Adaptive sliding mode formation control for a MAV Swarm System

Abstract

Swarm and multi-agent systems are inspired by the collective behavior of different life forms, like the flocking of migratory birds, the foraging of multiple insect species and even the slime mold aggregation. Applications using swarm robotics like surveillance and transportation usually require of robotic agents with a high degree of mobility, and quadrotor Unmanned Aerial Vehicles are good candidates due to their useful features like their ability to hover, vertical take off and landing. Since the amount of the robots in the system may vary from a couple to a hundred of agents, their size may have a representative impact to the cost and the performance of a task. Thus, quadrotors in the category of Micro Aerial Vehicles are an attractive option in the emerging field of swarm robotics. In multiple applications of swarm robotics, the agents are required to keep a desired position respect to a given reference. The desired positions can be defined by multiple approaches, where leader-follower, behavior based and virtual structure are some of the most common methods. In most of the applications, the desired formation must be maintained in every moment while the system is performing a task or, if different formations are required over time, the system must be able to switch formations and reach the new desired references. Moreover, disturbances and uncertainties in the system can compromise the success of the task if the formation control method is not able to deal with such conditions. In this dissertation, an adaptive sliding mode formation control for a quadrotor Micro Aerial Vehicle swarm system is designed. Due to its robust properties against disturbances and unmodeled dynamics, sliding mode is chosen as the control developed for the formation. Furthermore, since the control has adaptive gains, its has the advantage of managing the control effort as it is required. The formation is defined using the leader-follower approach and a geometrical description, allowing to define relative distance vectors to set the desired figure pattern. From this geometrical approach, the dynamics of the formation are obtained based on the formation error. The first stage develops the formation dynamics for a 2D figure, then, the work is extended to a 3D formation. From the error dynamics obtained from both, 2D and 3D models, the adaptive sliding mode control, considering a disturbance term, is designed. Since the agents for the system are quadrotor MAVs, the dynamics are presented along with a trajectory tracking control. Simulations are performed using Matlab and Simulink in order to validate the whole proposal. The first simulation shows the performance of the formation control achieving a 2D triangular pattern, while the second simulation shows an arrow like 2D pattern and then, it switches to a 3D crystalline shape. Both simulations include a disturbance with the objective to illustrate the performance of the proposed formation control. The last simulation shows a MAV agent tracking a trajectory obtained from the formation control. The results of the MAV simulation validate the feasibility of quadrotor MAV's as agents of the system under the proposed control.