Solving Vehicle Platooning Problem with Neural Networks

Abstract

This research seeks to solve the platoon control problem using the abilities of Artificial Neural Networks. Because it is difficult to obtain an analytical solution to the platoon control problem, the function approximation ability of Neural Networks is preferred to search for approximate solutions. In this thesis, we create an error function that encompasses all of the Pontryagin minimum principle (PMP) derived conditions for the platoon formation control under the Predecessor-following topology. The experimental solution for the state function, control function, and Lagrange multipliers is derived from the error function. State, control, and cost functions are implemented using Neural Networks. To optimize weights for the error function, optimization techniques such as the Genetic Algorithm (GA), Differential Evolution (DE), and variations of the self-organizing migrating algorithm (SOMA) algorithm are used. We substitute the optimized weights in the approximated functions and get the solutions to the platoon formation control problem.