This letter investigates the computation of information measures for distributed communication-aware information gathering by robotic sensor networks. The mutual information between an unknown target state and measurements received over a lossy network is derived in order to combine sensing and communication into a single objective function that can be optimized by the robot sensor network. Communication is modeled as a packet erasure channel, leading to a formulation of the mutual information that scales exponentially in the number of agents and time horizon. Several approximations to reduce this computational complexity are introduced and evaluated. Further, sample-based methods are introduced in order to derive approximate solutions of the mutual information when the system is nonlinear and uncertainties have non-Gaussian distributions. Finally, the information measure is evaluated through simulation experiments of multiple aerial robots tracking stationary radio emitters.