Simple and Effective Simulation Methodology to Optimize Obstacle-Avoiding Flocks

This article presents a simple and effective methodology to optimize bio-inspired behavior of drone swarms when the search space of operational parameters has low dimensionality. We exemplify our methodology by optimizing the operational parameters of a drone swarm that traverses a region while swarming and avoiding obstacles. To this end, we devise a new fitness function that favors drone swarms navigating together the obstacle as fast as possible and producing stable swarms. Next, we optimize the swarm operational parameters using gradient descent methods in a three-dimensional search space. This enables us to produce robust desired behavior of our drone swarm in the presence of external and internal noise in the operation of our system.