Cells maintain their morphology and produce forces thanks to the cytoskeleton, which is composed of three types of protein polymers, amongst which the actin microfilaments. Actin filaments assemble into diverse architectures, which assembly and disassembly is tightly controlled in space and time. Indeed, the progressive hydrolysis of ATP in the monomers causes the actin filaments to age. Thus, actin needs to be recycled. When assembly and disassembly compensate, different actin architectures are in a dynamic steady state, in which the pool of actin monomers is renewed. Disassembly of actin structures also maintains a large reservoir of polymerization-ready monomers ready to assemble when needed by the cell.The lamellipodium is the locomotory organelle of the cell, and is made of a thin yet very dense sheet of dendritic actin network, with regulatory proteins. A pivotal protein is ADF/Cofilin, which is responsible of the disassembly of old actin filaments by fragmentation and debranching. To date, there have been extensive studies about the microscopic mechanisms, but if one wants to understand cell motility, one must decipher the collective and macroscopic disassembly of the dendritic actin network.By combining motility media reconstituted from purified proteins, and a new surface micro-patterning technique, I was able to reconstitute lamellipodium-like dendritic networks in vitro. During this thesis I explored the parameters that control the macroscopic disassembly of these networks. This work shows that the disassembly of dendritic actin networks depends on their architecture (density) and geometry (size): dense or extended networks are less efficiently disassembled and remain cohesive longer. Simulations show that these effects can be explains by a local depletion of ADF/Cofilin in the volume surrounding the network. Besides, networks of heterogeneous densities acquire directionality. This steering is modulated by selective disassembly of the networks by ADF/Cofilin. In parallel, these studies established a ratio at which networks are at a dynamic steady state.This work goes further than the fundamentally important studies about fragmentation of individual actin filaments, and establishes new parameters that control the disassembly of dendritic actin at the macroscopic scale.