Adaptive Long-range UAV Flight Planning using Monte Carlo Search Trees

Overview & Synopsis

Long-range fixed-wing UAV missions require energy-efficient, dynamically feasible trajectories that can adapt online to evolving wind and airspace constraints. We present a Monte Carlo Tree Search (MCTS)-based framework that addresses this challenge by formulating mission planning as a finite-horizon Markov decision process incorporating aircraft kinematics, onboard energy dynamics, forecast wind fields, terrain, and time-varying no-fly zones. The MCTS planner searches over waypoint references generated by pre-stabilizing feedback controllers and a constraint-aware navigation field, yielding dynamically feasible trajectories. Simulation results from medical delivery scenarios show that the MCTS planner outperforms baseline geometric planners and population-based optimization algorithms. Moreover, the MCTS planner can be used in closed-loop to rapidly compute safe diversion trajectories under evolving constraints.