Game semantics characterize the interactive behaviour of proofs and programs, by modeling them as strategies which describe the way they react to their environment. In order to take in account concurrent aspects of proofs in linear logic, usual notions and techniques in game semantics are recasted in an asynchronous and non-alternating framework. In a first part, we define a family of asynchronous strategies giving rise to a model of linear logic, which is fully complete for the multiplicative fragment. These strategies are defined in a purely local way by a series of diagrammatic axioms. Then, they are refined by a dynamic scheduling criterion, which is shown to constrain strategies to satisfy an oriented variant of the correctness criteri...