Flocking Towards the Source: Indoor Experiments with Quadrotors

In this paper, we propose an algorithm and present experimental results to address the source-seeking problem in which a group of small quadrotors, deployed at random initial conditions, should cooperatively find and flock towards a source of an underlying spatial scalar field without any centralized information. By implementing a local velocity tracking controller, we can model the quadrotor by double integrator dynamics and use the flocking framework. Borrowing ideas from the literature on momentum methods in optimization, we propose a control algorithm and analyze the stability of the closed-loop dynamics under suitable assumptions. We then show in experiments, that by using estimated gradients and Hessians in the presence of noisy measurements of the field strength, reasonable results for convex, log-convex and non-convex scalar fields can be achieved.