This paper addresses the multi-robot planning problem to generate trajectories for maximum information gathering from an area of interest. To solve this problem, the proposed framework first leverages a Gaussian Mixture Model (GMM) as prior knowledge for modeling informative regions in the target area. Taking samples from the GMM, informative robot paths are then computed that optimize the travel costs subject to energy budgets. Decomposing these paths into waypoints, robot trajectories are planned while respecting kinematic constraints and subsequently replanned online to avoid collisions among themselves. The GMM model is incrementally updated by incorporating the gathered information. We demonstrate that our framework can achieve a significant amount of information gain with the optimal travel distance. We also provide a realistic simulation with a team of mobile robots in a port infrastructure monitoring setting.