Model Predictive Control for Vision-Based Platooning Utilizing Road Topography

Sammanfattning: Platooning is when vehicles are driving close aftereach other at a set distance and it is a promising method toimprove the traffic of todays infrastructure. Several approachesfor platooning can be taken and in this paper a vision-basedimplementation has been studied. With a camera that detectsthe orientation of a marker attached to a small vehicle, it hasbeen examined how the pitch of the marker can be exploitedto perform vision-based platooning considering the road grade.A model predictive control strategy is presented to maintain aplatooning distance with the potential of utilizing road topography.The aim of the project was to use this information tominimize brake and motor forces of the platooning vehicle. Thestrategy was based on relative vehicle states, detectable by acamera. The model predictive controller was implemented onsmall robotic vehicles and tested on a flat surface. The controllerwas successful in converging towards the wanted distance andcapable of reaching a steady state speed. The results showed thatit took 15 seconds for the system to reach a steady state.