We present a model for simple earth mover robots modifying continuous 2D granular structures and propose a multi-robot planner which can modify structures according to user-defined goals. The proposed planner decomposes a long-horizon construction task into subtasks that are easier to solve using optimal transport theory and Wasserstein geodesics. We test the algorithm with 100 randomly generated construction tasks and demonstrate the algorithm with both single and many robots. The simulations show that the multi-robot algorithm can achieve \(85\%\) construction progress on average even under the presence of action noise and motion constraints. In addition, the deployment of multiple robots can reduce the overall traveling distance by \(70\%\) compared with the single-robot case, which is important as motion also affects the structure.