Autonomous swarm testbed with multiple quadcopters

Clark, Ruaridh and Punzo, Giuliano and Dobie, Gordon and Summan, Rahul and MacLeod, Charles Norman and Pierce, Gareth and Macdonald, Malcolm (2014) Autonomous swarm testbed with multiple quadcopters. In: 1st World Congress on Unmanned Systems Enginenering, 2014-WCUSEng, 2014-07-30 - 2014-08-01.

[thumbnail of Clark R et al - Pure - Autonomous swarm testbed with multiple quadcopters 30 Jul 2014] PDF. Filename: Clark_R_et_al_Pure_Autonomous_swarm_testbed_with_multiple_quadcopters_30_Jul_2014.pdf
Accepted Author Manuscript
Download (992kB)

Abstract

A testbed has been developed to validate and trial swarm engineering concepts. The vehicles used in this testbed are commercially available Parrot AR.Drone quadcopters that are controlled through a computer interface connected over a 65Mbps, IEEE 802.11n (Wi-Fi), network. The testbed architecture is presented with the implementation of a distributed controller, for each vehicle, discussed. The controller relies on a tracking system to provide precise information on the position and orientation of the aerial vehicles within the enclosure. This enables the autonomous and distributed elements of the control scheme to be retained, whilst alleviating the drones of the control algorithm's computational load. The testbed is used to control 3 drones effectively, where the control, communication and tracking systems are scalable to at least 12 drones. The paper also introduces the application of swarm engineering in remote visual inspection, with multiple airborne platforms visually inspecting a target by completing coverage bands before transitioning to another height. A kinematic field enables the drones to follow this path autonomously with the field being asymmetrically modified as a result of drone interactions. This modification of the field for a single drone is shown to prevent vehicle collisions by enabling queuing behind the leader. Using a swarm of 3 quadcopters, the coverage time for a target can be reduced by around 60% when compared with a solitary drone. Finally the three-dimensional model of a target that is generated from drone footage is presented; this surface-meshed model is constructed in post-processing, through photogrammetric analysis of the collected images.

CORE (COnnecting REpositories)