Driving Risk Assessment Using Intervals and Weighted Fuzzy Rules

Risk assessment is important to guide the definition of control measures that prevent harm, especially in safety-critical domains in transportation, such as driving behaviour identification. With the widespread of sensors constantly gathering data, artificial intelligence is now used to produce online driving risk assessment. For many applications, however, current approaches found in the literature often do not consider the complex interactions between multiple driving risk factors and their context; and such information are usually not captured in sensor data. Our hypothesis to address this data gap is that information about the interaction of risk factors and their contextual impact can at least be partially obtained from stakeholders' expertise in the domain. In this paper, we introduce a fuzzy driving risk assessment framework that captures the synergistic impact of risky driving behaviours on road safety. The information about those effects are obtained from stakeholders using intervals that capture uncertainty in expert knowledge. The intervals are modelled using weighted fuzzy rules in a rule-based fuzzy inference system. The resulting system automatically maps simultaneous occurrence of driving behaviours to their corresponding levels of risk. An application of the framework to heavy goods vehicle driving is presented as a case study.