Average-consensus tracking problem is investigated for sensor network, in which the computing units are modeled by heterogeneous multi-agent systems composed of first-order agents and second-order agents. To track the average value of sensors' measured values, the proportional-integral consensus tracking algorithms are proposed for first-order and second-order agents, respectively. By using generalized Nyquist stability criterion, sufficient conditions of average-consensus convergence are obtained for the heterogeneous multi-agent systems under a fixed, symmetric and connected topology. Furthermore, delay-dependent convergence conditions are obtained for the multi-agent systems with identical communication delay. Numerical simulations illustrate the correctness of our results.