How much training is needed for interference coordination in cellular networks?

Cooperative techniques for cellular networks promise very high data rates, but require additional and precise knowledge of the serving and interference channels. We show, how the pilot symbols required for achieving this information affect the possible data rates. The measurements of the channels are suffering from pilot contamination, due to the measurements in adjacent cells. On the one hand, with a too short pilot length, cooperation is not possible and the channels are learned too poorly, degrading the possible data rates. On the other hand, a too long pilot length reduces the efficiency of the system, leaving no resources for the data transmission. In addition, the channel measurements are outdated before they can be applied. With an upper bound to the sum rate of a system with interference coordination and a sub-optimal pilot allocation strategy, we discuss the pilot length trade-off.