Recent progress in microprocessors and reliable communication technologies provides cooperation of autonomous systems such as unmanned aerial vehicles, robots, spacecrafts and etc. The ability of Cooperation in autonomous systems to perform a task, which one system is not able to do alone, causes advent of multi-agent systems in scientific literature. Demands for a controller in each subsystem that makes it work toward of group's goal, yields, considering these systems in control engineering. Depending on the system conditions, various control methods proposed to control of multi-agent systems. Methods like adaptive control, optimal control, model predictive control and robust control are some of these methods. Since the adaptive control has the ability to overcome parametric uncertainties and slow variations, it is an appropriate method to control of multi-agent systems. In this thesis, a new neuro-adaptive backstepping method is proposed for higher order multi-agent systems. Dynamics of the agents (nodes) are assumed to be unknown to the controllers, so they are estimated using neural networks. In the proposed method, the multi-agent systems follow the proposed trajectory of the leader, which supposed to be an independent agent. Agents connected through a weighted directed graph with time invariant topology. In addition, only a subset of agents can access to the leader directly. Furthermore, because of complexity explosion in backstepping method, we use adaptive DSC method to control the multi-agent system. Some practical aspects are considered in the proposed method. In each section, simulation results are presented to verify the effectiveness of the proposed controllers.