LAUSR

Drink driving is an infamous factor in road crashes and fatalities. Alcohol testing is a major countermeasure, and random breath tests (RBTs) deter tested drivers and passersby (observers who are not tested). We propose a genetic algorithm (GA)-based RBT scheduling optimisation method to achieve maximal deterrence of drink driving. The RBT schedule denotes the daily plan of where, when, and for how long tests should occur in the road network. The test results (positive and negative) and observing drivers are considered in the fitness function. The limited testing resource capacity is modeled by a number of constraints that consider the total duration of tests, the minimum and maximum duration of a single test site, and the total number of test sites during the day. Clustering of the alcohol-related crash data is used to estimate the matrix for drink driving on the scheduled day. The crash data and traffic flow data from Victoria, Australia are analysed and used to describe sober/drink driving. A detailed synthetic example is developed and a significant improvement with 150% more positive results and 59% more overall tests is observed using the proposed scheduling optimisation method.