Autonomous agents in moving subspace should move in a particular formation. This article provides a brief overview of current Reinforcement Learning methods for formation control. The approaches to creating the formation and avoiding collisions between individual agents are considered.