Optimizing the allocation of emergency vehicles is essential for enhancing route-planning efficiency and ensuring road safety during traffic incidents. Traditional dispatch methods often struggle with complex scenarios due to their inability to integrate and balance multiple conflicting factors. This study proposes a multi-objective dispatch framework for emergency vehicles that integrates regression analysis, deep learning, and an enhanced ant colony algorithm. Key environmental factors (e.g., weather, visibility) are selected through logistic regression, and a BP neural network predicts the impact ranges of accidents. The adaptive ant colony algorithm optimizes dynamic routing through innovations such as adjusting state transition probability and implementing pheromone reward—penalty strategies. It achieves faster convergence (with a comprehensive index of 86 in 8 iterations compared to 158 in 20 iterations) and superior path quality (a 9% reduction in rescue time and a 12% decrease in costs). Compared with existing hybrid frameworks, this study is the first to integrate logistic regression-selected environmental factors with BP neural network-predicted accident impact ranges, and further proposes adaptive state transition and pheromone reward-penalty update mechanisms, thereby achieving faster convergence speed and superior path quality in dynamic multi-objective rescue route planning.
Zhang et al. (Wed,) studied this question.