During the last decade, real-time video streaming over ad-hoc networks has gathered an increasing interest, because of the attractive property of being able to deploy a visual communication system anytime and anywhere, without the need for a pre-existing infrastructure. A wide range of target applications, from military and rescue operations, to business, educational, and recreational scenarios, has been envisaged, which has created great expectations with respect to the involved technologies. The goal of this thesis is to provide an efficient and robust real-time video streaming system over mobile ad-hoc networks, proposing cross-layer solutions that overcome the limitations of both the application and network solutions available at this time. Our contributions cover several aspects of the mobile video streaming paradigm: a new multiple description video coding technique, which provides an acceptable video quality even in presence of high loss rates; a novel cross-layer design for an overlay creation and maintenance protocol, which, with a low overhead, distributedly manages a set of multicast trees, one for each description of the stream; an original distributed congestion-distortion optimisation framework, which, through a compact representation of the topology information, enables the nodes to learn the structure of the overlay and optimise their behaviour accordingly; and, finally, an integration with the emerging network coding paradigm.
