AI Enhanced Mobility program (Vinnova), with an additional 100 kSEK in-kind, totaling 200 kSEK.
- Timing: The project will run from the present until the end of December.