An Application of Ant Colony Optimization, Kalman Filter and Artificial Neural Network for Multiple Target Tracking Problems

Abstract

The main problem in multi-sensor-multi-target tracking is data association problem that leads to multidimensional assignment problems. In this paper, a multidimensional assignment problem is solved by merging ant colony optimization, Kalman Filter and artificial neural network. The ant colony optimization algorithm is first applied to acquire the number of targets, next the states of the targets are estimated using the Kalman filter, and finally, the estimation results are improved by using the artificial neural network. In ant colony optimization algorithm, it has been investigated the process of choosing initial pheromone value of ants, ants placement in the second cycle and afterward. It is found that the ant colony optimization algorithm will produce better and faster search results if the initial pheromone value equals to the visibility and new ants are placed at the initial position in the next cycle on the point of the last visit of the previous cycle of ants. The proposed traget tracking system leads to improved performances in the simulation cases.