This paper investigates the finite-time formation control problem for high-order nonlinear multiagent systems (MASs) with consideration of obstacle avoidance, unmeasurable states and dead-zone input. A neural networks  k-filter observer is designed to estimate the unmeasurable states and cope with the problem of dead-zone input. Also, by using a tangent type Lyapunov barrier function (LBF), the obstacle avoidance mission can be completed for MASs without dynamic mismatching.