A Practical Framework for Multi Agent Experiments in Aerial Robotics

Abstract

A framework is presented for distributed control of rotary wing aerial robots using open-source flight controller firmware, commercial communication and computing hardware in an outdoor visual line of sight test environment. The work builds upon existing documented frameworks that focus mainly on laboratory based experiments. The research context is evaluation of novel rules for safe intersection of high-density aerial traffic flows. The framework is based on the use of a PX4 flight controller, Linux onboard computer running bespoke flight management code, MAVlink WiFi Bridge and WiFi module on each flight vehicle, and a ground control station running QGroundControl and a bespoke graphical user interface for interaction with vehicles and control of the overall experiment. SITL functionality is provided using Gazebo for physical simulation. Communication is via WiFi, with an individual radio control transmitter for each vehicle provided as backup. The developed framework was evaluated using a fleet of 10 multirotor vehicles based on 1kg class racing drones operating in a test volume approximately 100x100x100m. Engineering evaluation results are presented in the form of an technical issue register which highlights the engineering challenges that had to be solved to achieve a practical working system. Results are then presented for an example drone traffic intersection experiment in which two helical traffic flows are joined without conflict using flight rules running local to each drone. The proposed framework has been shown to be scalable to up to 10 vehicles, with addition of further vehicles being limited by WiFi router performance. The framework is hardware agnostic as far as possible and is extensible to use for a range of different distribute control experiments.