Coordinating multiple quadrotors through constrained passages is a difficult planning problem: it requires deciding not only where each vehicle flies but also when it passes through key points, all while respecting dynamics and actuation limits. The authors frame this as a joint optimization that couples temporal allocation with nonlinear motion, making it computationally challenging.
Their proposed method learns waypoints and passage events directly, rather than solving the full coupled optimization from scratch. This learning-based approach is intended for decentralized multi-quadrotor systems, meaning each vehicle can plan its own trajectory using shared or predicted information. The abstract does not provide experimental results, so the reported contribution is the formulation and the learning-based planning strategy itself.