Text this: DETEAMSK: A Model-Based Reinforcement Learning Approach to Intelligent Top-Level Planning and Decisions for Multi-Drone Ad Hoc Teamwork by Decoupling the Identification of Teammate and Task