Optimization of Platoon Controllers using Stochastic Models

A platoon is a group of vehicles driving close to each other in order to reduce air drag, save fuel (or electric energy) and increase road capacity. In order to avoid accidents due to the reduced gap between the vehicles, the distance is controlled automatically, supported by wireless messages between the vehicles. There exist many different control algorithms for platooning. The control performance of these controllers depends on their parameters. In previous theses, an optimization framework has been developed that can be used to find the best parameter values for a controller based on simulation results. The simulations are done with the platoon simulation software Plexe (see https://plexe.car2x.org/).

If packet losses occur in the wireless network used by the vehicles in the platoon, the control performance is degraded. Due to the stochastic nature of such packet losses, it is not enough to evaluate only one simulation run. Instead, many simulations runs with different random seed should be evaluated to get more insight both in the mean behavior and the worst case behavior of the platoon. This has not been considered yet in the optimization framework and it is even unclear how the implemented optimization algorithm will perform for such stochastic models.

Goals of the thesis

The goal of this thesis is to answer two questions:

• How many repetions of a simulation should be done depending on the packet loss rate in order to get reliable results regarding the mean and worst case behavior of a platoon?
• How does the available optimization algorithm perform in the presence of stochastic packet losses, also compared to Monte Carlo simulation, i.e. simulating with random parameters and selecting the best result of these cases?

Keywords

Simulation, Plexe, Platooning, Control Performance, Optimization, Parametrization, Python, C++

Contact

• B. Hensel