Undergrad Research Project - Supervised control of quadcopter formation within adverse environments with obstacles

Spring 2016

Antoine Alix
Bruno Sinopoli
Project description

The main idea is to develop a supervised obstacle avoidance control strategy capable to drive a formation of autonomous vehicles to a common goal avoiding the obstacles along the path and collision among the agents. By resorting to Command Governor (CG) ideas, a primal controller is used to stabilize each agent without taking care of physical constraints or obstacles. Then a supervisor module (one or several CG) has to be developed in order to guarantee formation coordination and constraint satisfaction (state/input constraints, collision avoidance with obstacle, mutual collision between agents). A preliminary version of such a strategy, restricted to a single agent, has been recently proposed in, where the real-time applicability of the designed control architecture has been proved with a laboratory experiments involving the quadrotor Qball-X4 from Quaser. The main points that should be properly addressed in order to extend the supervised strategy to formation of autonomous vehicles are the following: • Architecture and communication: Centralized Command Governor (only one CG collecting all the information) vs decentralized Command Governor (a CG, with neighbor- hood information for each agent). If a Decentralized Architecture is used the exchanged information between the agents (e.g. directed graph) must be defined and a consensus algorithm is required; • Procedure to avoid collision among agents. Each agent can freely move within the exterior region to the obstacles and to the other agents. An agent can be viewed as an additional moving obstacle which motion is limited to lie inside a specific region (this region can be imposed as a proper state constraint exploiting the CG capabilities).

Return to project list