Model Predictive Control for Vehicle Platooning

A Practical Comparison against Traditional Methods

More Info


Due to the increase in traffic, road congestion has gone up. Vehicle platooning is a possible way to increase the capacity of a given road, by decreasing the distance between the vehicles in the platoon. At the moment, the control of vehicle platoons is commonly done using PDD controllers. The advantage of this is that it requires little computational resources. With improvements in computing technology in recent years, the possibility of using a more computationally costly method has opened up. MPC is such a method, with a couple of important advantages for vehicle platooning. Firstly it is (finite-time) optimal so should lead to smaller deviations from the desired trajectory, thereby decreasing the minimum safe distance between the vehicles. Secondly, it is inherently able to include constraints on the behaviour of the platoon. This is important because it allows for explicitly enforcing the minimum safe distance between the vehicles, thereby preventing collisions.
While many studies have looked into the possibility of using MPC for vehicle platooning before. None of them have made the direct comparison with the current standard e.i. PDD control in a practical setup. This is important as it will give a good indication of both the possible advantages and pitfalls MPC could have in real platoons. To investigate this both a PDD and MPC controller are implemented on a platoon of RC-vehicles and their performance is compared.
To compare the performance five metrics are used on four scenarios. The metrics used are coherence, combined local error, velocity error, energy expenditure and minimum headway. The scenarios are a step up, step down, ramp up and a ramp down.

From the simulation results, it is clear that in an idealised case the MPC outperforms the PDD controller. However, when looking at the lab results this difference is less pronounced. In the lab setting, it is also clear that both controllers struggle to reliably deal with the added uncertainties caused by measurement noise and inconsistent behaviour of the vehicles.
In future research more emphasis should be placed on the robustness of the controllers used, to mitigate these problems.