This work concerns the use of multi-agent system models as a paradigm to support dynamic scheduling decisions. A dynamic class of scheduling problems has been considered, and a just-in-time manufacturing production has been assumed. In particular, the problem of scheduling independent jobs on identical parallel machines, to optimise a non-regular objective, including both job earliness and tardiness penalties, has been faced. Different classes of agents compose the multi-agent scheduling system, and a schedule is progressively defined by a negotiation protocol among the agents associated with the jobs and the machines.