Visual Cue based (vehicle to vehicle) cooperative positioning

Abstract

Formation flight helps multi-agents cooperate visually and accomplish missions effectively. But in order to achieve a good formation shape, the agents forming the swarm must have good inter-communication among them. Currently, the main way of communication between vehicles is done by Radio Frequency, but due to its various drawbacks, the use of RF wants to be limited. Therefore, another way of communication is proposed: the visual cue based communication.

In this project, a set of autonomous vehicles will be forced to adopt a certain shape that is related to the received visual cue based marker, that is, a lead vehicle will show a marker to the followers and these will have to perform the shape that is related to the received marker. Therefore, first, a marker detection algorithm is developed which is able of detecting markers. Then, depending on the marker that has been identified, one or another formation will be executed using a potential function based approach. The approach has some visual constraints included in order to adapt it to the real scenario and only relative distances and angles between vehicles will be employed in the potential functions.

The whole thesis has been developed in a simulation environment in Matlab and Python. The results show that besides the visual constraints included, the agents are able to position themselves in a formation with equal inter-agent distance.