In this work we introduce the concept of spatial pinning control for a network of mobile chaotic agents. In a planar space, where N agents move as random walkers, the control input is applied to agents only when they enter a given area, called control region. The control is effective to drive all agents to a reference evolution provided that the velocity of the agents is high enough, and has better performance than pinning control on a fixed set of agents. We derive analytical conditions on two important parameters (the relative size of the control region and the agent density) for the global convergence of the system to the reference evolution.