A modified bacterial foraging algorithm for improving road networks

Abstract

One of the challenges that transportation planners face is deciding which improvements to make to existing road networks. A solution methodology for this problem must consider multiple factors, such as motorist travel behaviour and budget constraints. This is commonly represented as a bi-level optimization problem called the Discrete Network Design Problem (DNDP). In this paper, we devise and implement a Modified Bacterial Foraging Optimization (MBFO) algorithm that provides the decision maker with n-optimal solutions to the DNDP. This enables transportation planners to evaluate and compare various improvements before deciding which ones to implement. The use of a binary mapping function facilitates the implementation of Bacterial Foraging Optimization (BFO) to solve the DNDP. The MBFO employs a modified chemotaxis step with a single mutation operation to represent the tumbling motion and a revised swimming operation to ensure a diverse search of the solution space. To rank and eliminate the solutions, a gain index that measures improvement in fitness function value is used. The MBFO is tested on the 24-node, 76-arc Sioux Falls network, and the results show that it can converge to the n-optimal solution.